diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,120456 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 20000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.0000000000000004e-08,
+      "loss": 2.1196,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 2.1101,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 2.1226,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 2.1191,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-07,
+      "loss": 2.1353,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 2.1187,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5000000000000004e-07,
+      "loss": 2.1528,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 2.1257,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-07,
+      "loss": 2.0833,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-07,
+      "loss": 2.0664,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.5e-07,
+      "loss": 2.0781,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 2.0918,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5e-07,
+      "loss": 2.1162,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 2.1326,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-07,
+      "loss": 2.0793,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 2.0891,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.5e-07,
+      "loss": 2.1221,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-07,
+      "loss": 2.1155,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.500000000000001e-07,
+      "loss": 2.0991,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-06,
+      "loss": 2.1252,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 2.1335,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1e-06,
+      "loss": 2.1018,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.15e-06,
+      "loss": 2.1138,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.1228,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2499999999999999e-06,
+      "loss": 2.1113,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3e-06,
+      "loss": 2.105,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.35e-06,
+      "loss": 2.1169,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.062,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.45e-06,
+      "loss": 2.094,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-06,
+      "loss": 2.0876,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.55e-06,
+      "loss": 2.1489,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 2.1162,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.65e-06,
+      "loss": 2.1643,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7e-06,
+      "loss": 2.0994,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.75e-06,
+      "loss": 2.1389,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-06,
+      "loss": 2.1467,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.85e-06,
+      "loss": 2.0889,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 2.0996,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.95e-06,
+      "loss": 2.0947,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-06,
+      "loss": 2.0933,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.05e-06,
+      "loss": 2.1208,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 2.1221,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.15e-06,
+      "loss": 2.1074,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2e-06,
+      "loss": 2.0906,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.25e-06,
+      "loss": 2.0896,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3e-06,
+      "loss": 2.1101,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3500000000000004e-06,
+      "loss": 2.1482,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 2.1055,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.45e-06,
+      "loss": 2.0574,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4999999999999998e-06,
+      "loss": 2.0652,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.55e-06,
+      "loss": 2.0972,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6e-06,
+      "loss": 2.071,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.65e-06,
+      "loss": 2.0737,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7e-06,
+      "loss": 2.0901,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.75e-06,
+      "loss": 2.0935,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 2.0706,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8500000000000002e-06,
+      "loss": 2.1323,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9e-06,
+      "loss": 2.1135,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9499999999999997e-06,
+      "loss": 2.1206,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-06,
+      "loss": 2.1309,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.05e-06,
+      "loss": 2.0986,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1e-06,
+      "loss": 2.1074,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.15e-06,
+      "loss": 2.0996,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 2.1101,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 2.0603,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3e-06,
+      "loss": 2.0632,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.35e-06,
+      "loss": 2.0974,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4e-06,
+      "loss": 2.1108,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4500000000000004e-06,
+      "loss": 2.0852,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5e-06,
+      "loss": 2.0952,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.55e-06,
+      "loss": 2.1042,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-06,
+      "loss": 2.1108,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.65e-06,
+      "loss": 2.1218,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.7e-06,
+      "loss": 2.0962,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.75e-06,
+      "loss": 2.1372,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 2.1294,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8499999999999996e-06,
+      "loss": 2.0886,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9e-06,
+      "loss": 2.1069,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9499999999999995e-06,
+      "loss": 2.073,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-06,
+      "loss": 2.0989,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.05e-06,
+      "loss": 2.0986,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.1e-06,
+      "loss": 2.1274,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.15e-06,
+      "loss": 2.0955,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 2.0693,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.25e-06,
+      "loss": 2.0991,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.3e-06,
+      "loss": 2.0793,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.35e-06,
+      "loss": 2.0906,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4e-06,
+      "loss": 2.0754,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.450000000000001e-06,
+      "loss": 2.1438,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-06,
+      "loss": 2.1208,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5500000000000005e-06,
+      "loss": 2.0864,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6e-06,
+      "loss": 2.1057,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.65e-06,
+      "loss": 2.0337,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.700000000000001e-06,
+      "loss": 2.0222,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.75e-06,
+      "loss": 2.1206,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 2.1318,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.849999999999999e-06,
+      "loss": 2.1648,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.9e-06,
+      "loss": 2.1621,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.95e-06,
+      "loss": 2.1047,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9999999999999996e-06,
+      "loss": 2.0867,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.05e-06,
+      "loss": 2.0747,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1e-06,
+      "loss": 2.0903,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.15e-06,
+      "loss": 2.0845,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2e-06,
+      "loss": 2.0776,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.25e-06,
+      "loss": 2.1196,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3e-06,
+      "loss": 2.136,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3500000000000004e-06,
+      "loss": 2.1138,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4e-06,
+      "loss": 2.0994,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.45e-06,
+      "loss": 2.0811,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.5e-06,
+      "loss": 2.1133,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.55e-06,
+      "loss": 2.0544,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 2.0916,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.65e-06,
+      "loss": 2.0945,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7000000000000005e-06,
+      "loss": 2.1162,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.750000000000001e-06,
+      "loss": 2.0642,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8e-06,
+      "loss": 2.0879,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.850000000000001e-06,
+      "loss": 2.1284,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.899999999999999e-06,
+      "loss": 2.1433,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.95e-06,
+      "loss": 2.093,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6e-06,
+      "loss": 2.084,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.05e-06,
+      "loss": 2.0806,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.1e-06,
+      "loss": 2.1101,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.1499999999999996e-06,
+      "loss": 2.0667,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2e-06,
+      "loss": 2.0588,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.25e-06,
+      "loss": 2.0764,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3e-06,
+      "loss": 2.095,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.35e-06,
+      "loss": 2.0955,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 2.1072,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.45e-06,
+      "loss": 2.0999,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 2.1008,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.55e-06,
+      "loss": 2.0862,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.6e-06,
+      "loss": 2.0635,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.650000000000001e-06,
+      "loss": 2.0869,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.7e-06,
+      "loss": 2.0942,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.750000000000001e-06,
+      "loss": 2.082,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.8e-06,
+      "loss": 2.0693,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.8500000000000005e-06,
+      "loss": 2.0632,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.900000000000001e-06,
+      "loss": 2.0784,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.95e-06,
+      "loss": 2.0779,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7e-06,
+      "loss": 2.0872,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.049999999999999e-06,
+      "loss": 2.0933,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1e-06,
+      "loss": 2.1216,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.15e-06,
+      "loss": 2.0522,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2e-06,
+      "loss": 2.0708,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.25e-06,
+      "loss": 2.0959,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.3e-06,
+      "loss": 2.1025,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.35e-06,
+      "loss": 2.0872,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.4e-06,
+      "loss": 2.0857,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.45e-06,
+      "loss": 2.0613,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5e-06,
+      "loss": 2.0862,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.55e-06,
+      "loss": 2.0984,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 2.1057,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.65e-06,
+      "loss": 2.0864,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.699999999999999e-06,
+      "loss": 2.1238,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.75e-06,
+      "loss": 2.0913,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.8e-06,
+      "loss": 2.0728,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.85e-06,
+      "loss": 2.1221,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.899999999999999e-06,
+      "loss": 2.1108,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.95e-06,
+      "loss": 2.0642,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8e-06,
+      "loss": 2.0706,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.05e-06,
+      "loss": 2.1218,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.1e-06,
+      "loss": 2.1328,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.15e-06,
+      "loss": 2.0786,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.2e-06,
+      "loss": 2.0928,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.25e-06,
+      "loss": 2.054,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.3e-06,
+      "loss": 2.0852,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.35e-06,
+      "loss": 2.072,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 2.0728,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.45e-06,
+      "loss": 2.0786,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.5e-06,
+      "loss": 2.0784,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.55e-06,
+      "loss": 2.0923,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.6e-06,
+      "loss": 2.1055,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.65e-06,
+      "loss": 2.0732,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.7e-06,
+      "loss": 2.0671,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 2.0488,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.8e-06,
+      "loss": 2.0745,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.85e-06,
+      "loss": 2.1118,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 2.0903,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.95e-06,
+      "loss": 2.0581,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9e-06,
+      "loss": 2.0518,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.050000000000001e-06,
+      "loss": 2.0452,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.100000000000001e-06,
+      "loss": 2.0564,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.15e-06,
+      "loss": 2.0635,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.2e-06,
+      "loss": 2.0759,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 2.1018,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.3e-06,
+      "loss": 2.1392,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.35e-06,
+      "loss": 2.0947,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.400000000000001e-06,
+      "loss": 2.1267,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.450000000000001e-06,
+      "loss": 2.0996,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.5e-06,
+      "loss": 2.0994,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.550000000000002e-06,
+      "loss": 2.0869,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 2.0823,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.649999999999999e-06,
+      "loss": 2.0847,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.699999999999999e-06,
+      "loss": 2.0706,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.75e-06,
+      "loss": 2.0884,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.8e-06,
+      "loss": 2.0769,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.849999999999999e-06,
+      "loss": 2.0706,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9e-06,
+      "loss": 2.073,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.95e-06,
+      "loss": 2.0747,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999999999999999e-06,
+      "loss": 2.0781,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.005e-05,
+      "loss": 2.1013,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.01e-05,
+      "loss": 2.0867,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.015e-05,
+      "loss": 2.0979,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.02e-05,
+      "loss": 2.1345,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.025e-05,
+      "loss": 2.0742,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.03e-05,
+      "loss": 2.0791,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.035e-05,
+      "loss": 2.1206,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.04e-05,
+      "loss": 2.054,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.045e-05,
+      "loss": 2.1062,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.05e-05,
+      "loss": 2.0891,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.055e-05,
+      "loss": 2.0767,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.06e-05,
+      "loss": 2.0676,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.065e-05,
+      "loss": 2.0488,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0700000000000001e-05,
+      "loss": 2.0989,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.075e-05,
+      "loss": 2.0642,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.08e-05,
+      "loss": 2.0515,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0850000000000001e-05,
+      "loss": 2.0999,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.09e-05,
+      "loss": 2.1287,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.095e-05,
+      "loss": 2.0615,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1e-05,
+      "loss": 2.0532,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1050000000000001e-05,
+      "loss": 2.0759,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.11e-05,
+      "loss": 2.0554,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.115e-05,
+      "loss": 2.0784,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 2.0784,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.125e-05,
+      "loss": 2.0291,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.13e-05,
+      "loss": 2.0544,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1350000000000001e-05,
+      "loss": 2.1079,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1400000000000001e-05,
+      "loss": 2.1306,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.145e-05,
+      "loss": 2.0442,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 2.0422,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1550000000000001e-05,
+      "loss": 2.0825,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.16e-05,
+      "loss": 2.0955,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.165e-05,
+      "loss": 2.0891,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1700000000000001e-05,
+      "loss": 2.0759,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1750000000000001e-05,
+      "loss": 2.0564,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1799999999999999e-05,
+      "loss": 2.0671,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.185e-05,
+      "loss": 2.0754,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.19e-05,
+      "loss": 2.1252,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1949999999999999e-05,
+      "loss": 2.1077,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2e-05,
+      "loss": 2.0952,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.205e-05,
+      "loss": 2.072,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.21e-05,
+      "loss": 2.0542,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.215e-05,
+      "loss": 2.1052,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.22e-05,
+      "loss": 2.0713,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.225e-05,
+      "loss": 2.0806,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2299999999999999e-05,
+      "loss": 2.092,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.235e-05,
+      "loss": 2.124,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.24e-05,
+      "loss": 2.1216,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.245e-05,
+      "loss": 2.0835,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.25e-05,
+      "loss": 2.0879,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.255e-05,
+      "loss": 2.1062,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.26e-05,
+      "loss": 2.113,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2650000000000001e-05,
+      "loss": 2.0779,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.27e-05,
+      "loss": 2.0874,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.275e-05,
+      "loss": 2.075,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 2.0588,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.285e-05,
+      "loss": 2.1008,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.29e-05,
+      "loss": 2.1042,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.295e-05,
+      "loss": 2.0879,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 2.0586,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.305e-05,
+      "loss": 2.0396,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.31e-05,
+      "loss": 2.0481,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3150000000000001e-05,
+      "loss": 2.0869,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.32e-05,
+      "loss": 2.0972,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.325e-05,
+      "loss": 2.1008,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3300000000000001e-05,
+      "loss": 2.1006,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3350000000000001e-05,
+      "loss": 2.1121,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.34e-05,
+      "loss": 2.0847,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.345e-05,
+      "loss": 2.041,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 2.0271,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.355e-05,
+      "loss": 2.1213,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.36e-05,
+      "loss": 2.074,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3650000000000001e-05,
+      "loss": 2.0427,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3700000000000001e-05,
+      "loss": 2.0527,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.375e-05,
+      "loss": 2.0413,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3800000000000002e-05,
+      "loss": 2.0518,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3850000000000001e-05,
+      "loss": 2.0737,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.39e-05,
+      "loss": 2.0742,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.395e-05,
+      "loss": 2.1255,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4e-05,
+      "loss": 2.0889,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.405e-05,
+      "loss": 2.0752,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4099999999999999e-05,
+      "loss": 2.095,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.415e-05,
+      "loss": 2.1074,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.42e-05,
+      "loss": 2.074,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 2.0129,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.43e-05,
+      "loss": 2.0139,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.435e-05,
+      "loss": 2.0623,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.44e-05,
+      "loss": 2.0693,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.445e-05,
+      "loss": 2.1255,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.45e-05,
+      "loss": 2.1016,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.455e-05,
+      "loss": 2.063,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.46e-05,
+      "loss": 2.0742,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.465e-05,
+      "loss": 2.0713,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.47e-05,
+      "loss": 2.072,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.475e-05,
+      "loss": 2.0618,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.48e-05,
+      "loss": 2.0632,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.485e-05,
+      "loss": 2.1345,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.49e-05,
+      "loss": 2.1382,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4950000000000001e-05,
+      "loss": 2.0833,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5e-05,
+      "loss": 2.0869,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5050000000000002e-05,
+      "loss": 2.0706,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.51e-05,
+      "loss": 2.0847,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.515e-05,
+      "loss": 2.0906,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5200000000000002e-05,
+      "loss": 2.0437,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.525e-05,
+      "loss": 2.0784,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.53e-05,
+      "loss": 2.0767,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.535e-05,
+      "loss": 2.0991,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5399999999999998e-05,
+      "loss": 2.0657,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.545e-05,
+      "loss": 2.0803,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.55e-05,
+      "loss": 2.1003,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.555e-05,
+      "loss": 2.0867,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.56e-05,
+      "loss": 2.0996,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5649999999999998e-05,
+      "loss": 2.1313,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.57e-05,
+      "loss": 2.1401,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.575e-05,
+      "loss": 2.0728,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5799999999999998e-05,
+      "loss": 2.0894,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.585e-05,
+      "loss": 2.074,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.59e-05,
+      "loss": 2.0518,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5949999999999998e-05,
+      "loss": 2.0623,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6e-05,
+      "loss": 2.0764,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.605e-05,
+      "loss": 2.0869,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.61e-05,
+      "loss": 2.0962,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.615e-05,
+      "loss": 2.0872,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.62e-05,
+      "loss": 2.1042,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.625e-05,
+      "loss": 2.0938,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.63e-05,
+      "loss": 2.1199,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.635e-05,
+      "loss": 2.0764,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.64e-05,
+      "loss": 2.104,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.645e-05,
+      "loss": 2.0337,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.65e-05,
+      "loss": 2.0298,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.655e-05,
+      "loss": 2.0959,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.66e-05,
+      "loss": 2.0906,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.665e-05,
+      "loss": 2.115,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.67e-05,
+      "loss": 2.0986,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.675e-05,
+      "loss": 2.0474,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 2.0786,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.685e-05,
+      "loss": 2.1113,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.69e-05,
+      "loss": 2.103,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.695e-05,
+      "loss": 2.052,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7e-05,
+      "loss": 2.0186,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.705e-05,
+      "loss": 2.1267,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.71e-05,
+      "loss": 2.095,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.715e-05,
+      "loss": 2.0796,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.72e-05,
+      "loss": 2.0774,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.725e-05,
+      "loss": 2.1028,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.73e-05,
+      "loss": 2.0784,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.735e-05,
+      "loss": 2.0361,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.74e-05,
+      "loss": 2.0415,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.745e-05,
+      "loss": 2.0588,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 2.0505,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.755e-05,
+      "loss": 2.0671,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.76e-05,
+      "loss": 2.0618,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7650000000000002e-05,
+      "loss": 2.085,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.77e-05,
+      "loss": 2.0649,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.775e-05,
+      "loss": 2.0872,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7800000000000002e-05,
+      "loss": 2.0737,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.785e-05,
+      "loss": 2.0669,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.79e-05,
+      "loss": 2.0676,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7950000000000003e-05,
+      "loss": 2.0244,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8e-05,
+      "loss": 2.0684,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.805e-05,
+      "loss": 2.1057,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8100000000000003e-05,
+      "loss": 2.093,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.815e-05,
+      "loss": 2.0867,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8200000000000002e-05,
+      "loss": 2.0881,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.825e-05,
+      "loss": 2.0291,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.83e-05,
+      "loss": 2.0688,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8350000000000002e-05,
+      "loss": 2.084,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.84e-05,
+      "loss": 2.0847,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.845e-05,
+      "loss": 2.0596,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8500000000000002e-05,
+      "loss": 2.0359,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.855e-05,
+      "loss": 2.0178,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.86e-05,
+      "loss": 2.0415,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8650000000000003e-05,
+      "loss": 2.0508,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.87e-05,
+      "loss": 2.0378,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 2.0957,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8800000000000003e-05,
+      "loss": 2.0759,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.885e-05,
+      "loss": 2.0862,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8900000000000002e-05,
+      "loss": 2.0742,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8950000000000003e-05,
+      "loss": 2.0818,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9e-05,
+      "loss": 2.0698,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9050000000000002e-05,
+      "loss": 2.0576,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9100000000000003e-05,
+      "loss": 2.0217,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.915e-05,
+      "loss": 2.0674,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 2.0825,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9250000000000004e-05,
+      "loss": 2.1169,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9299999999999998e-05,
+      "loss": 2.1506,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.935e-05,
+      "loss": 2.1296,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9399999999999997e-05,
+      "loss": 2.0847,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.945e-05,
+      "loss": 2.0618,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.95e-05,
+      "loss": 2.0562,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9549999999999997e-05,
+      "loss": 2.0234,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.96e-05,
+      "loss": 2.0278,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.965e-05,
+      "loss": 2.0522,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9699999999999998e-05,
+      "loss": 2.0991,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.975e-05,
+      "loss": 2.0833,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98e-05,
+      "loss": 2.0537,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9849999999999998e-05,
+      "loss": 2.094,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99e-05,
+      "loss": 2.1125,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.995e-05,
+      "loss": 2.0752,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9999999999999998e-05,
+      "loss": 2.074,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.005e-05,
+      "loss": 2.0872,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.01e-05,
+      "loss": 2.0251,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.015e-05,
+      "loss": 2.0513,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.02e-05,
+      "loss": 2.071,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.025e-05,
+      "loss": 2.0579,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.03e-05,
+      "loss": 2.0713,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.035e-05,
+      "loss": 2.0339,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.04e-05,
+      "loss": 2.0637,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.045e-05,
+      "loss": 2.0754,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.05e-05,
+      "loss": 2.0781,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.055e-05,
+      "loss": 2.075,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.06e-05,
+      "loss": 2.0798,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.065e-05,
+      "loss": 2.0293,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.07e-05,
+      "loss": 2.0544,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.075e-05,
+      "loss": 2.0439,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.08e-05,
+      "loss": 2.0476,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.085e-05,
+      "loss": 2.0718,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.09e-05,
+      "loss": 2.0371,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.095e-05,
+      "loss": 2.0693,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1e-05,
+      "loss": 2.0569,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.105e-05,
+      "loss": 2.0605,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.11e-05,
+      "loss": 2.0811,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.115e-05,
+      "loss": 2.0886,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.12e-05,
+      "loss": 2.0662,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.125e-05,
+      "loss": 2.0989,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.13e-05,
+      "loss": 2.0625,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.135e-05,
+      "loss": 2.0583,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1400000000000002e-05,
+      "loss": 2.0635,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.145e-05,
+      "loss": 2.1292,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.15e-05,
+      "loss": 2.1467,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1550000000000002e-05,
+      "loss": 2.02,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.16e-05,
+      "loss": 2.0327,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.165e-05,
+      "loss": 2.031,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1700000000000002e-05,
+      "loss": 2.0481,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.175e-05,
+      "loss": 2.105,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.18e-05,
+      "loss": 2.0227,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1850000000000003e-05,
+      "loss": 2.0361,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.19e-05,
+      "loss": 2.0435,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.195e-05,
+      "loss": 2.0383,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2e-05,
+      "loss": 2.0188,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.205e-05,
+      "loss": 2.0806,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2100000000000002e-05,
+      "loss": 2.0571,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.215e-05,
+      "loss": 2.0496,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.22e-05,
+      "loss": 2.0591,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2250000000000002e-05,
+      "loss": 2.0784,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.23e-05,
+      "loss": 2.1008,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.235e-05,
+      "loss": 2.041,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2400000000000002e-05,
+      "loss": 2.0127,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.245e-05,
+      "loss": 2.1008,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.25e-05,
+      "loss": 2.0796,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2550000000000003e-05,
+      "loss": 2.0879,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.26e-05,
+      "loss": 2.0466,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.265e-05,
+      "loss": 2.0754,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2700000000000003e-05,
+      "loss": 2.0696,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.275e-05,
+      "loss": 2.078,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 2.092,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2850000000000003e-05,
+      "loss": 2.1138,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.29e-05,
+      "loss": 2.1309,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2950000000000002e-05,
+      "loss": 2.1089,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 2.0881,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.305e-05,
+      "loss": 2.0522,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3100000000000002e-05,
+      "loss": 2.0483,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.315e-05,
+      "loss": 2.032,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.32e-05,
+      "loss": 2.072,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3250000000000003e-05,
+      "loss": 2.0464,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.33e-05,
+      "loss": 2.061,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3350000000000002e-05,
+      "loss": 2.0935,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3400000000000003e-05,
+      "loss": 2.0933,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.345e-05,
+      "loss": 2.0842,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3500000000000002e-05,
+      "loss": 2.1001,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3550000000000003e-05,
+      "loss": 2.0349,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3599999999999998e-05,
+      "loss": 2.0437,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.365e-05,
+      "loss": 2.1145,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.37e-05,
+      "loss": 2.1006,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3749999999999998e-05,
+      "loss": 2.0081,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.38e-05,
+      "loss": 2.0623,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.385e-05,
+      "loss": 2.0613,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3899999999999998e-05,
+      "loss": 2.0818,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.395e-05,
+      "loss": 2.1162,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4e-05,
+      "loss": 2.1316,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.405e-05,
+      "loss": 2.1147,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.41e-05,
+      "loss": 2.0957,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.415e-05,
+      "loss": 2.0447,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.42e-05,
+      "loss": 2.0366,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.425e-05,
+      "loss": 2.0879,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.43e-05,
+      "loss": 2.0737,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.435e-05,
+      "loss": 2.0576,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.44e-05,
+      "loss": 2.0334,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4449999999999998e-05,
+      "loss": 2.0876,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.45e-05,
+      "loss": 2.0745,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.455e-05,
+      "loss": 2.0906,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 2.0894,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.465e-05,
+      "loss": 2.0786,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.47e-05,
+      "loss": 2.0623,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.475e-05,
+      "loss": 2.0891,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.48e-05,
+      "loss": 2.0449,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.485e-05,
+      "loss": 2.0928,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.49e-05,
+      "loss": 2.083,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.495e-05,
+      "loss": 2.0298,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.5e-05,
+      "loss": 2.0398,
+      "step": 500
+    },
+    {
+      "Bleu_1": 0.7705747287055579,
+      "Bleu_2": 0.6094714575847039,
+      "Bleu_3": 0.4680369762510702,
+      "Bleu_4": 0.35550558448219516,
+      "CIDEr": 1.195837021300063,
+      "METEOR": 0.27824173472618197,
+      "ROUGE_L": 0.5561148690287793,
+      "epoch": 0.03,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.505e-05,
+      "loss": 2.0627,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.51e-05,
+      "loss": 2.0774,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.515e-05,
+      "loss": 2.0659,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.52e-05,
+      "loss": 2.083,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.525e-05,
+      "loss": 2.0632,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.5300000000000002e-05,
+      "loss": 2.0537,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.535e-05,
+      "loss": 2.0847,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.54e-05,
+      "loss": 2.0674,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.5450000000000002e-05,
+      "loss": 2.0503,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.55e-05,
+      "loss": 2.075,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.555e-05,
+      "loss": 2.0642,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.5600000000000002e-05,
+      "loss": 2.0718,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.565e-05,
+      "loss": 2.1001,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.57e-05,
+      "loss": 2.1079,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.575e-05,
+      "loss": 2.0361,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.58e-05,
+      "loss": 2.0518,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.585e-05,
+      "loss": 2.0737,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.59e-05,
+      "loss": 2.0842,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.595e-05,
+      "loss": 2.0327,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 2.0254,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.605e-05,
+      "loss": 2.0884,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.61e-05,
+      "loss": 2.0784,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6150000000000002e-05,
+      "loss": 2.1096,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.62e-05,
+      "loss": 2.1238,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.625e-05,
+      "loss": 2.0818,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6300000000000002e-05,
+      "loss": 2.1074,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.635e-05,
+      "loss": 1.9722,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.64e-05,
+      "loss": 1.9863,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6450000000000003e-05,
+      "loss": 2.0854,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.65e-05,
+      "loss": 2.0083,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.655e-05,
+      "loss": 2.0908,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6600000000000003e-05,
+      "loss": 2.0811,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.665e-05,
+      "loss": 2.1035,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6700000000000002e-05,
+      "loss": 2.0894,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6750000000000003e-05,
+      "loss": 2.1174,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.68e-05,
+      "loss": 2.125,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6850000000000002e-05,
+      "loss": 2.0349,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.69e-05,
+      "loss": 2.0527,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.695e-05,
+      "loss": 2.0889,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 2.0884,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.705e-05,
+      "loss": 2.1084,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.71e-05,
+      "loss": 2.1042,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7150000000000003e-05,
+      "loss": 2.0645,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.72e-05,
+      "loss": 2.0544,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.725e-05,
+      "loss": 2.0386,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7300000000000003e-05,
+      "loss": 2.0571,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.735e-05,
+      "loss": 2.126,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7400000000000002e-05,
+      "loss": 2.0801,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7450000000000003e-05,
+      "loss": 2.0652,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.75e-05,
+      "loss": 2.0608,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7550000000000002e-05,
+      "loss": 2.0271,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7600000000000003e-05,
+      "loss": 2.0073,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.765e-05,
+      "loss": 2.0945,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7700000000000002e-05,
+      "loss": 2.092,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7750000000000004e-05,
+      "loss": 2.0859,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.78e-05,
+      "loss": 2.0962,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7850000000000003e-05,
+      "loss": 2.0728,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.79e-05,
+      "loss": 2.0535,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.795e-05,
+      "loss": 2.061,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8e-05,
+      "loss": 2.054,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.805e-05,
+      "loss": 2.0879,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.81e-05,
+      "loss": 2.0874,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.815e-05,
+      "loss": 2.0747,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 2.0972,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.825e-05,
+      "loss": 2.0791,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.83e-05,
+      "loss": 2.0566,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8349999999999998e-05,
+      "loss": 2.0461,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.84e-05,
+      "loss": 2.0618,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.845e-05,
+      "loss": 2.0642,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 2.0598,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.855e-05,
+      "loss": 2.0999,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.86e-05,
+      "loss": 2.0876,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.865e-05,
+      "loss": 2.0681,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.87e-05,
+      "loss": 2.0913,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.875e-05,
+      "loss": 2.0654,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.88e-05,
+      "loss": 2.053,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.885e-05,
+      "loss": 2.1062,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.89e-05,
+      "loss": 2.1023,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.895e-05,
+      "loss": 2.0967,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9e-05,
+      "loss": 2.0891,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.905e-05,
+      "loss": 2.0366,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.91e-05,
+      "loss": 2.0239,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.915e-05,
+      "loss": 2.1165,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.92e-05,
+      "loss": 2.1287,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.925e-05,
+      "loss": 2.0637,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.93e-05,
+      "loss": 2.0881,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9350000000000002e-05,
+      "loss": 2.0547,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.94e-05,
+      "loss": 2.0911,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.945e-05,
+      "loss": 2.0786,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.95e-05,
+      "loss": 2.1018,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.955e-05,
+      "loss": 2.1206,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.96e-05,
+      "loss": 2.1233,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.965e-05,
+      "loss": 2.0415,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.97e-05,
+      "loss": 2.0156,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.975e-05,
+      "loss": 2.0393,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.98e-05,
+      "loss": 2.0632,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.985e-05,
+      "loss": 2.0408,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9900000000000002e-05,
+      "loss": 2.0547,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.995e-05,
+      "loss": 2.0842,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3e-05,
+      "loss": 2.0735,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999999981315514e-05,
+      "loss": 2.0576,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999999252620568e-05,
+      "loss": 2.0479,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.99999983183963e-05,
+      "loss": 2.0945,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999999701048235e-05,
+      "loss": 2.0779,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999995328878764e-05,
+      "loss": 2.105,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999993273585586e-05,
+      "loss": 2.0815,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999999084460286e-05,
+      "loss": 2.0498,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999998804193066e-05,
+      "loss": 1.9861,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999998486556905e-05,
+      "loss": 2.1111,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999981315518125e-05,
+      "loss": 2.1104,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999977391777967e-05,
+      "loss": 2.0942,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999973094348686e-05,
+      "loss": 2.0928,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999968423230393e-05,
+      "loss": 2.1028,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999963378423207e-05,
+      "loss": 2.0942,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999995795992726e-05,
+      "loss": 2.0286,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999952167742704e-05,
+      "loss": 2.0127,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999994600186968e-05,
+      "loss": 2.0864,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999939462308354e-05,
+      "loss": 2.0876,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999932549058898e-05,
+      "loss": 2.0637,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999992526212149e-05,
+      "loss": 2.0874,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999917601496328e-05,
+      "loss": 2.1326,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999909567183604e-05,
+      "loss": 2.0896,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999901159183535e-05,
+      "loss": 2.0979,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999892377496346e-05,
+      "loss": 2.0974,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999988322212225e-05,
+      "loss": 2.0503,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999873693061507e-05,
+      "loss": 2.0317,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999986379031435e-05,
+      "loss": 2.0381,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999985351388105e-05,
+      "loss": 2.0381,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999842863761874e-05,
+      "loss": 2.0718,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999831839957097e-05,
+      "loss": 2.0576,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999820442467018e-05,
+      "loss": 2.0771,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999808671291922e-05,
+      "loss": 2.0244,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999796526432128e-05,
+      "loss": 2.0679,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999978400788795e-05,
+      "loss": 2.0359,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999771115659724e-05,
+      "loss": 2.1338,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999975784974778e-05,
+      "loss": 2.0955,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999744210152463e-05,
+      "loss": 2.1323,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999730196874135e-05,
+      "loss": 2.1321,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999971580991317e-05,
+      "loss": 2.0613,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999701049269934e-05,
+      "loss": 2.0479,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999968591494482e-05,
+      "loss": 2.0833,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999670406938224e-05,
+      "loss": 2.0535,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999654525250556e-05,
+      "loss": 2.0576,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999638269882228e-05,
+      "loss": 2.0491,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999621640833663e-05,
+      "loss": 2.0198,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999604638105306e-05,
+      "loss": 2.033,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.99995872616976e-05,
+      "loss": 2.0171,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999569511611e-05,
+      "loss": 2.064,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999551387845964e-05,
+      "loss": 2.0708,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999953289040298e-05,
+      "loss": 2.0435,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999514019282523e-05,
+      "loss": 2.0344,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999494774485094e-05,
+      "loss": 2.0298,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999947515601119e-05,
+      "loss": 2.0339,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999455163861342e-05,
+      "loss": 2.0103,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999434798036054e-05,
+      "loss": 2.0444,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999414058535876e-05,
+      "loss": 2.0583,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999392945361344e-05,
+      "loss": 2.0613,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999937145851301e-05,
+      "loss": 2.0195,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999349597991444e-05,
+      "loss": 2.0623,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999932736379721e-05,
+      "loss": 1.9885,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999304755930898e-05,
+      "loss": 2.0803,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999281774393107e-05,
+      "loss": 2.0483,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999258419184426e-05,
+      "loss": 2.0583,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999923469030547e-05,
+      "loss": 2.0417,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999210587756873e-05,
+      "loss": 2.0962,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999186111539254e-05,
+      "loss": 2.0999,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999916126165326e-05,
+      "loss": 2.0991,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999136038099546e-05,
+      "loss": 2.0764,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999110440878765e-05,
+      "loss": 2.0876,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999084469991598e-05,
+      "loss": 2.0862,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999905812543872e-05,
+      "loss": 2.0408,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9999031407220824e-05,
+      "loss": 2.0522,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999900431533861e-05,
+      "loss": 2.0415,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998976849792783e-05,
+      "loss": 2.0754,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998949010584076e-05,
+      "loss": 2.075,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998920797713207e-05,
+      "loss": 2.0454,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999889221118092e-05,
+      "loss": 2.0686,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998863250987965e-05,
+      "loss": 2.075,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998833917135107e-05,
+      "loss": 2.1155,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998804209623103e-05,
+      "loss": 2.0857,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998774128452743e-05,
+      "loss": 2.0105,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999874367362481e-05,
+      "loss": 2.0461,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998712845140104e-05,
+      "loss": 2.1267,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998681642999437e-05,
+      "loss": 2.1096,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998650067203622e-05,
+      "loss": 2.0188,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999861811775349e-05,
+      "loss": 2.0222,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998585794649875e-05,
+      "loss": 2.1514,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998553097893626e-05,
+      "loss": 2.1169,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998520027485608e-05,
+      "loss": 2.1011,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998486583426678e-05,
+      "loss": 2.1116,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.999845276571772e-05,
+      "loss": 2.0759,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998418574359614e-05,
+      "loss": 2.0803,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998384009353263e-05,
+      "loss": 2.0459,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998349070699575e-05,
+      "loss": 2.0471,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998313758399456e-05,
+      "loss": 2.0994,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998278072453843e-05,
+      "loss": 2.0598,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998242012863664e-05,
+      "loss": 2.0527,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998205579629867e-05,
+      "loss": 2.0752,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9998168772753412e-05,
+      "loss": 2.0815,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999813159223526e-05,
+      "loss": 2.0747,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9998094038076384e-05,
+      "loss": 2.0576,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999805611027777e-05,
+      "loss": 2.0415,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999801780884042e-05,
+      "loss": 2.0681,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999797913376533e-05,
+      "loss": 2.0566,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997940085053515e-05,
+      "loss": 2.0425,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997900662705997e-05,
+      "loss": 2.0256,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997860866723817e-05,
+      "loss": 2.1021,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997820697108015e-05,
+      "loss": 2.0881,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997780153859643e-05,
+      "loss": 2.0569,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997739236979767e-05,
+      "loss": 2.084,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999769794646946e-05,
+      "loss": 2.0569,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997656282329802e-05,
+      "loss": 2.0635,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997614244561882e-05,
+      "loss": 2.0955,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997571833166814e-05,
+      "loss": 2.062,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99975290481457e-05,
+      "loss": 2.1001,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999748588949967e-05,
+      "loss": 2.0571,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997442357229847e-05,
+      "loss": 2.0828,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997398451337375e-05,
+      "loss": 2.0918,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999735417182341e-05,
+      "loss": 2.0522,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999730951868911e-05,
+      "loss": 2.0271,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997264491935648e-05,
+      "loss": 2.043,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99972190915642e-05,
+      "loss": 2.0212,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997173317575963e-05,
+      "loss": 2.0706,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997127169972135e-05,
+      "loss": 2.0586,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997080648753918e-05,
+      "loss": 2.03,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9997033753922543e-05,
+      "loss": 2.0315,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996986485479237e-05,
+      "loss": 2.0972,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996938843425242e-05,
+      "loss": 2.0901,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99968908277618e-05,
+      "loss": 2.1121,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996842438490175e-05,
+      "loss": 2.085,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996793675611634e-05,
+      "loss": 2.0186,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999674453912746e-05,
+      "loss": 2.0278,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996695029038932e-05,
+      "loss": 2.1006,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999664514534736e-05,
+      "loss": 2.123,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996594888054045e-05,
+      "loss": 2.0718,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996544257160308e-05,
+      "loss": 2.0632,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996493252667472e-05,
+      "loss": 2.0747,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996441874576878e-05,
+      "loss": 2.0596,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996390122889877e-05,
+      "loss": 2.0938,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999633799760782e-05,
+      "loss": 2.1018,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996285498732076e-05,
+      "loss": 2.0713,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996232626264025e-05,
+      "loss": 2.0596,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999617938020505e-05,
+      "loss": 2.1206,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996125760556546e-05,
+      "loss": 2.1477,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996071767319923e-05,
+      "loss": 2.1106,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9996017400496593e-05,
+      "loss": 2.1177,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995962660087987e-05,
+      "loss": 2.0581,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995907546095535e-05,
+      "loss": 2.0598,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995852058520687e-05,
+      "loss": 2.0732,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999579619736489e-05,
+      "loss": 2.0547,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999573996262962e-05,
+      "loss": 2.0369,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999568335431634e-05,
+      "loss": 2.0449,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995626372426545e-05,
+      "loss": 2.1106,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995569016961723e-05,
+      "loss": 2.0818,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995511287923383e-05,
+      "loss": 2.0835,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995453185313034e-05,
+      "loss": 2.0759,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99953947091322e-05,
+      "loss": 2.092,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999533585938242e-05,
+      "loss": 2.075,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999527663606523e-05,
+      "loss": 2.0671,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999521703918219e-05,
+      "loss": 2.0688,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995157068734853e-05,
+      "loss": 2.0991,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995096724724804e-05,
+      "loss": 2.0679,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9995036007153615e-05,
+      "loss": 2.0437,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994974916022885e-05,
+      "loss": 2.0195,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994913451334213e-05,
+      "loss": 2.0701,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994851613089212e-05,
+      "loss": 2.0562,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994789401289505e-05,
+      "loss": 2.0518,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994726815936718e-05,
+      "loss": 2.0613,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994663857032498e-05,
+      "loss": 2.0535,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999460052457849e-05,
+      "loss": 2.0674,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999453681857636e-05,
+      "loss": 2.0791,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999447273902778e-05,
+      "loss": 2.0427,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994408285934427e-05,
+      "loss": 2.011,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999434345929799e-05,
+      "loss": 2.0078,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9994278259120174e-05,
+      "loss": 2.0422,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999421268540268e-05,
+      "loss": 2.054,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999414673814724e-05,
+      "loss": 2.0894,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999408041735557e-05,
+      "loss": 2.0769,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999401372302942e-05,
+      "loss": 2.0769,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993946655170535e-05,
+      "loss": 2.0845,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993879213780674e-05,
+      "loss": 2.0894,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993811398861602e-05,
+      "loss": 2.0613,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99937432104151e-05,
+      "loss": 2.1006,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999367464844296e-05,
+      "loss": 2.0728,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999360571294698e-05,
+      "loss": 2.0872,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993536403928957e-05,
+      "loss": 2.0513,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993466721390718e-05,
+      "loss": 1.9998,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993396665334086e-05,
+      "loss": 1.9956,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999332623576091e-05,
+      "loss": 2.085,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993255432673015e-05,
+      "loss": 2.0903,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999318425607228e-05,
+      "loss": 2.0601,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993112705960554e-05,
+      "loss": 2.0688,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9993040782339727e-05,
+      "loss": 2.0862,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992968485211675e-05,
+      "loss": 2.0613,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99928958145783e-05,
+      "loss": 2.019,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992822770441506e-05,
+      "loss": 2.0125,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992749352803208e-05,
+      "loss": 2.0908,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999267556166533e-05,
+      "loss": 2.0398,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999260139702981e-05,
+      "loss": 2.0669,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999252685889859e-05,
+      "loss": 2.0388,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992451947273622e-05,
+      "loss": 2.0715,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992376662156882e-05,
+      "loss": 2.0786,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992301003550332e-05,
+      "loss": 2.1182,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9992224971455964e-05,
+      "loss": 2.0972,
+      "step": 804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999214856587577e-05,
+      "loss": 2.0627,
+      "step": 805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999207178681175e-05,
+      "loss": 2.0605,
+      "step": 806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999199463426592e-05,
+      "loss": 2.1221,
+      "step": 807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99919171082403e-05,
+      "loss": 2.0781,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999183920873693e-05,
+      "loss": 2.1123,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991760935757846e-05,
+      "loss": 2.1116,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999168228930511e-05,
+      "loss": 2.0823,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999160326938077e-05,
+      "loss": 2.0745,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999152387598691e-05,
+      "loss": 2.1289,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991444109125605e-05,
+      "loss": 2.1143,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991363968798956e-05,
+      "loss": 2.1245,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991283455009052e-05,
+      "loss": 2.115,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991202567758016e-05,
+      "loss": 2.0488,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991121307047962e-05,
+      "loss": 2.0562,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9991039672881023e-05,
+      "loss": 2.085,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999095766525934e-05,
+      "loss": 2.0784,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990875284185062e-05,
+      "loss": 2.0549,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990792529660353e-05,
+      "loss": 2.0674,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.999070940168738e-05,
+      "loss": 2.042,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990625900268323e-05,
+      "loss": 2.0564,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990542025405373e-05,
+      "loss": 2.0833,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990457777100727e-05,
+      "loss": 2.0273,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990373155356603e-05,
+      "loss": 2.0964,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990288160175208e-05,
+      "loss": 2.0798,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990202791558772e-05,
+      "loss": 2.0811,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990117049509546e-05,
+      "loss": 2.05,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9990030934029768e-05,
+      "loss": 2.0837,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989944445121698e-05,
+      "loss": 2.0933,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989857582787604e-05,
+      "loss": 2.0962,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989770347029768e-05,
+      "loss": 2.0818,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989682737850474e-05,
+      "loss": 2.1101,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989594755252017e-05,
+      "loss": 2.1086,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998950639923671e-05,
+      "loss": 2.1165,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989417669806866e-05,
+      "loss": 2.1147,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998932856696481e-05,
+      "loss": 2.0811,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989239090712882e-05,
+      "loss": 2.054,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9989149241053435e-05,
+      "loss": 2.1333,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998905901798881e-05,
+      "loss": 2.0979,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998896842152139e-05,
+      "loss": 2.0754,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988877451653533e-05,
+      "loss": 2.1077,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998878610838764e-05,
+      "loss": 2.0659,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988694391726094e-05,
+      "loss": 2.0754,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988602301671314e-05,
+      "loss": 2.1201,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99885098382257e-05,
+      "loss": 2.0894,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988417001391684e-05,
+      "loss": 2.0815,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988323791171702e-05,
+      "loss": 2.0627,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988230207568196e-05,
+      "loss": 2.031,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988136250583625e-05,
+      "loss": 2.0625,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9988041920220443e-05,
+      "loss": 2.0984,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987947216481133e-05,
+      "loss": 2.0645,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998785213936818e-05,
+      "loss": 2.0393,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998775668888406e-05,
+      "loss": 2.0342,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987660865031294e-05,
+      "loss": 2.0654,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998756466781239e-05,
+      "loss": 2.064,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987468097229868e-05,
+      "loss": 2.0769,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987371153286263e-05,
+      "loss": 2.0513,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987273835984117e-05,
+      "loss": 1.9878,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987176145325982e-05,
+      "loss": 1.9963,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9987078081314417e-05,
+      "loss": 2.0764,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986979643952e-05,
+      "loss": 2.075,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986880833241304e-05,
+      "loss": 2.0444,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986781649184926e-05,
+      "loss": 2.0105,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998668209178547e-05,
+      "loss": 2.0674,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986582161045537e-05,
+      "loss": 2.0837,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986481856967756e-05,
+      "loss": 2.0696,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998638117955475e-05,
+      "loss": 2.061,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986280128809164e-05,
+      "loss": 2.0847,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998617870473365e-05,
+      "loss": 2.0667,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9986076907330866e-05,
+      "loss": 2.0903,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998597473660348e-05,
+      "loss": 2.0754,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985872192554167e-05,
+      "loss": 2.052,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985769275185628e-05,
+      "loss": 2.053,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998566598450055e-05,
+      "loss": 2.0798,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985562320501647e-05,
+      "loss": 2.0813,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985458283191637e-05,
+      "loss": 2.0542,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998535387257325e-05,
+      "loss": 2.0359,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998524908864922e-05,
+      "loss": 2.0823,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985143931422304e-05,
+      "loss": 2.0647,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9985038400895245e-05,
+      "loss": 2.0696,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998493249707082e-05,
+      "loss": 2.0613,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998482621995181e-05,
+      "loss": 2.0559,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998471956954099e-05,
+      "loss": 2.0718,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9984612545841167e-05,
+      "loss": 2.0542,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998450514885514e-05,
+      "loss": 2.0161,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9984397378585725e-05,
+      "loss": 2.0571,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9984289235035763e-05,
+      "loss": 2.0859,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9984180718208074e-05,
+      "loss": 2.0984,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998407182810551e-05,
+      "loss": 2.0642,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9983962564730926e-05,
+      "loss": 2.0757,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998385292808718e-05,
+      "loss": 2.1064,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998374291817716e-05,
+      "loss": 2.0542,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9983632535003746e-05,
+      "loss": 2.0498,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9983521778569828e-05,
+      "loss": 2.083,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998341064887832e-05,
+      "loss": 2.0745,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.998329914593212e-05,
+      "loss": 2.0493,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9983187269734168e-05,
+      "loss": 2.0271,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998307502028739e-05,
+      "loss": 2.0259,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982962397594735e-05,
+      "loss": 2.0325,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982849401659152e-05,
+      "loss": 2.0593,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982736032483604e-05,
+      "loss": 2.0691,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982622290071064e-05,
+      "loss": 2.0879,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982508174424516e-05,
+      "loss": 2.0872,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982393685546956e-05,
+      "loss": 2.0657,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982278823441378e-05,
+      "loss": 2.0649,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9982163588110802e-05,
+      "loss": 2.0518,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998204797955825e-05,
+      "loss": 2.0847,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981931997786743e-05,
+      "loss": 2.0725,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981815642799337e-05,
+      "loss": 2.0425,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998169891459907e-05,
+      "loss": 2.0278,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981581813189016e-05,
+      "loss": 2.0154,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981464338572234e-05,
+      "loss": 2.0464,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981346490751814e-05,
+      "loss": 2.033,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981228269730835e-05,
+      "loss": 2.0684,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9981109675512412e-05,
+      "loss": 2.0461,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9980990708099644e-05,
+      "loss": 2.082,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9980871367495656e-05,
+      "loss": 2.0715,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9980751653703574e-05,
+      "loss": 2.0771,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998063156672654e-05,
+      "loss": 2.103,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.99805111065677e-05,
+      "loss": 2.0532,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998039027323022e-05,
+      "loss": 2.021,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998026906671726e-05,
+      "loss": 2.0747,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9980147487032005e-05,
+      "loss": 2.0757,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.998002553417764e-05,
+      "loss": 2.033,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979903208157363e-05,
+      "loss": 2.0483,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979780508974382e-05,
+      "loss": 2.0947,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979657436631918e-05,
+      "loss": 2.1035,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979533991133194e-05,
+      "loss": 2.1177,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979410172481448e-05,
+      "loss": 2.0718,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979285980679932e-05,
+      "loss": 2.0309,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9979161415731894e-05,
+      "loss": 2.052,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997903647764061e-05,
+      "loss": 2.0898,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978911166409345e-05,
+      "loss": 2.0933,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978785482041396e-05,
+      "loss": 2.0688,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978659424540055e-05,
+      "loss": 2.1072,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978532993908625e-05,
+      "loss": 2.0181,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978406190150427e-05,
+      "loss": 2.0325,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978279013268784e-05,
+      "loss": 2.0439,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978151463267025e-05,
+      "loss": 2.0674,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9978023540148505e-05,
+      "loss": 2.022,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977895243916574e-05,
+      "loss": 2.0305,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977766574574595e-05,
+      "loss": 2.0835,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977637532125942e-05,
+      "loss": 2.1057,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977508116574e-05,
+      "loss": 2.1155,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977378327922167e-05,
+      "loss": 2.1323,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997724816617384e-05,
+      "loss": 2.0532,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9977117631332434e-05,
+      "loss": 2.052,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976986723401376e-05,
+      "loss": 2.0872,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976855442384093e-05,
+      "loss": 2.1099,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976723788284033e-05,
+      "loss": 2.0315,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997659176110465e-05,
+      "loss": 2.0605,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976459360849396e-05,
+      "loss": 2.1123,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976326587521748e-05,
+      "loss": 2.0942,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9976193441125194e-05,
+      "loss": 2.0479,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997605992166322e-05,
+      "loss": 2.0581,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997592602913933e-05,
+      "loss": 2.084,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9975791763557034e-05,
+      "loss": 2.0535,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9975657124919847e-05,
+      "loss": 2.0894,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997552211323131e-05,
+      "loss": 2.1165,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9975386728494958e-05,
+      "loss": 2.0867,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9975250970714344e-05,
+      "loss": 2.0913,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9975114839893024e-05,
+      "loss": 2.103,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997497833603457e-05,
+      "loss": 2.0872,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9974841459142563e-05,
+      "loss": 2.1042,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9974704209220588e-05,
+      "loss": 2.0798,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9974566586272248e-05,
+      "loss": 2.1089,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9974428590301153e-05,
+      "loss": 2.0793,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997429022131092e-05,
+      "loss": 2.0977,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997415147930518e-05,
+      "loss": 2.094,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9974012364287567e-05,
+      "loss": 2.0784,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997387287626173e-05,
+      "loss": 2.0371,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997373301523133e-05,
+      "loss": 2.0398,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997359278120003e-05,
+      "loss": 2.0356,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9973452174171515e-05,
+      "loss": 2.0427,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9973311194149465e-05,
+      "loss": 2.054,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997316984113758e-05,
+      "loss": 2.0603,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9973028115139566e-05,
+      "loss": 2.0515,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972886016159136e-05,
+      "loss": 2.0911,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972743544200027e-05,
+      "loss": 2.0884,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972600699265968e-05,
+      "loss": 2.0842,
+      "step": 983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972457481360703e-05,
+      "loss": 2.072,
+      "step": 984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972313890487994e-05,
+      "loss": 2.1057,
+      "step": 985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9972169926651594e-05,
+      "loss": 2.0767,
+      "step": 986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997202558985529e-05,
+      "loss": 2.0159,
+      "step": 987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997188088010287e-05,
+      "loss": 2.0139,
+      "step": 988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997173579739812e-05,
+      "loss": 2.0989,
+      "step": 989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9971590341744847e-05,
+      "loss": 2.0908,
+      "step": 990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9971444513146867e-05,
+      "loss": 2.0615,
+      "step": 991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9971298311608e-05,
+      "loss": 2.0896,
+      "step": 992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9971151737132087e-05,
+      "loss": 2.0227,
+      "step": 993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997100478972296e-05,
+      "loss": 2.0396,
+      "step": 994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997085746938449e-05,
+      "loss": 2.0767,
+      "step": 995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9970709776120527e-05,
+      "loss": 2.0452,
+      "step": 996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.997056170993495e-05,
+      "loss": 2.0996,
+      "step": 997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9970413270831635e-05,
+      "loss": 2.0718,
+      "step": 998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9970264458814486e-05,
+      "loss": 2.1028,
+      "step": 999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9970115273887393e-05,
+      "loss": 2.0425,
+      "step": 1000
+    },
+    {
+      "Bleu_1": 0.772208312148033,
+      "Bleu_2": 0.6130826759096824,
+      "Bleu_3": 0.47280508889725675,
+      "Bleu_4": 0.3608290127105116,
+      "CIDEr": 1.2020997313239483,
+      "METEOR": 0.27812394683892355,
+      "ROUGE_L": 0.5585914536812749,
+      "epoch": 0.05,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969965716054274e-05,
+      "loss": 2.0786,
+      "step": 1001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969815785319055e-05,
+      "loss": 2.0701,
+      "step": 1002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969665481685658e-05,
+      "loss": 2.0591,
+      "step": 1003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969514805158034e-05,
+      "loss": 2.0459,
+      "step": 1004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996936375574013e-05,
+      "loss": 2.0654,
+      "step": 1005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969212333435906e-05,
+      "loss": 2.0583,
+      "step": 1006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9969060538249337e-05,
+      "loss": 2.0203,
+      "step": 1007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9968908370184398e-05,
+      "loss": 1.9944,
+      "step": 1008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9968755829245083e-05,
+      "loss": 2.0898,
+      "step": 1009
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9968602915435396e-05,
+      "loss": 2.105,
+      "step": 1010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996844962875934e-05,
+      "loss": 2.0649,
+      "step": 1011
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9968295969220935e-05,
+      "loss": 2.0557,
+      "step": 1012
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9968141936824214e-05,
+      "loss": 2.0359,
+      "step": 1013
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996798753157322e-05,
+      "loss": 2.0417,
+      "step": 1014
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996783275347199e-05,
+      "loss": 2.0713,
+      "step": 1015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9967677602524596e-05,
+      "loss": 2.0972,
+      "step": 1016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.99675220787351e-05,
+      "loss": 2.009,
+      "step": 1017
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996736618210758e-05,
+      "loss": 2.0288,
+      "step": 1018
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996720991264613e-05,
+      "loss": 2.1211,
+      "step": 1019
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996705327035484e-05,
+      "loss": 2.0962,
+      "step": 1020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996689625523782e-05,
+      "loss": 2.054,
+      "step": 1021
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996673886729919e-05,
+      "loss": 2.0425,
+      "step": 1022
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9966581106543077e-05,
+      "loss": 2.0911,
+      "step": 1023
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9966422972973616e-05,
+      "loss": 2.0801,
+      "step": 1024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9966264466594956e-05,
+      "loss": 2.1133,
+      "step": 1025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996610558741125e-05,
+      "loss": 2.1094,
+      "step": 1026
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9965946335426673e-05,
+      "loss": 2.0796,
+      "step": 1027
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996578671064539e-05,
+      "loss": 2.0696,
+      "step": 1028
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9965626713071598e-05,
+      "loss": 2.02,
+      "step": 1029
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9965466342709486e-05,
+      "loss": 2.0232,
+      "step": 1030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996530559956326e-05,
+      "loss": 2.0564,
+      "step": 1031
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9965144483637133e-05,
+      "loss": 2.0994,
+      "step": 1032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9964982994935335e-05,
+      "loss": 2.0845,
+      "step": 1033
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.99648211334621e-05,
+      "loss": 2.0637,
+      "step": 1034
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9964658899221672e-05,
+      "loss": 2.084,
+      "step": 1035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9964496292218303e-05,
+      "loss": 2.082,
+      "step": 1036
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996433331245626e-05,
+      "loss": 2.0562,
+      "step": 1037
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9964169959939815e-05,
+      "loss": 2.0337,
+      "step": 1038
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9964006234673256e-05,
+      "loss": 2.0183,
+      "step": 1039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996384213666087e-05,
+      "loss": 2.0247,
+      "step": 1040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996367766590697e-05,
+      "loss": 2.0881,
+      "step": 1041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9963512822415857e-05,
+      "loss": 2.0833,
+      "step": 1042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9963347606191855e-05,
+      "loss": 2.0601,
+      "step": 1043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9963182017239308e-05,
+      "loss": 2.0549,
+      "step": 1044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996301605556255e-05,
+      "loss": 2.0669,
+      "step": 1045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9962849721165936e-05,
+      "loss": 2.0671,
+      "step": 1046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9962683014053825e-05,
+      "loss": 2.0796,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996251593423059e-05,
+      "loss": 2.0725,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9962348481700614e-05,
+      "loss": 2.0657,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9962180656468282e-05,
+      "loss": 2.0876,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9962012458538005e-05,
+      "loss": 2.0947,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9961843887914192e-05,
+      "loss": 2.1013,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9961674944601253e-05,
+      "loss": 2.0317,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9961505628603633e-05,
+      "loss": 1.9988,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996133593992576e-05,
+      "loss": 2.0349,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9961165878572088e-05,
+      "loss": 2.0803,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996099544454708e-05,
+      "loss": 2.0854,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.99608246378552e-05,
+      "loss": 2.0637,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9960653458500934e-05,
+      "loss": 2.1094,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9960481906488765e-05,
+      "loss": 2.0725,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9960309981823194e-05,
+      "loss": 2.0903,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.996013768450873e-05,
+      "loss": 2.0476,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9959965014549887e-05,
+      "loss": 2.0784,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9959791971951204e-05,
+      "loss": 2.0872,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9959618556717205e-05,
+      "loss": 2.0339,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9959444768852447e-05,
+      "loss": 2.0562,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995927060836148e-05,
+      "loss": 2.1182,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995909607524888e-05,
+      "loss": 2.1064,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995892116951922e-05,
+      "loss": 2.0625,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9958745891177087e-05,
+      "loss": 2.0842,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9958570240227075e-05,
+      "loss": 2.041,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995839421667379e-05,
+      "loss": 1.9927,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9958217820521853e-05,
+      "loss": 2.0103,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9958041051775884e-05,
+      "loss": 2.0144,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9957863910440525e-05,
+      "loss": 2.074,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9957686396520413e-05,
+      "loss": 2.0615,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995750851002021e-05,
+      "loss": 1.998,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9957330250944584e-05,
+      "loss": 2.0061,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9957151619298196e-05,
+      "loss": 2.053,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9956972615085743e-05,
+      "loss": 2.0813,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9956793238311915e-05,
+      "loss": 2.0757,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9956613488981413e-05,
+      "loss": 2.0542,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995643336709896e-05,
+      "loss": 2.1052,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9956252872669265e-05,
+      "loss": 2.1155,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9956072005697075e-05,
+      "loss": 2.0514,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9955890766187124e-05,
+      "loss": 2.021,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995570915414417e-05,
+      "loss": 2.1101,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9955527169572972e-05,
+      "loss": 2.075,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9955344812478307e-05,
+      "loss": 2.072,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995516208286495e-05,
+      "loss": 2.0535,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.99549789807377e-05,
+      "loss": 2.0691,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9954795506101355e-05,
+      "loss": 2.0815,
+      "step": 1092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9954611658960726e-05,
+      "loss": 2.0054,
+      "step": 1093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9954427439320638e-05,
+      "loss": 2.0181,
+      "step": 1094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.995424284718591e-05,
+      "loss": 2.0654,
+      "step": 1095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9954057882561403e-05,
+      "loss": 2.1062,
+      "step": 1096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9953872545451953e-05,
+      "loss": 2.0337,
+      "step": 1097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9953686835862425e-05,
+      "loss": 2.0239,
+      "step": 1098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9953500753797683e-05,
+      "loss": 2.0693,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9953314299262614e-05,
+      "loss": 2.0654,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9953127472262105e-05,
+      "loss": 2.0601,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9952940272801057e-05,
+      "loss": 2.0684,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9952752700884372e-05,
+      "loss": 2.1248,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9952564756516977e-05,
+      "loss": 2.0989,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99523764397038e-05,
+      "loss": 2.0415,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.995218775044978e-05,
+      "loss": 2.0505,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9951998688759855e-05,
+      "loss": 2.0537,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9951809254638993e-05,
+      "loss": 2.0415,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.995161944809216e-05,
+      "loss": 2.1331,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9951429269124335e-05,
+      "loss": 2.1328,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99512387177405e-05,
+      "loss": 2.0405,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9951047793945656e-05,
+      "loss": 2.0513,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99508564977448e-05,
+      "loss": 2.095,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9950664829142967e-05,
+      "loss": 2.0662,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9950472788145174e-05,
+      "loss": 2.0361,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.995028037475645e-05,
+      "loss": 2.072,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9950087588981847e-05,
+      "loss": 2.0859,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9949894430826422e-05,
+      "loss": 2.0906,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9949700900295242e-05,
+      "loss": 2.0559,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994950699739338e-05,
+      "loss": 2.0281,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9949312722125916e-05,
+      "loss": 2.0791,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994911807449795e-05,
+      "loss": 2.0767,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9948923054514583e-05,
+      "loss": 2.0813,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9948727662180936e-05,
+      "loss": 2.0955,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9948531897502126e-05,
+      "loss": 2.0867,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994833576048329e-05,
+      "loss": 2.0674,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9948139251129567e-05,
+      "loss": 2.0813,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994794236944612e-05,
+      "loss": 2.063,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9947745115438096e-05,
+      "loss": 2.1392,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9947547489110684e-05,
+      "loss": 2.1294,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994734949046906e-05,
+      "loss": 2.02,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994715111951841e-05,
+      "loss": 2.0398,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994695237626395e-05,
+      "loss": 2.0923,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994675326071088e-05,
+      "loss": 2.0725,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9946553772864425e-05,
+      "loss": 2.0798,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994635391272982e-05,
+      "loss": 2.064,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9946153680312302e-05,
+      "loss": 2.0776,
+      "step": 1137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994595307561712e-05,
+      "loss": 2.071,
+      "step": 1138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994575209864954e-05,
+      "loss": 2.0452,
+      "step": 1139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994555074941483e-05,
+      "loss": 2.0237,
+      "step": 1140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994534902791827e-05,
+      "loss": 2.0386,
+      "step": 1141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9945146934165148e-05,
+      "loss": 2.0356,
+      "step": 1142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9944944468160767e-05,
+      "loss": 2.0452,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9944741629910438e-05,
+      "loss": 2.0283,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9944538419419472e-05,
+      "loss": 2.0933,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9944334836693206e-05,
+      "loss": 2.0984,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9944130881736977e-05,
+      "loss": 2.074,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9943926554556134e-05,
+      "loss": 2.0808,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994372185515603e-05,
+      "loss": 2.0679,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9943516783542037e-05,
+      "loss": 2.0713,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9943311339719533e-05,
+      "loss": 2.0798,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9943105523693903e-05,
+      "loss": 2.0693,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9942899335470548e-05,
+      "loss": 1.9954,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9942692775054878e-05,
+      "loss": 1.9731,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9942485842452297e-05,
+      "loss": 2.0593,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9942278537668244e-05,
+      "loss": 2.0444,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994207086070815e-05,
+      "loss": 2.0317,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9941862811577464e-05,
+      "loss": 2.0145,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9941654390281638e-05,
+      "loss": 2.0798,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9941445596826137e-05,
+      "loss": 2.0557,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9941236431216446e-05,
+      "loss": 2.0613,
+      "step": 1161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994102689345804e-05,
+      "loss": 2.0291,
+      "step": 1162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9940816983556413e-05,
+      "loss": 2.1196,
+      "step": 1163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994060670151708e-05,
+      "loss": 2.0923,
+      "step": 1164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9940396047345538e-05,
+      "loss": 2.062,
+      "step": 1165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.994018502104733e-05,
+      "loss": 2.0588,
+      "step": 1166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9939973622627988e-05,
+      "loss": 2.0425,
+      "step": 1167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9939761852093038e-05,
+      "loss": 2.0444,
+      "step": 1168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9939549709448055e-05,
+      "loss": 2.073,
+      "step": 1169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9939337194698584e-05,
+      "loss": 2.0544,
+      "step": 1170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9939124307850213e-05,
+      "loss": 2.0679,
+      "step": 1171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9938911048908515e-05,
+      "loss": 2.0464,
+      "step": 1172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9938697417879083e-05,
+      "loss": 2.0828,
+      "step": 1173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9938483414767526e-05,
+      "loss": 2.0735,
+      "step": 1174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993826903957945e-05,
+      "loss": 2.1248,
+      "step": 1175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993805429232048e-05,
+      "loss": 2.0917,
+      "step": 1176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9937839172996247e-05,
+      "loss": 2.1135,
+      "step": 1177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9937623681612385e-05,
+      "loss": 2.0808,
+      "step": 1178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9937407818174557e-05,
+      "loss": 2.0723,
+      "step": 1179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9937191582688415e-05,
+      "loss": 2.0835,
+      "step": 1180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9936974975159634e-05,
+      "loss": 1.9944,
+      "step": 1181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9936757995593893e-05,
+      "loss": 1.9792,
+      "step": 1182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993654064399688e-05,
+      "loss": 2.0823,
+      "step": 1183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9936322920374295e-05,
+      "loss": 2.0706,
+      "step": 1184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993610482473185e-05,
+      "loss": 2.0754,
+      "step": 1185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993588635707527e-05,
+      "loss": 2.073,
+      "step": 1186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9935667517410274e-05,
+      "loss": 2.0876,
+      "step": 1187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99354483057426e-05,
+      "loss": 2.0767,
+      "step": 1188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9935228722078004e-05,
+      "loss": 2.0803,
+      "step": 1189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9935008766422244e-05,
+      "loss": 2.0598,
+      "step": 1190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9934788438781082e-05,
+      "loss": 2.0408,
+      "step": 1191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99345677391603e-05,
+      "loss": 2.0583,
+      "step": 1192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993434666756568e-05,
+      "loss": 2.0769,
+      "step": 1193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993412522400303e-05,
+      "loss": 2.064,
+      "step": 1194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9933903408478152e-05,
+      "loss": 2.0994,
+      "step": 1195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993368122099686e-05,
+      "loss": 2.0654,
+      "step": 1196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9933458661564982e-05,
+      "loss": 2.0901,
+      "step": 1197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9933235730188356e-05,
+      "loss": 2.0894,
+      "step": 1198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9933012426872827e-05,
+      "loss": 2.0928,
+      "step": 1199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993278875162425e-05,
+      "loss": 2.1018,
+      "step": 1200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993256470444849e-05,
+      "loss": 2.085,
+      "step": 1201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993234028535143e-05,
+      "loss": 2.0857,
+      "step": 1202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9932115494338947e-05,
+      "loss": 2.0652,
+      "step": 1203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9931890331416934e-05,
+      "loss": 2.0479,
+      "step": 1204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99316647965913e-05,
+      "loss": 2.1067,
+      "step": 1205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9931438889867962e-05,
+      "loss": 2.0991,
+      "step": 1206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9931212611252844e-05,
+      "loss": 2.0349,
+      "step": 1207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993098596075187e-05,
+      "loss": 2.0425,
+      "step": 1208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9930758938370997e-05,
+      "loss": 2.0867,
+      "step": 1209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993053154411617e-05,
+      "loss": 2.0745,
+      "step": 1210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.993030377799335e-05,
+      "loss": 2.0771,
+      "step": 1211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9930075640008518e-05,
+      "loss": 2.095,
+      "step": 1212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9929847130167657e-05,
+      "loss": 2.0442,
+      "step": 1213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9929618248476747e-05,
+      "loss": 2.0479,
+      "step": 1214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9929388994941806e-05,
+      "loss": 2.0957,
+      "step": 1215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9929159369568833e-05,
+      "loss": 2.0916,
+      "step": 1216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9928929372363856e-05,
+      "loss": 2.0498,
+      "step": 1217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9928699003332906e-05,
+      "loss": 2.032,
+      "step": 1218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9928468262482025e-05,
+      "loss": 2.0667,
+      "step": 1219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992823714981726e-05,
+      "loss": 2.019,
+      "step": 1220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9928005665344674e-05,
+      "loss": 2.1064,
+      "step": 1221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992777380907034e-05,
+      "loss": 2.0874,
+      "step": 1222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9927541581000336e-05,
+      "loss": 2.0376,
+      "step": 1223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992730898114075e-05,
+      "loss": 2.0491,
+      "step": 1224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9927076009497683e-05,
+      "loss": 2.0266,
+      "step": 1225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9926842666077244e-05,
+      "loss": 2.0154,
+      "step": 1226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9926608950885557e-05,
+      "loss": 2.0935,
+      "step": 1227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992637486392874e-05,
+      "loss": 2.0867,
+      "step": 1228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9926140405212946e-05,
+      "loss": 2.0859,
+      "step": 1229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9925905574744307e-05,
+      "loss": 2.0967,
+      "step": 1230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9925670372529e-05,
+      "loss": 2.1038,
+      "step": 1231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9925434798573175e-05,
+      "loss": 2.1035,
+      "step": 1232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992519885288302e-05,
+      "loss": 2.0006,
+      "step": 1233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992496253546472e-05,
+      "loss": 2.0542,
+      "step": 1234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9924725846324475e-05,
+      "loss": 2.0645,
+      "step": 1235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9924488785468484e-05,
+      "loss": 2.0752,
+      "step": 1236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992425135290297e-05,
+      "loss": 2.0376,
+      "step": 1237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9924013548634156e-05,
+      "loss": 2.0276,
+      "step": 1238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9923775372668288e-05,
+      "loss": 2.0591,
+      "step": 1239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99235368250116e-05,
+      "loss": 2.043,
+      "step": 1240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992329790567035e-05,
+      "loss": 2.1157,
+      "step": 1241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9923058614650803e-05,
+      "loss": 2.1099,
+      "step": 1242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992281895195924e-05,
+      "loss": 2.0762,
+      "step": 1243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992257891760194e-05,
+      "loss": 2.0219,
+      "step": 1244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9922338511585203e-05,
+      "loss": 2.0747,
+      "step": 1245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9922097733915324e-05,
+      "loss": 2.0698,
+      "step": 1246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.992185658459863e-05,
+      "loss": 2.0803,
+      "step": 1247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9921615063641434e-05,
+      "loss": 2.0649,
+      "step": 1248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9921373171050075e-05,
+      "loss": 2.0886,
+      "step": 1249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9921130906830894e-05,
+      "loss": 2.1052,
+      "step": 1250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9920888270990245e-05,
+      "loss": 2.0649,
+      "step": 1251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9920645263534494e-05,
+      "loss": 2.061,
+      "step": 1252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9920401884470006e-05,
+      "loss": 2.0942,
+      "step": 1253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9920158133803173e-05,
+      "loss": 2.0566,
+      "step": 1254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9919914011540376e-05,
+      "loss": 2.1113,
+      "step": 1255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9919669517688027e-05,
+      "loss": 2.1284,
+      "step": 1256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9919424652252533e-05,
+      "loss": 2.1091,
+      "step": 1257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9919179415240314e-05,
+      "loss": 2.0585,
+      "step": 1258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9918933806657803e-05,
+      "loss": 2.0813,
+      "step": 1259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991868782651144e-05,
+      "loss": 2.0896,
+      "step": 1260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9918441474807676e-05,
+      "loss": 2.0874,
+      "step": 1261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9918194751552976e-05,
+      "loss": 2.0508,
+      "step": 1262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9917947656753798e-05,
+      "loss": 2.0947,
+      "step": 1263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9917700190416635e-05,
+      "loss": 2.0691,
+      "step": 1264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991745235254797e-05,
+      "loss": 1.9912,
+      "step": 1265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9917204143154297e-05,
+      "loss": 1.9802,
+      "step": 1266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9916955562242136e-05,
+      "loss": 2.0906,
+      "step": 1267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9916706609818e-05,
+      "loss": 2.0654,
+      "step": 1268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991645728588842e-05,
+      "loss": 2.0508,
+      "step": 1269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9916207590459928e-05,
+      "loss": 2.0645,
+      "step": 1270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9915957523539083e-05,
+      "loss": 2.0969,
+      "step": 1271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991570708513243e-05,
+      "loss": 2.0955,
+      "step": 1272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9915456275246543e-05,
+      "loss": 2.0376,
+      "step": 1273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9915205093888003e-05,
+      "loss": 2.0469,
+      "step": 1274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991495354106339e-05,
+      "loss": 2.0544,
+      "step": 1275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.99147016167793e-05,
+      "loss": 2.017,
+      "step": 1276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9914449321042346e-05,
+      "loss": 2.082,
+      "step": 1277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991419665385914e-05,
+      "loss": 2.1003,
+      "step": 1278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9913943615236312e-05,
+      "loss": 2.0432,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991369020518049e-05,
+      "loss": 2.0564,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9913436423698327e-05,
+      "loss": 2.0571,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991318227079647e-05,
+      "loss": 2.0176,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9912927746481595e-05,
+      "loss": 2.0107,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9912672850760372e-05,
+      "loss": 2.0046,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991241758363948e-05,
+      "loss": 2.1079,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991216194512562e-05,
+      "loss": 2.0759,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9911905935225487e-05,
+      "loss": 2.051,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9911649553945803e-05,
+      "loss": 2.0519,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9911392801293288e-05,
+      "loss": 2.0376,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991113567727467e-05,
+      "loss": 2.04,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991087818189671e-05,
+      "loss": 2.0862,
+      "step": 1291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.991062031516614e-05,
+      "loss": 2.0496,
+      "step": 1292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9910362077089736e-05,
+      "loss": 2.0879,
+      "step": 1293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9910103467674254e-05,
+      "loss": 2.0571,
+      "step": 1294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.990984448692649e-05,
+      "loss": 2.0181,
+      "step": 1295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9909585134853234e-05,
+      "loss": 2.0481,
+      "step": 1296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9909325411461287e-05,
+      "loss": 2.043,
+      "step": 1297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9909065316757453e-05,
+      "loss": 2.0583,
+      "step": 1298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.990880485074856e-05,
+      "loss": 2.0598,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990854401344143e-05,
+      "loss": 2.0508,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9908282804842914e-05,
+      "loss": 2.092,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9908021224959858e-05,
+      "loss": 2.1199,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9907759273799117e-05,
+      "loss": 2.0706,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9907496951367562e-05,
+      "loss": 2.0823,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9907234257672078e-05,
+      "loss": 2.1431,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9906971192719546e-05,
+      "loss": 2.1021,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990670775651687e-05,
+      "loss": 2.0862,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9906443949070957e-05,
+      "loss": 2.0969,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9906179770388725e-05,
+      "loss": 2.0715,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9905915220477098e-05,
+      "loss": 2.0659,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990565029934302e-05,
+      "loss": 2.04,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990538500699344e-05,
+      "loss": 2.0615,
+      "step": 1312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.99051193434353e-05,
+      "loss": 2.0796,
+      "step": 1313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990485330867558e-05,
+      "loss": 2.0618,
+      "step": 1314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9904586902721254e-05,
+      "loss": 2.1155,
+      "step": 1315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9904320125579308e-05,
+      "loss": 2.083,
+      "step": 1316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9904052977256737e-05,
+      "loss": 2.0793,
+      "step": 1317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9903785457760544e-05,
+      "loss": 2.0864,
+      "step": 1318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9903517567097753e-05,
+      "loss": 2.0867,
+      "step": 1319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9903249305275383e-05,
+      "loss": 2.0779,
+      "step": 1320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9902980672300465e-05,
+      "loss": 2.0532,
+      "step": 1321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9902711668180047e-05,
+      "loss": 2.0669,
+      "step": 1322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990244229292119e-05,
+      "loss": 2.1147,
+      "step": 1323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990217254653095e-05,
+      "loss": 2.1191,
+      "step": 1324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9901902429016402e-05,
+      "loss": 2.1067,
+      "step": 1325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990163194038463e-05,
+      "loss": 2.094,
+      "step": 1326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9901361080642732e-05,
+      "loss": 2.1228,
+      "step": 1327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.99010898497978e-05,
+      "loss": 2.0806,
+      "step": 1328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990081824785696e-05,
+      "loss": 2.051,
+      "step": 1329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9900546274827326e-05,
+      "loss": 2.043,
+      "step": 1330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.990027393071603e-05,
+      "loss": 2.1138,
+      "step": 1331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9900001215530218e-05,
+      "loss": 2.1038,
+      "step": 1332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989972812927704e-05,
+      "loss": 2.053,
+      "step": 1333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9899454671963656e-05,
+      "loss": 2.0105,
+      "step": 1334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989918084359724e-05,
+      "loss": 2.0588,
+      "step": 1335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9898906644184974e-05,
+      "loss": 2.053,
+      "step": 1336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9898632073734038e-05,
+      "loss": 2.0525,
+      "step": 1337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9898357132251642e-05,
+      "loss": 2.0146,
+      "step": 1338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9898081819744997e-05,
+      "loss": 2.0718,
+      "step": 1339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989780613622132e-05,
+      "loss": 2.0664,
+      "step": 1340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9897530081687835e-05,
+      "loss": 2.0239,
+      "step": 1341
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989725365615179e-05,
+      "loss": 2.0298,
+      "step": 1342
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989697685962043e-05,
+      "loss": 2.0393,
+      "step": 1343
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9896699692101014e-05,
+      "loss": 2.054,
+      "step": 1344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989642215360081e-05,
+      "loss": 2.041,
+      "step": 1345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9896144244127096e-05,
+      "loss": 2.0747,
+      "step": 1346
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9895865963687162e-05,
+      "loss": 2.0896,
+      "step": 1347
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98955873122883e-05,
+      "loss": 2.0371,
+      "step": 1348
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9895308289937826e-05,
+      "loss": 2.0562,
+      "step": 1349
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989502889664305e-05,
+      "loss": 2.0403,
+      "step": 1350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98947491324113e-05,
+      "loss": 2.0603,
+      "step": 1351
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9894468997249916e-05,
+      "loss": 2.032,
+      "step": 1352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989418849116624e-05,
+      "loss": 2.0632,
+      "step": 1353
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9893907614167636e-05,
+      "loss": 2.0596,
+      "step": 1354
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9893626366261453e-05,
+      "loss": 2.1011,
+      "step": 1355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9893344747455084e-05,
+      "loss": 2.0833,
+      "step": 1356
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9893062757755905e-05,
+      "loss": 2.0139,
+      "step": 1357
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989278039717131e-05,
+      "loss": 2.0342,
+      "step": 1358
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989249766570871e-05,
+      "loss": 2.1189,
+      "step": 1359
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989221456337551e-05,
+      "loss": 2.0813,
+      "step": 1360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9891931090179145e-05,
+      "loss": 2.105,
+      "step": 1361
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9891647246127045e-05,
+      "loss": 2.0974,
+      "step": 1362
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9891363031226647e-05,
+      "loss": 2.0251,
+      "step": 1363
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9891078445485414e-05,
+      "loss": 2.0537,
+      "step": 1364
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98907934889108e-05,
+      "loss": 2.1135,
+      "step": 1365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.989050816151028e-05,
+      "loss": 2.0903,
+      "step": 1366
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9890222463291342e-05,
+      "loss": 2.0645,
+      "step": 1367
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9889936394261475e-05,
+      "loss": 2.0603,
+      "step": 1368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9889649954428177e-05,
+      "loss": 2.0715,
+      "step": 1369
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9889363143798965e-05,
+      "loss": 2.0491,
+      "step": 1370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988907596238136e-05,
+      "loss": 2.0645,
+      "step": 1371
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9888788410182882e-05,
+      "loss": 2.1099,
+      "step": 1372
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9888500487211088e-05,
+      "loss": 2.0903,
+      "step": 1373
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988821219347352e-05,
+      "loss": 2.0896,
+      "step": 1374
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9887923528977736e-05,
+      "loss": 2.0728,
+      "step": 1375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988763449373131e-05,
+      "loss": 2.0942,
+      "step": 1376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9887345087741823e-05,
+      "loss": 2.073,
+      "step": 1377
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988705531101686e-05,
+      "loss": 2.0681,
+      "step": 1378
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988676516356402e-05,
+      "loss": 2.0684,
+      "step": 1379
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9886474645390918e-05,
+      "loss": 2.0776,
+      "step": 1380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9886183756505165e-05,
+      "loss": 2.1238,
+      "step": 1381
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9885892496914395e-05,
+      "loss": 2.125,
+      "step": 1382
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9885600866626246e-05,
+      "loss": 2.1016,
+      "step": 1383
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9885308865648358e-05,
+      "loss": 2.0664,
+      "step": 1384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9885016493988395e-05,
+      "loss": 2.0322,
+      "step": 1385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9884723751654024e-05,
+      "loss": 2.0432,
+      "step": 1386
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9884430638652925e-05,
+      "loss": 2.0474,
+      "step": 1387
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988413715499277e-05,
+      "loss": 2.0535,
+      "step": 1388
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988384330068127e-05,
+      "loss": 2.0681,
+      "step": 1389
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9883549075726128e-05,
+      "loss": 2.0276,
+      "step": 1390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988325448013506e-05,
+      "loss": 1.9973,
+      "step": 1391
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9882959513915783e-05,
+      "loss": 1.9905,
+      "step": 1392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9882664177076045e-05,
+      "loss": 2.1111,
+      "step": 1393
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988236846962358e-05,
+      "loss": 2.1035,
+      "step": 1394
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9882072391566147e-05,
+      "loss": 2.0662,
+      "step": 1395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988177594291151e-05,
+      "loss": 2.0723,
+      "step": 1396
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9881479123667448e-05,
+      "loss": 2.0623,
+      "step": 1397
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988118193384173e-05,
+      "loss": 2.0601,
+      "step": 1398
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9880884373442173e-05,
+      "loss": 2.0791,
+      "step": 1399
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988058644247656e-05,
+      "loss": 2.041,
+      "step": 1400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.988028814095271e-05,
+      "loss": 1.9937,
+      "step": 1401
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9879989468878447e-05,
+      "loss": 2.022,
+      "step": 1402
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9879690426261603e-05,
+      "loss": 2.0552,
+      "step": 1403
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987939101311002e-05,
+      "loss": 2.0457,
+      "step": 1404
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987909122943155e-05,
+      "loss": 2.0459,
+      "step": 1405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987879107523405e-05,
+      "loss": 2.072,
+      "step": 1406
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98784905505254e-05,
+      "loss": 2.1331,
+      "step": 1407
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987818965531347e-05,
+      "loss": 2.073,
+      "step": 1408
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987788838960616e-05,
+      "loss": 1.9729,
+      "step": 1409
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987758675341137e-05,
+      "loss": 2.0002,
+      "step": 1410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9877284746737005e-05,
+      "loss": 2.0447,
+      "step": 1411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9876982369590985e-05,
+      "loss": 2.0669,
+      "step": 1412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987667962198124e-05,
+      "loss": 2.0558,
+      "step": 1413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987637650391571e-05,
+      "loss": 2.0886,
+      "step": 1414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9876073015402352e-05,
+      "loss": 2.0667,
+      "step": 1415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9875769156449114e-05,
+      "loss": 2.0544,
+      "step": 1416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987546492706396e-05,
+      "loss": 2.094,
+      "step": 1417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9875160327254885e-05,
+      "loss": 2.073,
+      "step": 1418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9874855357029864e-05,
+      "loss": 1.9985,
+      "step": 1419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9874550016396894e-05,
+      "loss": 1.9838,
+      "step": 1420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987424430536399e-05,
+      "loss": 1.9951,
+      "step": 1421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9873938223939163e-05,
+      "loss": 2.0166,
+      "step": 1422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9873631772130447e-05,
+      "loss": 2.0557,
+      "step": 1423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987332494994586e-05,
+      "loss": 2.0142,
+      "step": 1424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987301775739347e-05,
+      "loss": 2.1104,
+      "step": 1425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9872710194481326e-05,
+      "loss": 2.0947,
+      "step": 1426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9872402261217488e-05,
+      "loss": 2.0596,
+      "step": 1427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9872093957610035e-05,
+      "loss": 2.0669,
+      "step": 1428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9871785283667048e-05,
+      "loss": 2.0884,
+      "step": 1429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.987147623939663e-05,
+      "loss": 2.0476,
+      "step": 1430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9871166824806874e-05,
+      "loss": 2.0435,
+      "step": 1431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9870857039905902e-05,
+      "loss": 2.0166,
+      "step": 1432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9870546884701838e-05,
+      "loss": 2.1226,
+      "step": 1433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9870236359202813e-05,
+      "loss": 2.0833,
+      "step": 1434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986992546341697e-05,
+      "loss": 2.0823,
+      "step": 1435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986961419735246e-05,
+      "loss": 2.0837,
+      "step": 1436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9869302561017447e-05,
+      "loss": 2.0193,
+      "step": 1437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986899055442011e-05,
+      "loss": 2.0576,
+      "step": 1438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9868678177568623e-05,
+      "loss": 2.0684,
+      "step": 1439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9868365430471184e-05,
+      "loss": 2.0881,
+      "step": 1440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9868052313135987e-05,
+      "loss": 2.0813,
+      "step": 1441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9867738825571246e-05,
+      "loss": 2.0488,
+      "step": 1442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9867424967785182e-05,
+      "loss": 2.0608,
+      "step": 1443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9867110739786028e-05,
+      "loss": 2.0491,
+      "step": 1444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986679614158202e-05,
+      "loss": 2.0833,
+      "step": 1445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986648117318141e-05,
+      "loss": 2.0549,
+      "step": 1446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986616583459246e-05,
+      "loss": 2.033,
+      "step": 1447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986585012582344e-05,
+      "loss": 2.0505,
+      "step": 1448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986553404688262e-05,
+      "loss": 2.0359,
+      "step": 1449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98652175977783e-05,
+      "loss": 2.0383,
+      "step": 1450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9864900778518772e-05,
+      "loss": 2.0667,
+      "step": 1451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9864583589112345e-05,
+      "loss": 2.0576,
+      "step": 1452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9864266029567337e-05,
+      "loss": 2.0261,
+      "step": 1453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986394809989208e-05,
+      "loss": 2.0286,
+      "step": 1454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9863629800094904e-05,
+      "loss": 2.1077,
+      "step": 1455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9863311130184166e-05,
+      "loss": 2.085,
+      "step": 1456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9862992090168212e-05,
+      "loss": 2.0642,
+      "step": 1457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9862672680055412e-05,
+      "loss": 2.1047,
+      "step": 1458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9862352899854148e-05,
+      "loss": 2.0305,
+      "step": 1459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9862032749572798e-05,
+      "loss": 2.0129,
+      "step": 1460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9861712229219765e-05,
+      "loss": 2.033,
+      "step": 1461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9861391338803445e-05,
+      "loss": 2.0522,
+      "step": 1462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9861070078332258e-05,
+      "loss": 2.0376,
+      "step": 1463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.986074844781463e-05,
+      "loss": 2.0652,
+      "step": 1464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9860426447258994e-05,
+      "loss": 2.0884,
+      "step": 1465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9860104076673795e-05,
+      "loss": 2.115,
+      "step": 1466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.985978133606748e-05,
+      "loss": 2.0659,
+      "step": 1467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9859458225448528e-05,
+      "loss": 2.0408,
+      "step": 1468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9859134744825398e-05,
+      "loss": 2.0544,
+      "step": 1469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9858810894206574e-05,
+      "loss": 2.0498,
+      "step": 1470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9858486673600556e-05,
+      "loss": 2.103,
+      "step": 1471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9858162083015847e-05,
+      "loss": 2.0884,
+      "step": 1472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9857837122460946e-05,
+      "loss": 2.0942,
+      "step": 1473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9857511791944388e-05,
+      "loss": 2.0486,
+      "step": 1474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9857186091474697e-05,
+      "loss": 2.0537,
+      "step": 1475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9856860021060416e-05,
+      "loss": 2.1042,
+      "step": 1476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98565335807101e-05,
+      "loss": 2.1194,
+      "step": 1477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9856206770432306e-05,
+      "loss": 2.0425,
+      "step": 1478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98558795902356e-05,
+      "loss": 2.0991,
+      "step": 1479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.985555204012857e-05,
+      "loss": 2.0737,
+      "step": 1480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98552241201198e-05,
+      "loss": 2.0381,
+      "step": 1481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9854895830217894e-05,
+      "loss": 2.0801,
+      "step": 1482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9854567170431455e-05,
+      "loss": 2.0544,
+      "step": 1483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9854238140769107e-05,
+      "loss": 2.0635,
+      "step": 1484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9853908741239475e-05,
+      "loss": 2.04,
+      "step": 1485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98535789718512e-05,
+      "loss": 2.0554,
+      "step": 1486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9853248832612925e-05,
+      "loss": 2.0579,
+      "step": 1487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9852918323533314e-05,
+      "loss": 2.0581,
+      "step": 1488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.985258744462103e-05,
+      "loss": 2.0369,
+      "step": 1489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9852256195884746e-05,
+      "loss": 2.074,
+      "step": 1490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9851924577333158e-05,
+      "loss": 2.0447,
+      "step": 1491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.985159258897496e-05,
+      "loss": 2.0627,
+      "step": 1492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.985126023081885e-05,
+      "loss": 2.0493,
+      "step": 1493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9850927502873554e-05,
+      "loss": 2.0232,
+      "step": 1494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9850594405147794e-05,
+      "loss": 2.0613,
+      "step": 1495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9850260937650302e-05,
+      "loss": 2.042,
+      "step": 1496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.984992710038982e-05,
+      "loss": 2.0076,
+      "step": 1497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9849592893375117e-05,
+      "loss": 2.0403,
+      "step": 1498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9849258316614943e-05,
+      "loss": 1.9897,
+      "step": 1499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9848923370118074e-05,
+      "loss": 1.9946,
+      "step": 1500
+    },
+    {
+      "Bleu_1": 0.7722717658570937,
+      "Bleu_2": 0.6109760700278793,
+      "Bleu_3": 0.46907008266515443,
+      "Bleu_4": 0.3561962472918271,
+      "CIDEr": 1.2002801569387105,
+      "METEOR": 0.278256696083119,
+      "ROUGE_L": 0.5580042628274962,
+      "epoch": 0.07,
+      "step": 1500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9848588053893296e-05,
+      "loss": 2.0764,
+      "step": 1501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9848252367949407e-05,
+      "loss": 2.1118,
+      "step": 1502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9847916312295203e-05,
+      "loss": 2.1113,
+      "step": 1503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9847579886939496e-05,
+      "loss": 2.0889,
+      "step": 1504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9847243091891113e-05,
+      "loss": 2.0659,
+      "step": 1505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9846905927158887e-05,
+      "loss": 2.0701,
+      "step": 1506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9846568392751657e-05,
+      "loss": 2.0518,
+      "step": 1507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.984623048867827e-05,
+      "loss": 2.0984,
+      "step": 1508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9845892214947593e-05,
+      "loss": 2.1003,
+      "step": 1509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9845553571568494e-05,
+      "loss": 2.1013,
+      "step": 1510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.984521455854986e-05,
+      "loss": 2.105,
+      "step": 1511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9844875175900575e-05,
+      "loss": 2.0481,
+      "step": 1512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9844535423629537e-05,
+      "loss": 2.0796,
+      "step": 1513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9844195301745663e-05,
+      "loss": 2.0427,
+      "step": 1514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9843854810257865e-05,
+      "loss": 2.1216,
+      "step": 1515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9843513949175077e-05,
+      "loss": 2.0906,
+      "step": 1516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9843172718506234e-05,
+      "loss": 2.071,
+      "step": 1517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9842831118260287e-05,
+      "loss": 2.0715,
+      "step": 1518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9842489148446194e-05,
+      "loss": 2.0247,
+      "step": 1519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.984214680907292e-05,
+      "loss": 2.021,
+      "step": 1520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9841804100149446e-05,
+      "loss": 2.0942,
+      "step": 1521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9841461021684757e-05,
+      "loss": 2.0806,
+      "step": 1522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.984111757368785e-05,
+      "loss": 2.085,
+      "step": 1523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9840773756167732e-05,
+      "loss": 2.0479,
+      "step": 1524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9840429569133424e-05,
+      "loss": 2.0337,
+      "step": 1525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9840085012593938e-05,
+      "loss": 2.0483,
+      "step": 1526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983974008655832e-05,
+      "loss": 2.0762,
+      "step": 1527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983939479103562e-05,
+      "loss": 2.0571,
+      "step": 1528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9839049126034885e-05,
+      "loss": 2.1074,
+      "step": 1529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983870309156518e-05,
+      "loss": 2.0781,
+      "step": 1530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9838356687635582e-05,
+      "loss": 2.0664,
+      "step": 1531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9838009914255175e-05,
+      "loss": 2.094,
+      "step": 1532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983766277143305e-05,
+      "loss": 2.0906,
+      "step": 1533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983731525917831e-05,
+      "loss": 2.0652,
+      "step": 1534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9836967377500076e-05,
+      "loss": 2.0918,
+      "step": 1535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983661912640746e-05,
+      "loss": 2.0681,
+      "step": 1536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9836270505909603e-05,
+      "loss": 2.093,
+      "step": 1537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983592151601564e-05,
+      "loss": 2.0859,
+      "step": 1538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983557215673473e-05,
+      "loss": 2.0588,
+      "step": 1539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983522242807603e-05,
+      "loss": 2.0649,
+      "step": 1540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9834872330048707e-05,
+      "loss": 2.0403,
+      "step": 1541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9834521862661957e-05,
+      "loss": 2.0696,
+      "step": 1542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9834171025924955e-05,
+      "loss": 2.0161,
+      "step": 1543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9833819819846907e-05,
+      "loss": 2.0225,
+      "step": 1544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9833468244437023e-05,
+      "loss": 2.0989,
+      "step": 1545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9833116299704524e-05,
+      "loss": 2.1182,
+      "step": 1546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9832763985658636e-05,
+      "loss": 2.084,
+      "step": 1547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9832411302308603e-05,
+      "loss": 2.0564,
+      "step": 1548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.983205824966367e-05,
+      "loss": 2.0837,
+      "step": 1549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9831704827733093e-05,
+      "loss": 2.0908,
+      "step": 1550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9831351036526147e-05,
+      "loss": 2.0754,
+      "step": 1551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9830996876052105e-05,
+      "loss": 2.0845,
+      "step": 1552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9830642346320256e-05,
+      "loss": 2.0149,
+      "step": 1553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9830287447339898e-05,
+      "loss": 2.0259,
+      "step": 1554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9829932179120334e-05,
+      "loss": 2.0732,
+      "step": 1555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9829576541670886e-05,
+      "loss": 2.0151,
+      "step": 1556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982922053500087e-05,
+      "loss": 2.0574,
+      "step": 1557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9828864159119635e-05,
+      "loss": 2.0127,
+      "step": 1558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982850741403652e-05,
+      "loss": 2.0488,
+      "step": 1559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9828150299760877e-05,
+      "loss": 2.0676,
+      "step": 1560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9827792816302078e-05,
+      "loss": 2.0891,
+      "step": 1561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9827434963669494e-05,
+      "loss": 2.0933,
+      "step": 1562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982707674187251e-05,
+      "loss": 2.0942,
+      "step": 1563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982671815092052e-05,
+      "loss": 2.053,
+      "step": 1564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9826359190822924e-05,
+      "loss": 2.0715,
+      "step": 1565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9825999861589137e-05,
+      "loss": 2.0603,
+      "step": 1566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982564016322859e-05,
+      "loss": 2.05,
+      "step": 1567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9825280095750708e-05,
+      "loss": 2.0305,
+      "step": 1568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9824919659164927e-05,
+      "loss": 2.1143,
+      "step": 1569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9824558853480712e-05,
+      "loss": 2.1338,
+      "step": 1570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982419767870752e-05,
+      "loss": 2.064,
+      "step": 1571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982383613485482e-05,
+      "loss": 2.0569,
+      "step": 1572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9823474221932097e-05,
+      "loss": 2.0623,
+      "step": 1573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9823111939948835e-05,
+      "loss": 2.0266,
+      "step": 1574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982274928891454e-05,
+      "loss": 2.0601,
+      "step": 1575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9822386268838723e-05,
+      "loss": 2.0723,
+      "step": 1576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.98220228797309e-05,
+      "loss": 2.0969,
+      "step": 1577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9821659121600603e-05,
+      "loss": 2.0862,
+      "step": 1578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982129499445737e-05,
+      "loss": 2.0518,
+      "step": 1579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.982093049831075e-05,
+      "loss": 2.0449,
+      "step": 1580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9820565633170302e-05,
+      "loss": 2.0261,
+      "step": 1581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9820200399045593e-05,
+      "loss": 2.032,
+      "step": 1582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.98198347959462e-05,
+      "loss": 2.0774,
+      "step": 1583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9819468823881718e-05,
+      "loss": 2.0994,
+      "step": 1584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9819102482861732e-05,
+      "loss": 2.0732,
+      "step": 1585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981873577289586e-05,
+      "loss": 2.085,
+      "step": 1586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981836869399371e-05,
+      "loss": 2.0547,
+      "step": 1587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9818001246164913e-05,
+      "loss": 2.0227,
+      "step": 1588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.98176334294191e-05,
+      "loss": 2.0591,
+      "step": 1589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9817265243765925e-05,
+      "loss": 2.0913,
+      "step": 1590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981689668921504e-05,
+      "loss": 2.0913,
+      "step": 1591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9816527765776103e-05,
+      "loss": 2.1162,
+      "step": 1592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9816158473458796e-05,
+      "loss": 2.072,
+      "step": 1593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9815788812272806e-05,
+      "loss": 2.0542,
+      "step": 1594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9815418782227817e-05,
+      "loss": 2.0242,
+      "step": 1595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9815048383333542e-05,
+      "loss": 2.0347,
+      "step": 1596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9814677615599685e-05,
+      "loss": 2.1282,
+      "step": 1597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981430647903598e-05,
+      "loss": 2.0935,
+      "step": 1598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9813934973652147e-05,
+      "loss": 2.0642,
+      "step": 1599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9813563099457942e-05,
+      "loss": 2.0542,
+      "step": 1600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9813190856463108e-05,
+      "loss": 2.0671,
+      "step": 1601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981281824467741e-05,
+      "loss": 2.0688,
+      "step": 1602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981244526411062e-05,
+      "loss": 2.0085,
+      "step": 1603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981207191477251e-05,
+      "loss": 2.0525,
+      "step": 1604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9811698196672877e-05,
+      "loss": 2.0742,
+      "step": 1605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9811324109821528e-05,
+      "loss": 2.0146,
+      "step": 1606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9810949654228262e-05,
+      "loss": 2.0957,
+      "step": 1607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9810574829902906e-05,
+      "loss": 2.0957,
+      "step": 1608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.981019963685529e-05,
+      "loss": 2.0798,
+      "step": 1609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9809824075095244e-05,
+      "loss": 2.0884,
+      "step": 1610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9809448144632625e-05,
+      "loss": 2.0061,
+      "step": 1611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980907184547729e-05,
+      "loss": 1.991,
+      "step": 1612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980869517763911e-05,
+      "loss": 2.0828,
+      "step": 1613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980831814112795e-05,
+      "loss": 2.083,
+      "step": 1614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9807940735953713e-05,
+      "loss": 1.9756,
+      "step": 1615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9807562962126285e-05,
+      "loss": 1.9827,
+      "step": 1616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980718481965558e-05,
+      "loss": 2.0291,
+      "step": 1617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9806806308551508e-05,
+      "loss": 2.0171,
+      "step": 1618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9806427428823996e-05,
+      "loss": 2.021,
+      "step": 1619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9806048180482985e-05,
+      "loss": 2.0251,
+      "step": 1620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980566856353842e-05,
+      "loss": 2.0745,
+      "step": 1621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9805288578000248e-05,
+      "loss": 2.0884,
+      "step": 1622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980490822387844e-05,
+      "loss": 2.0544,
+      "step": 1623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980452750118297e-05,
+      "loss": 2.0828,
+      "step": 1624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980414640992382e-05,
+      "loss": 2.1008,
+      "step": 1625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980376495011099e-05,
+      "loss": 2.0532,
+      "step": 1626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980338312175447e-05,
+      "loss": 2.0686,
+      "step": 1627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9803000924864286e-05,
+      "loss": 2.0432,
+      "step": 1628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9802618359450456e-05,
+      "loss": 2.0579,
+      "step": 1629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9802235425523013e-05,
+      "loss": 1.9954,
+      "step": 1630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9801852123091996e-05,
+      "loss": 2.0264,
+      "step": 1631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.980146845216746e-05,
+      "loss": 2.073,
+      "step": 1632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9801084412759465e-05,
+      "loss": 2.1304,
+      "step": 1633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9800700004878077e-05,
+      "loss": 2.0928,
+      "step": 1634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9800315228533387e-05,
+      "loss": 2.0273,
+      "step": 1635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979993008373548e-05,
+      "loss": 2.0347,
+      "step": 1636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9799544570494456e-05,
+      "loss": 2.0718,
+      "step": 1637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9799158688820423e-05,
+      "loss": 2.0464,
+      "step": 1638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9798772438723504e-05,
+      "loss": 2.1013,
+      "step": 1639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9798385820213822e-05,
+      "loss": 2.1099,
+      "step": 1640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9797998833301526e-05,
+      "loss": 2.1155,
+      "step": 1641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9797611477996753e-05,
+      "loss": 2.0967,
+      "step": 1642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979722375430967e-05,
+      "loss": 2.0828,
+      "step": 1643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9796835662250434e-05,
+      "loss": 2.0723,
+      "step": 1644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9796447201829232e-05,
+      "loss": 2.1077,
+      "step": 1645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979605837305625e-05,
+      "loss": 2.0835,
+      "step": 1646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979566917594168e-05,
+      "loss": 2.0757,
+      "step": 1647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979527961049573e-05,
+      "loss": 2.0896,
+      "step": 1648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979488967672862e-05,
+      "loss": 2.0776,
+      "step": 1649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9794499374650568e-05,
+      "loss": 2.1006,
+      "step": 1650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9794108704271814e-05,
+      "loss": 2.1072,
+      "step": 1651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9793717665602602e-05,
+      "loss": 2.0872,
+      "step": 1652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9793326258653185e-05,
+      "loss": 2.1008,
+      "step": 1653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9792934483433833e-05,
+      "loss": 2.1091,
+      "step": 1654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9792542339954808e-05,
+      "loss": 2.0984,
+      "step": 1655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979214982822641e-05,
+      "loss": 2.0728,
+      "step": 1656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979175694825892e-05,
+      "loss": 2.0935,
+      "step": 1657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9791363700062644e-05,
+      "loss": 2.0767,
+      "step": 1658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9790970083647893e-05,
+      "loss": 2.0439,
+      "step": 1659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979057609902499e-05,
+      "loss": 2.0396,
+      "step": 1660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979018174620427e-05,
+      "loss": 2.04,
+      "step": 1661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9789787025196067e-05,
+      "loss": 2.0701,
+      "step": 1662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9789391936010745e-05,
+      "loss": 2.0894,
+      "step": 1663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9788996478658652e-05,
+      "loss": 2.0461,
+      "step": 1664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9788600653150163e-05,
+      "loss": 2.0874,
+      "step": 1665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978820445949566e-05,
+      "loss": 2.0552,
+      "step": 1666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978780789770553e-05,
+      "loss": 2.0796,
+      "step": 1667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9787410967790172e-05,
+      "loss": 2.053,
+      "step": 1668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9787013669759997e-05,
+      "loss": 2.0149,
+      "step": 1669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9786616003625427e-05,
+      "loss": 2.0383,
+      "step": 1670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978621796939688e-05,
+      "loss": 2.0588,
+      "step": 1671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9785819567084805e-05,
+      "loss": 2.0759,
+      "step": 1672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9785420796699643e-05,
+      "loss": 2.054,
+      "step": 1673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9785021658251852e-05,
+      "loss": 2.0527,
+      "step": 1674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9784622151751905e-05,
+      "loss": 2.0776,
+      "step": 1675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978422227721027e-05,
+      "loss": 2.0806,
+      "step": 1676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978382203463744e-05,
+      "loss": 2.0759,
+      "step": 1677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9783421424043905e-05,
+      "loss": 2.0493,
+      "step": 1678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9783020445440174e-05,
+      "loss": 2.0469,
+      "step": 1679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978261909883676e-05,
+      "loss": 2.0454,
+      "step": 1680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9782217384244195e-05,
+      "loss": 2.0603,
+      "step": 1681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9781815301673004e-05,
+      "loss": 2.0715,
+      "step": 1682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9781412851133734e-05,
+      "loss": 2.0779,
+      "step": 1683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9781010032636948e-05,
+      "loss": 2.1018,
+      "step": 1684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9780606846193193e-05,
+      "loss": 2.0564,
+      "step": 1685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9780203291813053e-05,
+      "loss": 2.0549,
+      "step": 1686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9779799369507105e-05,
+      "loss": 2.0549,
+      "step": 1687
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977939507928595e-05,
+      "loss": 2.0657,
+      "step": 1688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9778990421160184e-05,
+      "loss": 2.0681,
+      "step": 1689
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9778585395140418e-05,
+      "loss": 2.0647,
+      "step": 1690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9778180001237275e-05,
+      "loss": 2.0198,
+      "step": 1691
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9777774239461385e-05,
+      "loss": 2.0361,
+      "step": 1692
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9777368109823388e-05,
+      "loss": 2.063,
+      "step": 1693
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977696161233394e-05,
+      "loss": 2.0813,
+      "step": 1694
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9776554747003692e-05,
+      "loss": 2.0903,
+      "step": 1695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977614751384332e-05,
+      "loss": 2.0786,
+      "step": 1696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9775739912863505e-05,
+      "loss": 2.0571,
+      "step": 1697
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977533194407493e-05,
+      "loss": 2.0239,
+      "step": 1698
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9774923607488297e-05,
+      "loss": 2.0857,
+      "step": 1699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9774514903114307e-05,
+      "loss": 2.0964,
+      "step": 1700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9774105830963693e-05,
+      "loss": 2.0593,
+      "step": 1701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9773696391047165e-05,
+      "loss": 2.0635,
+      "step": 1702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.977328658337547e-05,
+      "loss": 2.0793,
+      "step": 1703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9772876407959353e-05,
+      "loss": 2.1086,
+      "step": 1704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9772465864809574e-05,
+      "loss": 1.9724,
+      "step": 1705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9772054953936894e-05,
+      "loss": 2.0303,
+      "step": 1706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9771643675352083e-05,
+      "loss": 2.0874,
+      "step": 1707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.977123202906594e-05,
+      "loss": 2.0825,
+      "step": 1708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9770820015089253e-05,
+      "loss": 2.0408,
+      "step": 1709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.977040763343283e-05,
+      "loss": 2.0322,
+      "step": 1710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9769994884107476e-05,
+      "loss": 2.0859,
+      "step": 1711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976958176712402e-05,
+      "loss": 2.0894,
+      "step": 1712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97691682824933e-05,
+      "loss": 2.0605,
+      "step": 1713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9768754430226157e-05,
+      "loss": 2.0662,
+      "step": 1714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976834021033344e-05,
+      "loss": 2.1084,
+      "step": 1715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9767925622826015e-05,
+      "loss": 2.0952,
+      "step": 1716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9767510667714754e-05,
+      "loss": 2.085,
+      "step": 1717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9767095345010534e-05,
+      "loss": 2.0386,
+      "step": 1718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9766679654724254e-05,
+      "loss": 2.053,
+      "step": 1719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9766263596866807e-05,
+      "loss": 2.0325,
+      "step": 1720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9765847171449112e-05,
+      "loss": 2.1021,
+      "step": 1721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976543037848208e-05,
+      "loss": 2.0942,
+      "step": 1722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976501321797665e-05,
+      "loss": 2.0833,
+      "step": 1723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9764595689943754e-05,
+      "loss": 2.0837,
+      "step": 1724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9764177794394348e-05,
+      "loss": 2.0767,
+      "step": 1725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9763759531339386e-05,
+      "loss": 2.0669,
+      "step": 1726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976334090078984e-05,
+      "loss": 2.1479,
+      "step": 1727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976292190275668e-05,
+      "loss": 2.1133,
+      "step": 1728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9762502537250905e-05,
+      "loss": 2.0757,
+      "step": 1729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9762082804283504e-05,
+      "loss": 2.0813,
+      "step": 1730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9761662703865486e-05,
+      "loss": 2.1055,
+      "step": 1731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9761242236007872e-05,
+      "loss": 2.0896,
+      "step": 1732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9760821400721688e-05,
+      "loss": 2.0938,
+      "step": 1733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9760400198017957e-05,
+      "loss": 2.0693,
+      "step": 1734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9759978627907746e-05,
+      "loss": 2.0752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975955669040209e-05,
+      "loss": 2.0762,
+      "step": 1736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9759134385512064e-05,
+      "loss": 2.0811,
+      "step": 1737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9758711713248742e-05,
+      "loss": 2.0593,
+      "step": 1738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9758288673623212e-05,
+      "loss": 2.0386,
+      "step": 1739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9757865266646557e-05,
+      "loss": 2.0391,
+      "step": 1740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9757441492329887e-05,
+      "loss": 2.1169,
+      "step": 1741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9757017350684318e-05,
+      "loss": 2.1123,
+      "step": 1742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9756592841720964e-05,
+      "loss": 2.0093,
+      "step": 1743
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9756167965450963e-05,
+      "loss": 2.012,
+      "step": 1744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9755742721885454e-05,
+      "loss": 2.0532,
+      "step": 1745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975531711103559e-05,
+      "loss": 2.0718,
+      "step": 1746
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9754891132912535e-05,
+      "loss": 2.0471,
+      "step": 1747
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975446478752746e-05,
+      "loss": 2.0654,
+      "step": 1748
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975403807489154e-05,
+      "loss": 2.0977,
+      "step": 1749
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9753610995015966e-05,
+      "loss": 2.1094,
+      "step": 1750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9753183547911944e-05,
+      "loss": 2.0989,
+      "step": 1751
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9752755733590676e-05,
+      "loss": 2.063,
+      "step": 1752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975232755206338e-05,
+      "loss": 2.0815,
+      "step": 1753
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975189900334129e-05,
+      "loss": 2.0974,
+      "step": 1754
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975147008743565e-05,
+      "loss": 2.0203,
+      "step": 1755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9751040804357696e-05,
+      "loss": 2.0234,
+      "step": 1756
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9750611154118684e-05,
+      "loss": 2.1143,
+      "step": 1757
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975018113672989e-05,
+      "loss": 2.0693,
+      "step": 1758
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9749750752202592e-05,
+      "loss": 2.074,
+      "step": 1759
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974932000054807e-05,
+      "loss": 2.0764,
+      "step": 1760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9748888881777618e-05,
+      "loss": 2.0349,
+      "step": 1761
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974845739590255e-05,
+      "loss": 2.0361,
+      "step": 1762
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974802554293417e-05,
+      "loss": 2.075,
+      "step": 1763
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9747593322883813e-05,
+      "loss": 2.0891,
+      "step": 1764
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9747160735762812e-05,
+      "loss": 2.0576,
+      "step": 1765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9746727781582506e-05,
+      "loss": 2.0476,
+      "step": 1766
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9746294460354255e-05,
+      "loss": 2.0298,
+      "step": 1767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9745860772089417e-05,
+      "loss": 2.0149,
+      "step": 1768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9745426716799364e-05,
+      "loss": 2.084,
+      "step": 1769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974499229449549e-05,
+      "loss": 2.1072,
+      "step": 1770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974455750518917e-05,
+      "loss": 2.0798,
+      "step": 1771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974412234889182e-05,
+      "loss": 2.1052,
+      "step": 1772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9743686825614845e-05,
+      "loss": 2.1074,
+      "step": 1773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9743250935369665e-05,
+      "loss": 2.0679,
+      "step": 1774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9742814678167714e-05,
+      "loss": 2.0437,
+      "step": 1775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9742378054020436e-05,
+      "loss": 2.0334,
+      "step": 1776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974194106293927e-05,
+      "loss": 2.0312,
+      "step": 1777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9741503704935684e-05,
+      "loss": 2.0496,
+      "step": 1778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9741065980021148e-05,
+      "loss": 2.0781,
+      "step": 1779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9740627888207136e-05,
+      "loss": 2.094,
+      "step": 1780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974018942950514e-05,
+      "loss": 2.0586,
+      "step": 1781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973975060392665e-05,
+      "loss": 2.0762,
+      "step": 1782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9739311411483187e-05,
+      "loss": 2.0767,
+      "step": 1783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9738871852186257e-05,
+      "loss": 2.0869,
+      "step": 1784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973843192604739e-05,
+      "loss": 1.9822,
+      "step": 1785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973799163307813e-05,
+      "loss": 2.0183,
+      "step": 1786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9737550973290016e-05,
+      "loss": 2.0129,
+      "step": 1787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97371099466946e-05,
+      "loss": 2.0083,
+      "step": 1788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9736668553303458e-05,
+      "loss": 2.0625,
+      "step": 1789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9736226793128153e-05,
+      "loss": 2.0483,
+      "step": 1790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973578466618028e-05,
+      "loss": 2.0991,
+      "step": 1791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973534217247143e-05,
+      "loss": 2.0701,
+      "step": 1792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9734899312013205e-05,
+      "loss": 2.0911,
+      "step": 1793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9734456084817218e-05,
+      "loss": 2.0925,
+      "step": 1794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9734012490895095e-05,
+      "loss": 2.0989,
+      "step": 1795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973356853025847e-05,
+      "loss": 2.0701,
+      "step": 1796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973312420291898e-05,
+      "loss": 2.0879,
+      "step": 1797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973267950888828e-05,
+      "loss": 2.0686,
+      "step": 1798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9732234448178034e-05,
+      "loss": 2.0798,
+      "step": 1799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9731789020799906e-05,
+      "loss": 2.1013,
+      "step": 1800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9731343226765586e-05,
+      "loss": 2.0974,
+      "step": 1801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9730897066086756e-05,
+      "loss": 2.0903,
+      "step": 1802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9730450538775123e-05,
+      "loss": 2.0769,
+      "step": 1803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973000364484239e-05,
+      "loss": 2.0691,
+      "step": 1804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9729556384300282e-05,
+      "loss": 2.1023,
+      "step": 1805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9729108757160527e-05,
+      "loss": 2.1008,
+      "step": 1806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9728660763434857e-05,
+      "loss": 2.0845,
+      "step": 1807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9728212403135032e-05,
+      "loss": 2.1047,
+      "step": 1808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97277636762728e-05,
+      "loss": 2.0989,
+      "step": 1809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9727314582859933e-05,
+      "loss": 2.1145,
+      "step": 1810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9726865122908203e-05,
+      "loss": 2.0618,
+      "step": 1811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9726415296429403e-05,
+      "loss": 2.0549,
+      "step": 1812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9725965103435324e-05,
+      "loss": 2.1445,
+      "step": 1813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9725514543937776e-05,
+      "loss": 2.0967,
+      "step": 1814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.972506361794857e-05,
+      "loss": 2.0652,
+      "step": 1815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.972461232547953e-05,
+      "loss": 2.0552,
+      "step": 1816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97241606665425e-05,
+      "loss": 2.1177,
+      "step": 1817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9723708641149318e-05,
+      "loss": 2.0994,
+      "step": 1818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9723256249311836e-05,
+      "loss": 2.0706,
+      "step": 1819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9722803491041917e-05,
+      "loss": 2.0688,
+      "step": 1820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.972235036635144e-05,
+      "loss": 2.0435,
+      "step": 1821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9721896875252283e-05,
+      "loss": 2.0217,
+      "step": 1822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9721443017756342e-05,
+      "loss": 2.0977,
+      "step": 1823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9720988793875514e-05,
+      "loss": 2.1099,
+      "step": 1824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9720534203621714e-05,
+      "loss": 2.0408,
+      "step": 1825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9720079247006862e-05,
+      "loss": 2.0618,
+      "step": 1826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971962392404288e-05,
+      "loss": 2.0154,
+      "step": 1827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9719168234741728e-05,
+      "loss": 2.0154,
+      "step": 1828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9718712179115342e-05,
+      "loss": 2.0913,
+      "step": 1829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9718255757175684e-05,
+      "loss": 2.116,
+      "step": 1830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9717798968934724e-05,
+      "loss": 2.0857,
+      "step": 1831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9717341814404438e-05,
+      "loss": 2.0334,
+      "step": 1832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9716884293596818e-05,
+      "loss": 2.1047,
+      "step": 1833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971642640652386e-05,
+      "loss": 2.1045,
+      "step": 1834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9715968153197578e-05,
+      "loss": 2.0305,
+      "step": 1835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9715509533629976e-05,
+      "loss": 2.0286,
+      "step": 1836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971505054783309e-05,
+      "loss": 2.0759,
+      "step": 1837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971459119581895e-05,
+      "loss": 2.071,
+      "step": 1838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9714131477599614e-05,
+      "loss": 2.0476,
+      "step": 1839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9713671393187128e-05,
+      "loss": 2.0359,
+      "step": 1840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9713210942593557e-05,
+      "loss": 2.0691,
+      "step": 1841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971275012583098e-05,
+      "loss": 2.0872,
+      "step": 1842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9712288942911477e-05,
+      "loss": 2.0544,
+      "step": 1843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971182739384715e-05,
+      "loss": 2.0669,
+      "step": 1844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971136547865009e-05,
+      "loss": 2.0549,
+      "step": 1845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9710903197332417e-05,
+      "loss": 2.0361,
+      "step": 1846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.971044054990626e-05,
+      "loss": 2.0891,
+      "step": 1847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.970997753638374e-05,
+      "loss": 2.0588,
+      "step": 1848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9709514156777005e-05,
+      "loss": 2.0327,
+      "step": 1849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9709050411098208e-05,
+      "loss": 2.0447,
+      "step": 1850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.970858629935951e-05,
+      "loss": 2.0984,
+      "step": 1851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9708121821573077e-05,
+      "loss": 2.0928,
+      "step": 1852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9707656977751086e-05,
+      "loss": 2.0786,
+      "step": 1853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9707191767905743e-05,
+      "loss": 2.0806,
+      "step": 1854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9706726192049233e-05,
+      "loss": 2.0271,
+      "step": 1855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.970626025019377e-05,
+      "loss": 2.0315,
+      "step": 1856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9705793942351574e-05,
+      "loss": 2.0579,
+      "step": 1857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9705327268534878e-05,
+      "loss": 2.0527,
+      "step": 1858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.970486022875591e-05,
+      "loss": 2.0696,
+      "step": 1859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9704392823026913e-05,
+      "loss": 2.0349,
+      "step": 1860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.970392505136016e-05,
+      "loss": 2.0635,
+      "step": 1861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9703456913767916e-05,
+      "loss": 2.0669,
+      "step": 1862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9702988410262445e-05,
+      "loss": 2.0544,
+      "step": 1863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9702519540856044e-05,
+      "loss": 2.0491,
+      "step": 1864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9702050305560998e-05,
+      "loss": 2.1145,
+      "step": 1865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9701580704389623e-05,
+      "loss": 2.1304,
+      "step": 1866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9701110737354232e-05,
+      "loss": 2.0288,
+      "step": 1867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9700640404467144e-05,
+      "loss": 2.0178,
+      "step": 1868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9700169705740695e-05,
+      "loss": 2.1111,
+      "step": 1869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9699698641187228e-05,
+      "loss": 2.0869,
+      "step": 1870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.96992272108191e-05,
+      "loss": 2.0305,
+      "step": 1871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969875541464867e-05,
+      "loss": 2.0659,
+      "step": 1872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9698283252688312e-05,
+      "loss": 2.0469,
+      "step": 1873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969781072495041e-05,
+      "loss": 2.0688,
+      "step": 1874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969733783144735e-05,
+      "loss": 2.0945,
+      "step": 1875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9696864572191534e-05,
+      "loss": 2.0686,
+      "step": 1876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9696390947195378e-05,
+      "loss": 2.0503,
+      "step": 1877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.96959169564713e-05,
+      "loss": 1.9978,
+      "step": 1878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9695442600031725e-05,
+      "loss": 2.0774,
+      "step": 1879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9694967877889096e-05,
+      "loss": 2.061,
+      "step": 1880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9694492790055863e-05,
+      "loss": 2.0679,
+      "step": 1881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9694017336544483e-05,
+      "loss": 2.0615,
+      "step": 1882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9693541517367428e-05,
+      "loss": 2.0383,
+      "step": 1883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969306533253717e-05,
+      "loss": 2.0288,
+      "step": 1884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.96925887820662e-05,
+      "loss": 1.9714,
+      "step": 1885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9692111865967017e-05,
+      "loss": 1.9595,
+      "step": 1886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969163458425212e-05,
+      "loss": 2.0691,
+      "step": 1887
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969115693693403e-05,
+      "loss": 2.0857,
+      "step": 1888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9690678924025274e-05,
+      "loss": 2.1001,
+      "step": 1889
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.969020054553839e-05,
+      "loss": 2.1038,
+      "step": 1890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9689721801485918e-05,
+      "loss": 2.0051,
+      "step": 1891
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9689242691880406e-05,
+      "loss": 2.0076,
+      "step": 1892
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9688763216734435e-05,
+      "loss": 2.0586,
+      "step": 1893
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9688283376060564e-05,
+      "loss": 2.0527,
+      "step": 1894
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.968780316987138e-05,
+      "loss": 2.0991,
+      "step": 1895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9687322598179478e-05,
+      "loss": 2.1018,
+      "step": 1896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9686841660997462e-05,
+      "loss": 2.1174,
+      "step": 1897
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.968636035833794e-05,
+      "loss": 2.1094,
+      "step": 1898
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9685878690213538e-05,
+      "loss": 2.0376,
+      "step": 1899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9685396656636886e-05,
+      "loss": 2.0188,
+      "step": 1900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9684914257620615e-05,
+      "loss": 2.1465,
+      "step": 1901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.968443149317739e-05,
+      "loss": 2.1162,
+      "step": 1902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9683948363319865e-05,
+      "loss": 2.0391,
+      "step": 1903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.968346486806071e-05,
+      "loss": 2.0234,
+      "step": 1904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96829810074126e-05,
+      "loss": 2.094,
+      "step": 1905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.968249678138822e-05,
+      "loss": 2.0977,
+      "step": 1906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.968201219000029e-05,
+      "loss": 2.0896,
+      "step": 1907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9681527233261494e-05,
+      "loss": 2.0784,
+      "step": 1908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9681041911184558e-05,
+      "loss": 2.1042,
+      "step": 1909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9680556223782216e-05,
+      "loss": 2.0715,
+      "step": 1910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9680070171067193e-05,
+      "loss": 2.0654,
+      "step": 1911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.967958375305224e-05,
+      "loss": 2.0452,
+      "step": 1912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9679096969750116e-05,
+      "loss": 2.0623,
+      "step": 1913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9678609821173582e-05,
+      "loss": 2.0793,
+      "step": 1914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.967812230733541e-05,
+      "loss": 2.0439,
+      "step": 1915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.967763442824839e-05,
+      "loss": 2.0781,
+      "step": 1916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.967714618392532e-05,
+      "loss": 2.0789,
+      "step": 1917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9676657574378997e-05,
+      "loss": 2.0933,
+      "step": 1918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9676168599622234e-05,
+      "loss": 2.1128,
+      "step": 1919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9675679259667857e-05,
+      "loss": 2.0933,
+      "step": 1920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9675189554528696e-05,
+      "loss": 2.0872,
+      "step": 1921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9674699484217592e-05,
+      "loss": 2.0479,
+      "step": 1922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96742090487474e-05,
+      "loss": 2.127,
+      "step": 1923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9673718248130976e-05,
+      "loss": 2.0964,
+      "step": 1924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9673227082381198e-05,
+      "loss": 2.0928,
+      "step": 1925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9672735551510942e-05,
+      "loss": 2.0645,
+      "step": 1926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9672243655533097e-05,
+      "loss": 2.0662,
+      "step": 1927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9671751394460562e-05,
+      "loss": 2.0588,
+      "step": 1928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9671258768306253e-05,
+      "loss": 2.0637,
+      "step": 1929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9670765777083074e-05,
+      "loss": 2.0503,
+      "step": 1930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9670272420803972e-05,
+      "loss": 2.0757,
+      "step": 1931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.966977869948187e-05,
+      "loss": 2.0708,
+      "step": 1932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9669284613129722e-05,
+      "loss": 2.0071,
+      "step": 1933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.966879016176048e-05,
+      "loss": 2.0288,
+      "step": 1934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9668295345387115e-05,
+      "loss": 2.082,
+      "step": 1935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9667800164022598e-05,
+      "loss": 2.0779,
+      "step": 1936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9667304617679926e-05,
+      "loss": 2.0615,
+      "step": 1937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9666808706372078e-05,
+      "loss": 2.0701,
+      "step": 1938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9666312430112072e-05,
+      "loss": 2.0654,
+      "step": 1939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.966581578891291e-05,
+      "loss": 2.0352,
+      "step": 1940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.966531878278763e-05,
+      "loss": 2.0779,
+      "step": 1941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9664821411749257e-05,
+      "loss": 2.0962,
+      "step": 1942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9664323675810834e-05,
+      "loss": 2.0288,
+      "step": 1943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.966382557498542e-05,
+      "loss": 2.0098,
+      "step": 1944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9663327109286063e-05,
+      "loss": 2.0642,
+      "step": 1945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9662828278725847e-05,
+      "loss": 2.0605,
+      "step": 1946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9662329083317855e-05,
+      "loss": 2.106,
+      "step": 1947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9661829523075165e-05,
+      "loss": 2.0737,
+      "step": 1948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9661329598010888e-05,
+      "loss": 2.0818,
+      "step": 1949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9660829308138133e-05,
+      "loss": 2.0693,
+      "step": 1950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9660328653470016e-05,
+      "loss": 2.0967,
+      "step": 1951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965982763401967e-05,
+      "loss": 2.0571,
+      "step": 1952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9659326249800228e-05,
+      "loss": 2.0723,
+      "step": 1953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965882450082484e-05,
+      "loss": 2.0793,
+      "step": 1954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9658322387106668e-05,
+      "loss": 2.1111,
+      "step": 1955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9657819908658875e-05,
+      "loss": 2.0771,
+      "step": 1956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9657317065494642e-05,
+      "loss": 2.1084,
+      "step": 1957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965681385762715e-05,
+      "loss": 2.0989,
+      "step": 1958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9656310285069602e-05,
+      "loss": 2.1304,
+      "step": 1959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9655806347835196e-05,
+      "loss": 2.0684,
+      "step": 1960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9655302045937154e-05,
+      "loss": 2.0195,
+      "step": 1961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9654797379388698e-05,
+      "loss": 2.0391,
+      "step": 1962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9654292348203057e-05,
+      "loss": 2.0435,
+      "step": 1963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9653786952393484e-05,
+      "loss": 2.0769,
+      "step": 1964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9653281191973225e-05,
+      "loss": 2.1047,
+      "step": 1965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965277506695555e-05,
+      "loss": 2.1042,
+      "step": 1966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9652268577353723e-05,
+      "loss": 2.0852,
+      "step": 1967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9651761723181036e-05,
+      "loss": 2.0779,
+      "step": 1968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.965125450445077e-05,
+      "loss": 2.0991,
+      "step": 1969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9650746921176235e-05,
+      "loss": 2.1101,
+      "step": 1970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9650238973370737e-05,
+      "loss": 2.0549,
+      "step": 1971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96497306610476e-05,
+      "loss": 2.0923,
+      "step": 1972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9649221984220153e-05,
+      "loss": 2.1025,
+      "step": 1973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9648712942901728e-05,
+      "loss": 2.0781,
+      "step": 1974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9648203537105682e-05,
+      "loss": 2.0996,
+      "step": 1975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9647693766845377e-05,
+      "loss": 2.0825,
+      "step": 1976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9647183632134173e-05,
+      "loss": 2.084,
+      "step": 1977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9646673132985448e-05,
+      "loss": 2.0403,
+      "step": 1978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9646162269412592e-05,
+      "loss": 2.0911,
+      "step": 1979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9645651041429007e-05,
+      "loss": 2.0935,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9645139449048088e-05,
+      "loss": 2.0957,
+      "step": 1981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.964462749228326e-05,
+      "loss": 2.0791,
+      "step": 1982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9644115171147944e-05,
+      "loss": 2.0693,
+      "step": 1983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9643602485655577e-05,
+      "loss": 2.0833,
+      "step": 1984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9643089435819604e-05,
+      "loss": 2.0715,
+      "step": 1985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9642576021653476e-05,
+      "loss": 2.082,
+      "step": 1986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.964206224317066e-05,
+      "loss": 2.0688,
+      "step": 1987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9641548100384624e-05,
+      "loss": 2.0771,
+      "step": 1988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9641033593308863e-05,
+      "loss": 2.1292,
+      "step": 1989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9640518721956855e-05,
+      "loss": 2.1208,
+      "step": 1990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.964000348634211e-05,
+      "loss": 2.0454,
+      "step": 1991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9639487886478137e-05,
+      "loss": 2.0493,
+      "step": 1992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.963897192237846e-05,
+      "loss": 2.0955,
+      "step": 1993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9638455594056604e-05,
+      "loss": 2.1011,
+      "step": 1994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9637938901526114e-05,
+      "loss": 2.0557,
+      "step": 1995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9637421844800537e-05,
+      "loss": 2.0813,
+      "step": 1996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9636904423893432e-05,
+      "loss": 2.0642,
+      "step": 1997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9636386638818372e-05,
+      "loss": 2.0603,
+      "step": 1998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.963586848958893e-05,
+      "loss": 2.0359,
+      "step": 1999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9635349976218697e-05,
+      "loss": 2.0576,
+      "step": 2000
+    },
+    {
+      "Bleu_1": 0.7707147270088032,
+      "Bleu_2": 0.6108141146166212,
+      "Bleu_3": 0.4694944458277585,
+      "Bleu_4": 0.3561662364085532,
+      "CIDEr": 1.1974656464004014,
+      "METEOR": 0.2780508251704735,
+      "ROUGE_L": 0.5571833692224261,
+      "epoch": 0.1,
+      "step": 2000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9634831098721268e-05,
+      "loss": 1.9958,
+      "step": 2001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9634311857110253e-05,
+      "loss": 1.9651,
+      "step": 2002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9633792251399266e-05,
+      "loss": 2.0276,
+      "step": 2003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9633272281601937e-05,
+      "loss": 1.9927,
+      "step": 2004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9632751947731895e-05,
+      "loss": 2.0627,
+      "step": 2005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9632231249802788e-05,
+      "loss": 2.0474,
+      "step": 2006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9631710187828272e-05,
+      "loss": 2.0837,
+      "step": 2007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9631188761822014e-05,
+      "loss": 2.0234,
+      "step": 2008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9630666971797683e-05,
+      "loss": 2.0522,
+      "step": 2009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.963014481776896e-05,
+      "loss": 2.0569,
+      "step": 2010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9629622299749546e-05,
+      "loss": 2.0449,
+      "step": 2011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9629099417753137e-05,
+      "loss": 2.0144,
+      "step": 2012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962857617179345e-05,
+      "loss": 2.1416,
+      "step": 2013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96280525618842e-05,
+      "loss": 2.1047,
+      "step": 2014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962752858803913e-05,
+      "loss": 2.0974,
+      "step": 2015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962700425027196e-05,
+      "loss": 2.0823,
+      "step": 2016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962647954859646e-05,
+      "loss": 2.0625,
+      "step": 2017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9625954483026377e-05,
+      "loss": 2.0544,
+      "step": 2018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962542905357549e-05,
+      "loss": 2.0481,
+      "step": 2019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962490326025757e-05,
+      "loss": 2.0447,
+      "step": 2020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9624377103086414e-05,
+      "loss": 2.041,
+      "step": 2021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962385058207581e-05,
+      "loss": 2.0596,
+      "step": 2022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9623323697239573e-05,
+      "loss": 2.0601,
+      "step": 2023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9622796448591515e-05,
+      "loss": 2.064,
+      "step": 2024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9622268836145462e-05,
+      "loss": 2.0723,
+      "step": 2025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9621740859915252e-05,
+      "loss": 2.0535,
+      "step": 2026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.962121251991473e-05,
+      "loss": 2.0557,
+      "step": 2027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9620683816157755e-05,
+      "loss": 2.02,
+      "step": 2028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9620154748658187e-05,
+      "loss": 2.061,
+      "step": 2029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9619625317429902e-05,
+      "loss": 2.0557,
+      "step": 2030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9619095522486783e-05,
+      "loss": 2.1135,
+      "step": 2031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9618565363842724e-05,
+      "loss": 2.1064,
+      "step": 2032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9618034841511627e-05,
+      "loss": 2.0476,
+      "step": 2033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961750395550741e-05,
+      "loss": 2.0415,
+      "step": 2034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961697270584398e-05,
+      "loss": 2.0415,
+      "step": 2035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9616441092535285e-05,
+      "loss": 2.071,
+      "step": 2036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9615909115595254e-05,
+      "loss": 2.0515,
+      "step": 2037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9615376775037846e-05,
+      "loss": 2.0071,
+      "step": 2038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961484407087702e-05,
+      "loss": 2.1201,
+      "step": 2039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9614311003126743e-05,
+      "loss": 2.0911,
+      "step": 2040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961377757180099e-05,
+      "loss": 2.0793,
+      "step": 2041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961324377691375e-05,
+      "loss": 2.0745,
+      "step": 2042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9612709618479036e-05,
+      "loss": 2.0825,
+      "step": 2043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9612175096510835e-05,
+      "loss": 2.0662,
+      "step": 2044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9611640211023177e-05,
+      "loss": 2.0896,
+      "step": 2045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9611104962030083e-05,
+      "loss": 2.0837,
+      "step": 2046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96105693495456e-05,
+      "loss": 2.0747,
+      "step": 2047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.961003337358376e-05,
+      "loss": 2.092,
+      "step": 2048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.960949703415862e-05,
+      "loss": 2.0007,
+      "step": 2049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9608960331284257e-05,
+      "loss": 2.0229,
+      "step": 2050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9608423264974734e-05,
+      "loss": 1.9824,
+      "step": 2051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9607885835244136e-05,
+      "loss": 2.022,
+      "step": 2052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9607348042106558e-05,
+      "loss": 2.083,
+      "step": 2053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9606809885576108e-05,
+      "loss": 2.0884,
+      "step": 2054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.960627136566689e-05,
+      "loss": 1.9951,
+      "step": 2055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9605732482393033e-05,
+      "loss": 2.0442,
+      "step": 2056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9605193235768666e-05,
+      "loss": 2.0754,
+      "step": 2057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.960465362580793e-05,
+      "loss": 2.0811,
+      "step": 2058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9604113652524976e-05,
+      "loss": 2.0496,
+      "step": 2059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9603573315933964e-05,
+      "loss": 2.0613,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.960303261604906e-05,
+      "loss": 2.0718,
+      "step": 2061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9602491552884448e-05,
+      "loss": 2.0789,
+      "step": 2062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9601950126454317e-05,
+      "loss": 2.0342,
+      "step": 2063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9601408336772866e-05,
+      "loss": 2.0352,
+      "step": 2064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.96008661838543e-05,
+      "loss": 2.0754,
+      "step": 2065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.960032366771283e-05,
+      "loss": 2.0554,
+      "step": 2066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9599780788362695e-05,
+      "loss": 2.0596,
+      "step": 2067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.959923754581813e-05,
+      "loss": 2.0793,
+      "step": 2068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9598693940093373e-05,
+      "loss": 2.0593,
+      "step": 2069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9598149971202683e-05,
+      "loss": 2.0652,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9597605639160325e-05,
+      "loss": 2.041,
+      "step": 2071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9597060943980573e-05,
+      "loss": 2.0718,
+      "step": 2072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9596515885677716e-05,
+      "loss": 2.042,
+      "step": 2073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.959597046426604e-05,
+      "loss": 2.0476,
+      "step": 2074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.959542467975985e-05,
+      "loss": 2.061,
+      "step": 2075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9594878532173464e-05,
+      "loss": 2.0667,
+      "step": 2076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.95943320215212e-05,
+      "loss": 2.0422,
+      "step": 2077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9593785147817387e-05,
+      "loss": 2.0454,
+      "step": 2078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.959323791107637e-05,
+      "loss": 2.0659,
+      "step": 2079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9592690311312497e-05,
+      "loss": 2.0679,
+      "step": 2080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9592142348540135e-05,
+      "loss": 2.0073,
+      "step": 2081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9591594022773643e-05,
+      "loss": 2.0337,
+      "step": 2082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9591045334027407e-05,
+      "loss": 2.0391,
+      "step": 2083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9590496282315817e-05,
+      "loss": 2.0491,
+      "step": 2084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.958994686765327e-05,
+      "loss": 2.0759,
+      "step": 2085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9589397090054165e-05,
+      "loss": 2.0535,
+      "step": 2086
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.958884694953293e-05,
+      "loss": 1.9802,
+      "step": 2087
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9588296446103988e-05,
+      "loss": 2.009,
+      "step": 2088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9587745579781775e-05,
+      "loss": 2.0828,
+      "step": 2089
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.958719435058074e-05,
+      "loss": 2.0947,
+      "step": 2090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9586642758515335e-05,
+      "loss": 2.0901,
+      "step": 2091
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9586090803600026e-05,
+      "loss": 2.071,
+      "step": 2092
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9585538485849283e-05,
+      "loss": 2.0925,
+      "step": 2093
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.95849858052776e-05,
+      "loss": 2.0522,
+      "step": 2094
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9584432761899463e-05,
+      "loss": 2.0444,
+      "step": 2095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9583879355729375e-05,
+      "loss": 2.0608,
+      "step": 2096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9583325586781847e-05,
+      "loss": 2.0574,
+      "step": 2097
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.958277145507141e-05,
+      "loss": 2.0459,
+      "step": 2098
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9582216960612584e-05,
+      "loss": 2.1001,
+      "step": 2099
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.958166210341992e-05,
+      "loss": 2.0725,
+      "step": 2100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9581106883507958e-05,
+      "loss": 2.072,
+      "step": 2101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.958055130089127e-05,
+      "loss": 2.0659,
+      "step": 2102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9579995355584415e-05,
+      "loss": 2.0615,
+      "step": 2103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9579439047601974e-05,
+      "loss": 2.0667,
+      "step": 2104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9578882376958542e-05,
+      "loss": 2.0649,
+      "step": 2105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.957832534366871e-05,
+      "loss": 2.0525,
+      "step": 2106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9577767947747092e-05,
+      "loss": 2.053,
+      "step": 2107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9577210189208298e-05,
+      "loss": 2.0693,
+      "step": 2108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9576652068066957e-05,
+      "loss": 2.0491,
+      "step": 2109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.957609358433771e-05,
+      "loss": 2.0107,
+      "step": 2110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9575534738035197e-05,
+      "loss": 2.0698,
+      "step": 2111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.957497552917407e-05,
+      "loss": 2.0474,
+      "step": 2112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9574415957769004e-05,
+      "loss": 2.0579,
+      "step": 2113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9573856023834665e-05,
+      "loss": 2.0891,
+      "step": 2114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9573295727385742e-05,
+      "loss": 2.0723,
+      "step": 2115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9572735068436923e-05,
+      "loss": 2.0735,
+      "step": 2116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9572174047002912e-05,
+      "loss": 2.0457,
+      "step": 2117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.957161266309842e-05,
+      "loss": 2.0522,
+      "step": 2118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9571050916738174e-05,
+      "loss": 2.0627,
+      "step": 2119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9570488807936903e-05,
+      "loss": 2.0381,
+      "step": 2120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9569926336709343e-05,
+      "loss": 2.1306,
+      "step": 2121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9569363503070253e-05,
+      "loss": 2.0869,
+      "step": 2122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9568800307034377e-05,
+      "loss": 2.0757,
+      "step": 2123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9568236748616502e-05,
+      "loss": 2.0686,
+      "step": 2124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9567672827831396e-05,
+      "loss": 2.0637,
+      "step": 2125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9567108544693856e-05,
+      "loss": 2.0288,
+      "step": 2126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9566543899218664e-05,
+      "loss": 2.0562,
+      "step": 2127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9565978891420643e-05,
+      "loss": 2.0137,
+      "step": 2128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9565413521314605e-05,
+      "loss": 2.0789,
+      "step": 2129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.956484778891537e-05,
+      "loss": 2.0906,
+      "step": 2130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9564281694237784e-05,
+      "loss": 2.0742,
+      "step": 2131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9563715237296684e-05,
+      "loss": 2.082,
+      "step": 2132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9563148418106928e-05,
+      "loss": 2.0684,
+      "step": 2133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.956258123668338e-05,
+      "loss": 2.0906,
+      "step": 2134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.956201369304091e-05,
+      "loss": 2.0371,
+      "step": 2135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9561445787194407e-05,
+      "loss": 2.0789,
+      "step": 2136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.956087751915876e-05,
+      "loss": 2.0845,
+      "step": 2137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9560308888948875e-05,
+      "loss": 2.0918,
+      "step": 2138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955973989657966e-05,
+      "loss": 2.0508,
+      "step": 2139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955917054206603e-05,
+      "loss": 2.0452,
+      "step": 2140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9558600825422933e-05,
+      "loss": 2.0334,
+      "step": 2141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9558030746665295e-05,
+      "loss": 2.0527,
+      "step": 2142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955746030580807e-05,
+      "loss": 2.0786,
+      "step": 2143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9556889502866213e-05,
+      "loss": 2.0686,
+      "step": 2144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.95563183378547e-05,
+      "loss": 2.0159,
+      "step": 2145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9555746810788507e-05,
+      "loss": 2.0144,
+      "step": 2146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9555174921682615e-05,
+      "loss": 2.0688,
+      "step": 2147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955460267055203e-05,
+      "loss": 2.0581,
+      "step": 2148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9554030057411755e-05,
+      "loss": 2.0508,
+      "step": 2149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9553457082276806e-05,
+      "loss": 2.0752,
+      "step": 2150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9552883745162205e-05,
+      "loss": 2.0542,
+      "step": 2151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9552310046082996e-05,
+      "loss": 2.0154,
+      "step": 2152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9551735985054214e-05,
+      "loss": 2.0488,
+      "step": 2153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955116156209092e-05,
+      "loss": 2.04,
+      "step": 2154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9550586777208172e-05,
+      "loss": 2.0305,
+      "step": 2155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.955001163042105e-05,
+      "loss": 2.0593,
+      "step": 2156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.954943612174463e-05,
+      "loss": 2.0527,
+      "step": 2157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9548860251194005e-05,
+      "loss": 2.075,
+      "step": 2158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9548284018784283e-05,
+      "loss": 2.0891,
+      "step": 2159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9547707424530566e-05,
+      "loss": 2.063,
+      "step": 2160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9547130468447983e-05,
+      "loss": 2.0706,
+      "step": 2161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9546553150551654e-05,
+      "loss": 2.0891,
+      "step": 2162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9545975470856728e-05,
+      "loss": 2.032,
+      "step": 2163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9545397429378347e-05,
+      "loss": 2.0364,
+      "step": 2164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9544819026131675e-05,
+      "loss": 2.0874,
+      "step": 2165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.954424026113188e-05,
+      "loss": 2.074,
+      "step": 2166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9543661134394133e-05,
+      "loss": 2.0266,
+      "step": 2167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9543081645933627e-05,
+      "loss": 2.033,
+      "step": 2168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9542501795765556e-05,
+      "loss": 2.1042,
+      "step": 2169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9541921583905127e-05,
+      "loss": 2.1011,
+      "step": 2170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9541341010367555e-05,
+      "loss": 2.0676,
+      "step": 2171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9540760075168063e-05,
+      "loss": 2.0935,
+      "step": 2172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9540178778321884e-05,
+      "loss": 2.0974,
+      "step": 2173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953959711984427e-05,
+      "loss": 2.0972,
+      "step": 2174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953901509975047e-05,
+      "loss": 2.0098,
+      "step": 2175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9538432718055738e-05,
+      "loss": 2.0107,
+      "step": 2176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9537849974775363e-05,
+      "loss": 2.053,
+      "step": 2177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9537266869924612e-05,
+      "loss": 2.0366,
+      "step": 2178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9536683403518783e-05,
+      "loss": 2.0696,
+      "step": 2179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953609957557318e-05,
+      "loss": 2.054,
+      "step": 2180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9535515386103103e-05,
+      "loss": 2.0251,
+      "step": 2181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953493083512388e-05,
+      "loss": 2.0566,
+      "step": 2182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953434592265084e-05,
+      "loss": 2.0403,
+      "step": 2183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953376064869932e-05,
+      "loss": 2.0459,
+      "step": 2184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9533175013284664e-05,
+      "loss": 2.1187,
+      "step": 2185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9532589016422237e-05,
+      "loss": 2.1121,
+      "step": 2186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.95320026581274e-05,
+      "loss": 2.0308,
+      "step": 2187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.953141593841553e-05,
+      "loss": 2.0435,
+      "step": 2188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9530828857302017e-05,
+      "loss": 2.0454,
+      "step": 2189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9530241414802256e-05,
+      "loss": 2.0608,
+      "step": 2190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9529653610931652e-05,
+      "loss": 2.0591,
+      "step": 2191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9529065445705616e-05,
+      "loss": 2.0574,
+      "step": 2192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9528476919139572e-05,
+      "loss": 2.0547,
+      "step": 2193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.952788803124896e-05,
+      "loss": 2.0562,
+      "step": 2194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9527298782049215e-05,
+      "loss": 2.021,
+      "step": 2195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.952670917155579e-05,
+      "loss": 2.0212,
+      "step": 2196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9526119199784156e-05,
+      "loss": 2.0222,
+      "step": 2197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9525528866749774e-05,
+      "loss": 2.0049,
+      "step": 2198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.952493817246813e-05,
+      "loss": 2.0789,
+      "step": 2199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9524347116954714e-05,
+      "loss": 2.072,
+      "step": 2200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9523755700225018e-05,
+      "loss": 2.0378,
+      "step": 2201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9523163922294563e-05,
+      "loss": 2.0911,
+      "step": 2202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9522571783178865e-05,
+      "loss": 2.0918,
+      "step": 2203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9521979282893445e-05,
+      "loss": 2.1064,
+      "step": 2204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.952138642145385e-05,
+      "loss": 2.0369,
+      "step": 2205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9520793198875616e-05,
+      "loss": 2.04,
+      "step": 2206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.952019961517431e-05,
+      "loss": 2.0146,
+      "step": 2207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9519605670365495e-05,
+      "loss": 2.0151,
+      "step": 2208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.951901136446474e-05,
+      "loss": 2.0886,
+      "step": 2209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9518416697487646e-05,
+      "loss": 2.0626,
+      "step": 2210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.951782166944979e-05,
+      "loss": 1.9783,
+      "step": 2211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9517226280366784e-05,
+      "loss": 2.0063,
+      "step": 2212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9516630530254243e-05,
+      "loss": 2.0669,
+      "step": 2213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9516034419127782e-05,
+      "loss": 2.0659,
+      "step": 2214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9515437947003042e-05,
+      "loss": 2.0332,
+      "step": 2215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.951484111389566e-05,
+      "loss": 2.0635,
+      "step": 2216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9514243919821287e-05,
+      "loss": 2.1279,
+      "step": 2217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.951364636479559e-05,
+      "loss": 2.1189,
+      "step": 2218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9513048448834232e-05,
+      "loss": 2.0549,
+      "step": 2219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9512450171952895e-05,
+      "loss": 2.0364,
+      "step": 2220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9511851534167268e-05,
+      "loss": 2.0339,
+      "step": 2221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9511252535493052e-05,
+      "loss": 2.0125,
+      "step": 2222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.951065317594595e-05,
+      "loss": 2.0708,
+      "step": 2223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9510053455541683e-05,
+      "loss": 2.0579,
+      "step": 2224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950945337429598e-05,
+      "loss": 2.1177,
+      "step": 2225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950885293222457e-05,
+      "loss": 2.106,
+      "step": 2226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9508252129343207e-05,
+      "loss": 2.0283,
+      "step": 2227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9507650965667642e-05,
+      "loss": 2.0117,
+      "step": 2228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9507049441213643e-05,
+      "loss": 2.1035,
+      "step": 2229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9506447555996978e-05,
+      "loss": 2.0981,
+      "step": 2230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950584531003344e-05,
+      "loss": 2.0684,
+      "step": 2231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9505242703338814e-05,
+      "loss": 2.0823,
+      "step": 2232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9504639735928908e-05,
+      "loss": 2.0842,
+      "step": 2233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950403640781953e-05,
+      "loss": 2.0701,
+      "step": 2234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.95034327190265e-05,
+      "loss": 2.1025,
+      "step": 2235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9502828669565655e-05,
+      "loss": 2.0845,
+      "step": 2236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950222425945284e-05,
+      "loss": 2.1028,
+      "step": 2237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9501619488703886e-05,
+      "loss": 2.1003,
+      "step": 2238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.950101435733467e-05,
+      "loss": 2.0525,
+      "step": 2239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9500408865361058e-05,
+      "loss": 2.0793,
+      "step": 2240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9499803012798922e-05,
+      "loss": 1.9573,
+      "step": 2241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9499196799664154e-05,
+      "loss": 1.9836,
+      "step": 2242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949859022597265e-05,
+      "loss": 2.1096,
+      "step": 2243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949798329174032e-05,
+      "loss": 2.0759,
+      "step": 2244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9497375996983073e-05,
+      "loss": 2.0435,
+      "step": 2245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949676834171684e-05,
+      "loss": 2.0811,
+      "step": 2246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949616032595756e-05,
+      "loss": 2.0505,
+      "step": 2247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949555194972117e-05,
+      "loss": 2.0178,
+      "step": 2248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949494321302362e-05,
+      "loss": 2.0562,
+      "step": 2249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949433411588089e-05,
+      "loss": 2.0906,
+      "step": 2250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9493724658308935e-05,
+      "loss": 2.0659,
+      "step": 2251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949311484032375e-05,
+      "loss": 2.0693,
+      "step": 2252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949250466194132e-05,
+      "loss": 2.0701,
+      "step": 2253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949189412317765e-05,
+      "loss": 2.0706,
+      "step": 2254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9491283224048746e-05,
+      "loss": 2.1052,
+      "step": 2255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9490671964570633e-05,
+      "loss": 2.1045,
+      "step": 2256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.949006034475934e-05,
+      "loss": 2.0491,
+      "step": 2257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.94894483646309e-05,
+      "loss": 2.0483,
+      "step": 2258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9488836024201368e-05,
+      "loss": 2.0918,
+      "step": 2259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9488223323486802e-05,
+      "loss": 2.0901,
+      "step": 2260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9487610262503262e-05,
+      "loss": 2.0715,
+      "step": 2261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9486996841266835e-05,
+      "loss": 2.0718,
+      "step": 2262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9486383059793604e-05,
+      "loss": 2.114,
+      "step": 2263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9485768918099656e-05,
+      "loss": 2.1194,
+      "step": 2264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9485154416201105e-05,
+      "loss": 2.0066,
+      "step": 2265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.948453955411407e-05,
+      "loss": 1.991,
+      "step": 2266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.948392433185466e-05,
+      "loss": 2.0564,
+      "step": 2267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9483308749439025e-05,
+      "loss": 2.0132,
+      "step": 2268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9482692806883297e-05,
+      "loss": 2.0396,
+      "step": 2269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.948207650420363e-05,
+      "loss": 2.0745,
+      "step": 2270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.948145984141619e-05,
+      "loss": 2.0876,
+      "step": 2271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9480842818537145e-05,
+      "loss": 2.0476,
+      "step": 2272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9480225435582676e-05,
+      "loss": 2.0039,
+      "step": 2273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9479607692568975e-05,
+      "loss": 2.0579,
+      "step": 2274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.947898958951224e-05,
+      "loss": 2.0459,
+      "step": 2275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9478371126428676e-05,
+      "loss": 2.0806,
+      "step": 2276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9477752303334506e-05,
+      "loss": 2.0776,
+      "step": 2277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9477133120245966e-05,
+      "loss": 2.0742,
+      "step": 2278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9476513577179276e-05,
+      "loss": 2.1106,
+      "step": 2279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9475893674150698e-05,
+      "loss": 2.052,
+      "step": 2280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.947527341117648e-05,
+      "loss": 2.1355,
+      "step": 2281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9474652788272886e-05,
+      "loss": 2.1179,
+      "step": 2282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9474031805456197e-05,
+      "loss": 2.0291,
+      "step": 2283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9473410462742697e-05,
+      "loss": 2.0449,
+      "step": 2284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9472788760148675e-05,
+      "loss": 2.1079,
+      "step": 2285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9472166697690443e-05,
+      "loss": 2.1104,
+      "step": 2286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9471544275384304e-05,
+      "loss": 2.0872,
+      "step": 2287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9470921493246588e-05,
+      "loss": 2.0698,
+      "step": 2288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9470298351293622e-05,
+      "loss": 2.0969,
+      "step": 2289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.946967484954175e-05,
+      "loss": 2.0974,
+      "step": 2290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.946905098800732e-05,
+      "loss": 2.0688,
+      "step": 2291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9468426766706694e-05,
+      "loss": 2.0549,
+      "step": 2292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.946780218565624e-05,
+      "loss": 2.0618,
+      "step": 2293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9467177244872338e-05,
+      "loss": 2.0422,
+      "step": 2294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.946655194437138e-05,
+      "loss": 2.0659,
+      "step": 2295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9465926284169755e-05,
+      "loss": 2.0894,
+      "step": 2296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.946530026428388e-05,
+      "loss": 2.0894,
+      "step": 2297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9464673884730164e-05,
+      "loss": 2.0964,
+      "step": 2298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9464047145525037e-05,
+      "loss": 2.0742,
+      "step": 2299
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9463420046684934e-05,
+      "loss": 2.0684,
+      "step": 2300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9462792588226296e-05,
+      "loss": 2.0867,
+      "step": 2301
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.946216477016558e-05,
+      "loss": 2.0632,
+      "step": 2302
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.946153659251926e-05,
+      "loss": 2.0327,
+      "step": 2303
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.946090805530379e-05,
+      "loss": 2.0142,
+      "step": 2304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9460279158535666e-05,
+      "loss": 2.0283,
+      "step": 2305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945964990223138e-05,
+      "loss": 2.03,
+      "step": 2306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945902028640742e-05,
+      "loss": 2.082,
+      "step": 2307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9458390311080316e-05,
+      "loss": 2.0491,
+      "step": 2308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945775997626658e-05,
+      "loss": 2.082,
+      "step": 2309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9457129281982732e-05,
+      "loss": 2.0737,
+      "step": 2310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9456498228245326e-05,
+      "loss": 2.0417,
+      "step": 2311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945586681507091e-05,
+      "loss": 2.0378,
+      "step": 2312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945523504247603e-05,
+      "loss": 2.1206,
+      "step": 2313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9454602910477263e-05,
+      "loss": 2.0659,
+      "step": 2314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9453970419091185e-05,
+      "loss": 2.0012,
+      "step": 2315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945333756833438e-05,
+      "loss": 2.0259,
+      "step": 2316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9452704358223445e-05,
+      "loss": 2.0291,
+      "step": 2317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9452070788774983e-05,
+      "loss": 2.0659,
+      "step": 2318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.945143686000561e-05,
+      "loss": 2.0369,
+      "step": 2319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9450802571931955e-05,
+      "loss": 2.0488,
+      "step": 2320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9450167924570643e-05,
+      "loss": 2.0493,
+      "step": 2321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9449532917938323e-05,
+      "loss": 2.0459,
+      "step": 2322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9448897552051647e-05,
+      "loss": 2.0708,
+      "step": 2323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9448261826927276e-05,
+      "loss": 2.0564,
+      "step": 2324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9447625742581873e-05,
+      "loss": 2.1086,
+      "step": 2325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.944698929903213e-05,
+      "loss": 2.0977,
+      "step": 2326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9446352496294732e-05,
+      "loss": 2.0332,
+      "step": 2327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.944571533438638e-05,
+      "loss": 2.0706,
+      "step": 2328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.944507781332378e-05,
+      "loss": 2.0452,
+      "step": 2329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9444439933123654e-05,
+      "loss": 2.0803,
+      "step": 2330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9443801693802726e-05,
+      "loss": 2.0322,
+      "step": 2331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9443163095377738e-05,
+      "loss": 2.0552,
+      "step": 2332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.944252413786543e-05,
+      "loss": 2.0789,
+      "step": 2333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9441884821282566e-05,
+      "loss": 2.1094,
+      "step": 2334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.94412451456459e-05,
+      "loss": 2.0271,
+      "step": 2335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.944060511097222e-05,
+      "loss": 2.0298,
+      "step": 2336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9439964717278302e-05,
+      "loss": 2.0625,
+      "step": 2337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9439323964580938e-05,
+      "loss": 2.0874,
+      "step": 2338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9438682852896937e-05,
+      "loss": 2.0356,
+      "step": 2339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.943804138224311e-05,
+      "loss": 2.0295,
+      "step": 2340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9437399552636273e-05,
+      "loss": 2.0715,
+      "step": 2341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.943675736409326e-05,
+      "loss": 2.082,
+      "step": 2342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9436114816630918e-05,
+      "loss": 2.0466,
+      "step": 2343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9435471910266095e-05,
+      "loss": 2.0283,
+      "step": 2344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9434828645015645e-05,
+      "loss": 2.0696,
+      "step": 2345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.943418502089644e-05,
+      "loss": 2.0649,
+      "step": 2346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.943354103792536e-05,
+      "loss": 2.0332,
+      "step": 2347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9432896696119286e-05,
+      "loss": 2.0337,
+      "step": 2348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.943225199549512e-05,
+      "loss": 2.0535,
+      "step": 2349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9431606936069772e-05,
+      "loss": 2.0835,
+      "step": 2350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9430961517860152e-05,
+      "loss": 2.0767,
+      "step": 2351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9430315740883188e-05,
+      "loss": 2.0796,
+      "step": 2352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9429669605155814e-05,
+      "loss": 2.0217,
+      "step": 2353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9429023110694974e-05,
+      "loss": 2.0225,
+      "step": 2354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9428376257517625e-05,
+      "loss": 2.0327,
+      "step": 2355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.942772904564073e-05,
+      "loss": 2.0415,
+      "step": 2356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9427081475081254e-05,
+      "loss": 2.0867,
+      "step": 2357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.942643354585618e-05,
+      "loss": 2.0686,
+      "step": 2358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.942578525798251e-05,
+      "loss": 2.0896,
+      "step": 2359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9425136611477227e-05,
+      "loss": 2.0378,
+      "step": 2360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9424487606357362e-05,
+      "loss": 2.0737,
+      "step": 2361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9423838242639918e-05,
+      "loss": 2.061,
+      "step": 2362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9423188520341928e-05,
+      "loss": 2.0093,
+      "step": 2363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9422538439480435e-05,
+      "loss": 2.0154,
+      "step": 2364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.942188800007248e-05,
+      "loss": 2.0781,
+      "step": 2365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9421237202135126e-05,
+      "loss": 2.0959,
+      "step": 2366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9420586045685436e-05,
+      "loss": 2.0967,
+      "step": 2367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9419934530740483e-05,
+      "loss": 2.0605,
+      "step": 2368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941928265731736e-05,
+      "loss": 2.0947,
+      "step": 2369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9418630425433155e-05,
+      "loss": 2.1052,
+      "step": 2370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941797783510497e-05,
+      "loss": 2.0781,
+      "step": 2371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9417324886349936e-05,
+      "loss": 2.1111,
+      "step": 2372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9416671579185147e-05,
+      "loss": 2.0498,
+      "step": 2373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941601791362776e-05,
+      "loss": 2.0632,
+      "step": 2374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941536388969491e-05,
+      "loss": 2.0532,
+      "step": 2375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9414709507403736e-05,
+      "loss": 2.0562,
+      "step": 2376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9414054766771413e-05,
+      "loss": 2.0205,
+      "step": 2377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9413399667815103e-05,
+      "loss": 2.0399,
+      "step": 2378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9412744210551988e-05,
+      "loss": 2.1455,
+      "step": 2379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941208839499926e-05,
+      "loss": 2.125,
+      "step": 2380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.941143222117411e-05,
+      "loss": 2.0491,
+      "step": 2381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9410775689093752e-05,
+      "loss": 2.0547,
+      "step": 2382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9410118798775397e-05,
+      "loss": 2.1023,
+      "step": 2383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9409461550236275e-05,
+      "loss": 2.0801,
+      "step": 2384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9408803943493622e-05,
+      "loss": 2.0745,
+      "step": 2385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940814597856468e-05,
+      "loss": 2.1006,
+      "step": 2386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9407487655466702e-05,
+      "loss": 2.0908,
+      "step": 2387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9406828974216958e-05,
+      "loss": 2.0996,
+      "step": 2388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9406169934832715e-05,
+      "loss": 2.0718,
+      "step": 2389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940551053733126e-05,
+      "loss": 2.0623,
+      "step": 2390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940485078172988e-05,
+      "loss": 2.0305,
+      "step": 2391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9404190668045885e-05,
+      "loss": 2.0576,
+      "step": 2392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9403530196296575e-05,
+      "loss": 2.1086,
+      "step": 2393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9402869366499275e-05,
+      "loss": 2.1123,
+      "step": 2394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940220817867132e-05,
+      "loss": 2.0615,
+      "step": 2395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940154663283004e-05,
+      "loss": 2.0745,
+      "step": 2396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940088472899279e-05,
+      "loss": 2.0945,
+      "step": 2397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.940022246717692e-05,
+      "loss": 2.0889,
+      "step": 2398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9399559847399807e-05,
+      "loss": 2.0513,
+      "step": 2399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9398896869678817e-05,
+      "loss": 2.0759,
+      "step": 2400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9398233534031342e-05,
+      "loss": 2.0833,
+      "step": 2401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9397569840474777e-05,
+      "loss": 2.0947,
+      "step": 2402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9396905789026525e-05,
+      "loss": 2.0237,
+      "step": 2403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9396241379704003e-05,
+      "loss": 2.0605,
+      "step": 2404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.939557661252463e-05,
+      "loss": 2.1111,
+      "step": 2405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.939491148750584e-05,
+      "loss": 2.0771,
+      "step": 2406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9394246004665078e-05,
+      "loss": 2.0139,
+      "step": 2407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9393580164019793e-05,
+      "loss": 2.0586,
+      "step": 2408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9392913965587443e-05,
+      "loss": 2.0537,
+      "step": 2409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9392247409385502e-05,
+      "loss": 2.0591,
+      "step": 2410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.939158049543145e-05,
+      "loss": 2.0491,
+      "step": 2411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9390913223742783e-05,
+      "loss": 2.0654,
+      "step": 2412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9390245594336983e-05,
+      "loss": 2.0552,
+      "step": 2413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9389577607231568e-05,
+      "loss": 2.0674,
+      "step": 2414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9388909262444055e-05,
+      "loss": 2.0361,
+      "step": 2415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9388240559991967e-05,
+      "loss": 2.0522,
+      "step": 2416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9387571499892845e-05,
+      "loss": 2.0793,
+      "step": 2417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9386902082164232e-05,
+      "loss": 2.0632,
+      "step": 2418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.938623230682368e-05,
+      "loss": 2.0132,
+      "step": 2419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9385562173888757e-05,
+      "loss": 2.0317,
+      "step": 2420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9384891683377033e-05,
+      "loss": 2.1006,
+      "step": 2421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9384220835306095e-05,
+      "loss": 2.0696,
+      "step": 2422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9383549629693534e-05,
+      "loss": 2.0981,
+      "step": 2423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9382878066556945e-05,
+      "loss": 2.0857,
+      "step": 2424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.938220614591395e-05,
+      "loss": 2.0959,
+      "step": 2425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9381533867782158e-05,
+      "loss": 2.0964,
+      "step": 2426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9380861232179207e-05,
+      "loss": 2.0339,
+      "step": 2427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9380188239122735e-05,
+      "loss": 2.0205,
+      "step": 2428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9379514888630387e-05,
+      "loss": 2.095,
+      "step": 2429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9378841180719825e-05,
+      "loss": 2.0872,
+      "step": 2430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9378167115408712e-05,
+      "loss": 2.0239,
+      "step": 2431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9377492692714728e-05,
+      "loss": 2.0442,
+      "step": 2432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.937681791265556e-05,
+      "loss": 2.1204,
+      "step": 2433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9376142775248894e-05,
+      "loss": 2.084,
+      "step": 2434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9375467280512447e-05,
+      "loss": 2.0354,
+      "step": 2435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9374791428463926e-05,
+      "loss": 2.0112,
+      "step": 2436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.937411521912106e-05,
+      "loss": 2.0806,
+      "step": 2437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.937343865250157e-05,
+      "loss": 2.1038,
+      "step": 2438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9372761728623214e-05,
+      "loss": 2.0857,
+      "step": 2439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.937208444750373e-05,
+      "loss": 2.05,
+      "step": 2440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.937140680916089e-05,
+      "loss": 2.0715,
+      "step": 2441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9370728813612452e-05,
+      "loss": 2.0869,
+      "step": 2442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9370050460876205e-05,
+      "loss": 2.0867,
+      "step": 2443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.936937175096994e-05,
+      "loss": 2.0354,
+      "step": 2444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9368692683911444e-05,
+      "loss": 2.0398,
+      "step": 2445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9368013259718533e-05,
+      "loss": 2.0552,
+      "step": 2446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9367333478409023e-05,
+      "loss": 2.0317,
+      "step": 2447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9366653340000746e-05,
+      "loss": 2.0488,
+      "step": 2448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9365972844511526e-05,
+      "loss": 2.084,
+      "step": 2449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9365291991959214e-05,
+      "loss": 2.0552,
+      "step": 2450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9364610782361667e-05,
+      "loss": 2.0427,
+      "step": 2451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.936392921573675e-05,
+      "loss": 2.0371,
+      "step": 2452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9363247292102326e-05,
+      "loss": 2.0461,
+      "step": 2453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.936256501147629e-05,
+      "loss": 2.0442,
+      "step": 2454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9361882373876526e-05,
+      "loss": 2.0549,
+      "step": 2455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9361199379320937e-05,
+      "loss": 2.0486,
+      "step": 2456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.936051602782744e-05,
+      "loss": 2.0647,
+      "step": 2457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9359832319413946e-05,
+      "loss": 2.0605,
+      "step": 2458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9359148254098388e-05,
+      "loss": 2.1013,
+      "step": 2459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9358463831898707e-05,
+      "loss": 2.085,
+      "step": 2460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9357779052832848e-05,
+      "loss": 2.0793,
+      "step": 2461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9357093916918774e-05,
+      "loss": 2.0608,
+      "step": 2462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.935640842417444e-05,
+      "loss": 2.0225,
+      "step": 2463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9355722574617835e-05,
+      "loss": 2.0295,
+      "step": 2464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9355036368266938e-05,
+      "loss": 2.0627,
+      "step": 2465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9354349805139746e-05,
+      "loss": 2.0366,
+      "step": 2466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9353662885254268e-05,
+      "loss": 2.0945,
+      "step": 2467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.935297560862851e-05,
+      "loss": 2.0767,
+      "step": 2468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9352287975280493e-05,
+      "loss": 2.1448,
+      "step": 2469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9351599985228256e-05,
+      "loss": 2.1042,
+      "step": 2470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.935091163848984e-05,
+      "loss": 2.0676,
+      "step": 2471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9350222935083295e-05,
+      "loss": 2.0798,
+      "step": 2472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.934953387502668e-05,
+      "loss": 2.106,
+      "step": 2473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.934884445833807e-05,
+      "loss": 2.0906,
+      "step": 2474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9348154685035537e-05,
+      "loss": 2.0576,
+      "step": 2475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9347464555137172e-05,
+      "loss": 2.0828,
+      "step": 2476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9346774068661074e-05,
+      "loss": 2.1128,
+      "step": 2477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9346083225625352e-05,
+      "loss": 2.0874,
+      "step": 2478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9345392026048117e-05,
+      "loss": 2.0986,
+      "step": 2479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9344700469947503e-05,
+      "loss": 2.0776,
+      "step": 2480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9344008557341636e-05,
+      "loss": 2.0896,
+      "step": 2481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9343316288248667e-05,
+      "loss": 2.0854,
+      "step": 2482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.934262366268675e-05,
+      "loss": 2.0603,
+      "step": 2483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9341930680674047e-05,
+      "loss": 2.0869,
+      "step": 2484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9341237342228725e-05,
+      "loss": 2.1021,
+      "step": 2485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9340543647368974e-05,
+      "loss": 2.0662,
+      "step": 2486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9339849596112983e-05,
+      "loss": 2.0891,
+      "step": 2487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933915518847895e-05,
+      "loss": 2.0859,
+      "step": 2488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933846042448509e-05,
+      "loss": 2.075,
+      "step": 2489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933776530414962e-05,
+      "loss": 2.0547,
+      "step": 2490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9337069827490765e-05,
+      "loss": 2.0564,
+      "step": 2491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933637399452677e-05,
+      "loss": 2.0391,
+      "step": 2492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9335677805275874e-05,
+      "loss": 2.0842,
+      "step": 2493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933498125975634e-05,
+      "loss": 2.0916,
+      "step": 2494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933428435798643e-05,
+      "loss": 2.0793,
+      "step": 2495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.933358709998443e-05,
+      "loss": 2.0708,
+      "step": 2496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9332889485768616e-05,
+      "loss": 2.0151,
+      "step": 2497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9332191515357273e-05,
+      "loss": 2.0186,
+      "step": 2498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9331493188768725e-05,
+      "loss": 2.0723,
+      "step": 2499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9330794506021268e-05,
+      "loss": 2.022,
+      "step": 2500
+    },
+    {
+      "Bleu_1": 0.7704474447405247,
+      "Bleu_2": 0.6086311512781954,
+      "Bleu_3": 0.46719339328913007,
+      "Bleu_4": 0.3548403280910617,
+      "CIDEr": 1.1957586325543035,
+      "METEOR": 0.2783552537256633,
+      "ROUGE_L": 0.5572698758838127,
+      "epoch": 0.12,
+      "step": 2500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9330095467133234e-05,
+      "loss": 2.0759,
+      "step": 2501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9329396072122948e-05,
+      "loss": 2.1099,
+      "step": 2502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9328696321008755e-05,
+      "loss": 2.0574,
+      "step": 2503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9327996213809002e-05,
+      "loss": 2.0562,
+      "step": 2504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932729575054205e-05,
+      "loss": 2.0234,
+      "step": 2505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9326594931226276e-05,
+      "loss": 1.967,
+      "step": 2506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932589375588004e-05,
+      "loss": 2.0925,
+      "step": 2507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9325192224521746e-05,
+      "loss": 2.0544,
+      "step": 2508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9324490337169782e-05,
+      "loss": 2.0801,
+      "step": 2509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932378809384256e-05,
+      "loss": 2.0706,
+      "step": 2510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932308549455849e-05,
+      "loss": 2.0737,
+      "step": 2511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9322382539335998e-05,
+      "loss": 2.0308,
+      "step": 2512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932167922819352e-05,
+      "loss": 2.0803,
+      "step": 2513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.93209755611495e-05,
+      "loss": 2.0955,
+      "step": 2514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.932027153822239e-05,
+      "loss": 2.0657,
+      "step": 2515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931956715943065e-05,
+      "loss": 2.0823,
+      "step": 2516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9318862424792757e-05,
+      "loss": 2.1147,
+      "step": 2517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931815733432719e-05,
+      "loss": 2.084,
+      "step": 2518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931745188805243e-05,
+      "loss": 2.0344,
+      "step": 2519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9316746085986986e-05,
+      "loss": 2.0186,
+      "step": 2520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931603992814937e-05,
+      "loss": 2.0981,
+      "step": 2521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931533341455809e-05,
+      "loss": 2.0994,
+      "step": 2522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931462654523168e-05,
+      "loss": 2.0437,
+      "step": 2523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931391932018868e-05,
+      "loss": 2.0581,
+      "step": 2524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9313211739447626e-05,
+      "loss": 2.1331,
+      "step": 2525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931250380302708e-05,
+      "loss": 2.0764,
+      "step": 2526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9311795510945608e-05,
+      "loss": 2.1174,
+      "step": 2527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9311086863221776e-05,
+      "loss": 2.1038,
+      "step": 2528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.931037785987418e-05,
+      "loss": 2.0591,
+      "step": 2529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.93096685009214e-05,
+      "loss": 2.0486,
+      "step": 2530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9308958786382053e-05,
+      "loss": 2.0964,
+      "step": 2531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9308248716274733e-05,
+      "loss": 2.0906,
+      "step": 2532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.930753829061807e-05,
+      "loss": 2.0686,
+      "step": 2533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9306827509430703e-05,
+      "loss": 2.0903,
+      "step": 2534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9306116372731252e-05,
+      "loss": 2.0791,
+      "step": 2535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.930540488053838e-05,
+      "loss": 2.0535,
+      "step": 2536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.930469303287074e-05,
+      "loss": 2.0503,
+      "step": 2537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9303980829746998e-05,
+      "loss": 2.073,
+      "step": 2538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9303268271185838e-05,
+      "loss": 2.1023,
+      "step": 2539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.930255535720594e-05,
+      "loss": 2.1074,
+      "step": 2540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9301842087826e-05,
+      "loss": 2.0837,
+      "step": 2541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.930112846306472e-05,
+      "loss": 2.0859,
+      "step": 2542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9300414482940822e-05,
+      "loss": 2.0759,
+      "step": 2543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9299700147473017e-05,
+      "loss": 2.0425,
+      "step": 2544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.929898545668005e-05,
+      "loss": 2.083,
+      "step": 2545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9298270410580656e-05,
+      "loss": 2.0667,
+      "step": 2546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.929755500919359e-05,
+      "loss": 2.0818,
+      "step": 2547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9296839252537605e-05,
+      "loss": 2.0725,
+      "step": 2548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9296123140631483e-05,
+      "loss": 2.0579,
+      "step": 2549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9295406673493993e-05,
+      "loss": 2.0774,
+      "step": 2550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9294689851143926e-05,
+      "loss": 1.9812,
+      "step": 2551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9293972673600086e-05,
+      "loss": 2.0059,
+      "step": 2552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9293255140881274e-05,
+      "loss": 2.0757,
+      "step": 2553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9292537253006304e-05,
+      "loss": 2.0557,
+      "step": 2554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.929181900999401e-05,
+      "loss": 2.0505,
+      "step": 2555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9291100411863216e-05,
+      "loss": 2.0325,
+      "step": 2556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.929038145863278e-05,
+      "loss": 2.0303,
+      "step": 2557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9289662150321546e-05,
+      "loss": 1.9839,
+      "step": 2558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9288942486948377e-05,
+      "loss": 2.0481,
+      "step": 2559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9288222468532154e-05,
+      "loss": 2.071,
+      "step": 2560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9287502095091746e-05,
+      "loss": 2.1023,
+      "step": 2561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9286781366646055e-05,
+      "loss": 2.0872,
+      "step": 2562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9286060283213975e-05,
+      "loss": 2.0969,
+      "step": 2563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9285338844814415e-05,
+      "loss": 2.115,
+      "step": 2564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9284617051466302e-05,
+      "loss": 2.1025,
+      "step": 2565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9283894903188555e-05,
+      "loss": 2.092,
+      "step": 2566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9283172400000116e-05,
+      "loss": 2.1057,
+      "step": 2567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9282449541919933e-05,
+      "loss": 2.0884,
+      "step": 2568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.928172632896696e-05,
+      "loss": 2.1003,
+      "step": 2569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9281002761160156e-05,
+      "loss": 2.0676,
+      "step": 2570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.928027883851851e-05,
+      "loss": 2.0369,
+      "step": 2571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9279554561060988e-05,
+      "loss": 2.0398,
+      "step": 2572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9278829928806603e-05,
+      "loss": 2.0693,
+      "step": 2573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9278104941774342e-05,
+      "loss": 2.0608,
+      "step": 2574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9277379599983226e-05,
+      "loss": 2.0601,
+      "step": 2575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9276653903452272e-05,
+      "loss": 2.0435,
+      "step": 2576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9275927852200515e-05,
+      "loss": 2.0405,
+      "step": 2577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9275201446246992e-05,
+      "loss": 2.0234,
+      "step": 2578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9274474685610746e-05,
+      "loss": 2.0747,
+      "step": 2579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9273747570310847e-05,
+      "loss": 2.0942,
+      "step": 2580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.927302010036635e-05,
+      "loss": 2.0962,
+      "step": 2581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9272292275796346e-05,
+      "loss": 2.0818,
+      "step": 2582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9271564096619918e-05,
+      "loss": 2.0459,
+      "step": 2583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9270835562856148e-05,
+      "loss": 2.0659,
+      "step": 2584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9270106674524156e-05,
+      "loss": 2.0703,
+      "step": 2585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9269377431643055e-05,
+      "loss": 2.0806,
+      "step": 2586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9268647834231953e-05,
+      "loss": 2.0322,
+      "step": 2587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9267917882310007e-05,
+      "loss": 2.0239,
+      "step": 2588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9267187575896343e-05,
+      "loss": 2.1208,
+      "step": 2589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9266456915010114e-05,
+      "loss": 2.1235,
+      "step": 2590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9265725899670487e-05,
+      "loss": 2.0999,
+      "step": 2591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9264994529896624e-05,
+      "loss": 2.0671,
+      "step": 2592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.926426280570771e-05,
+      "loss": 2.0725,
+      "step": 2593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9263530727122932e-05,
+      "loss": 2.0708,
+      "step": 2594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.926279829416149e-05,
+      "loss": 2.1079,
+      "step": 2595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9262065506842584e-05,
+      "loss": 2.0916,
+      "step": 2596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9261332365185443e-05,
+      "loss": 2.1208,
+      "step": 2597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.926059886920928e-05,
+      "loss": 2.0959,
+      "step": 2598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9259865018933336e-05,
+      "loss": 2.0496,
+      "step": 2599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9259130814376854e-05,
+      "loss": 2.0591,
+      "step": 2600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9258396255559086e-05,
+      "loss": 2.0684,
+      "step": 2601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9257661342499306e-05,
+      "loss": 2.0383,
+      "step": 2602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9256926075216775e-05,
+      "loss": 2.1079,
+      "step": 2603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9256190453730775e-05,
+      "loss": 2.0544,
+      "step": 2604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.92554544780606e-05,
+      "loss": 2.0776,
+      "step": 2605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.925471814822555e-05,
+      "loss": 2.0576,
+      "step": 2606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.925398146424493e-05,
+      "loss": 2.051,
+      "step": 2607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9253244426138064e-05,
+      "loss": 2.0696,
+      "step": 2608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.925250703392428e-05,
+      "loss": 2.0898,
+      "step": 2609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.925176928762291e-05,
+      "loss": 2.0986,
+      "step": 2610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.925103118725331e-05,
+      "loss": 2.0789,
+      "step": 2611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9250292732834825e-05,
+      "loss": 2.0461,
+      "step": 2612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9249553924386825e-05,
+      "loss": 2.0146,
+      "step": 2613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924881476192869e-05,
+      "loss": 2.0056,
+      "step": 2614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924807524547979e-05,
+      "loss": 2.0781,
+      "step": 2615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9247335375059527e-05,
+      "loss": 2.0613,
+      "step": 2616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9246595150687305e-05,
+      "loss": 2.1172,
+      "step": 2617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9245854572382536e-05,
+      "loss": 2.1106,
+      "step": 2618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9245113640164636e-05,
+      "loss": 2.0837,
+      "step": 2619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924437235405303e-05,
+      "loss": 2.0583,
+      "step": 2620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924363071406717e-05,
+      "loss": 2.073,
+      "step": 2621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.92428887202265e-05,
+      "loss": 2.0352,
+      "step": 2622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9242146372550472e-05,
+      "loss": 2.0742,
+      "step": 2623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924140367105856e-05,
+      "loss": 2.0784,
+      "step": 2624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.924066061577024e-05,
+      "loss": 2.061,
+      "step": 2625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.923991720670499e-05,
+      "loss": 2.0396,
+      "step": 2626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9239173443882316e-05,
+      "loss": 2.0486,
+      "step": 2627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9238429327321717e-05,
+      "loss": 2.0474,
+      "step": 2628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9237684857042707e-05,
+      "loss": 2.0503,
+      "step": 2629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9236940033064806e-05,
+      "loss": 2.103,
+      "step": 2630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9236194855407555e-05,
+      "loss": 2.0476,
+      "step": 2631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9235449324090488e-05,
+      "loss": 2.0183,
+      "step": 2632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9234703439133154e-05,
+      "loss": 2.0161,
+      "step": 2633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9233957200555115e-05,
+      "loss": 2.0071,
+      "step": 2634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9233210608375943e-05,
+      "loss": 2.0481,
+      "step": 2635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9232463662615217e-05,
+      "loss": 2.042,
+      "step": 2636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.923171636329252e-05,
+      "loss": 2.1047,
+      "step": 2637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9230968710427452e-05,
+      "loss": 2.1111,
+      "step": 2638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.923022070403962e-05,
+      "loss": 2.0342,
+      "step": 2639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.922947234414864e-05,
+      "loss": 2.061,
+      "step": 2640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9228723630774136e-05,
+      "loss": 2.0244,
+      "step": 2641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9227974563935736e-05,
+      "loss": 2.0542,
+      "step": 2642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9227225143653093e-05,
+      "loss": 2.0396,
+      "step": 2643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9226475369945855e-05,
+      "loss": 2.0579,
+      "step": 2644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.922572524283369e-05,
+      "loss": 2.1038,
+      "step": 2645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9224974762336254e-05,
+      "loss": 2.085,
+      "step": 2646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9224223928473244e-05,
+      "loss": 2.0225,
+      "step": 2647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9223472741264342e-05,
+      "loss": 2.0205,
+      "step": 2648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9222721200729246e-05,
+      "loss": 2.033,
+      "step": 2649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9221969306887668e-05,
+      "loss": 2.0232,
+      "step": 2650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9221217059759327e-05,
+      "loss": 2.054,
+      "step": 2651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9220464459363942e-05,
+      "loss": 2.0554,
+      "step": 2652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9219711505721252e-05,
+      "loss": 2.1094,
+      "step": 2653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.921895819885101e-05,
+      "loss": 2.0977,
+      "step": 2654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9218204538772967e-05,
+      "loss": 2.0159,
+      "step": 2655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9217450525506877e-05,
+      "loss": 2.033,
+      "step": 2656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9216696159072523e-05,
+      "loss": 2.0989,
+      "step": 2657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9215941439489687e-05,
+      "loss": 2.0635,
+      "step": 2658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9215186366778157e-05,
+      "loss": 2.0535,
+      "step": 2659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.921443094095774e-05,
+      "loss": 2.0339,
+      "step": 2660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9213675162048234e-05,
+      "loss": 2.0781,
+      "step": 2661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9212919030069475e-05,
+      "loss": 2.0696,
+      "step": 2662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9212162545041274e-05,
+      "loss": 2.064,
+      "step": 2663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9211405706983483e-05,
+      "loss": 2.0623,
+      "step": 2664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9210648515915943e-05,
+      "loss": 2.0913,
+      "step": 2665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920989097185851e-05,
+      "loss": 2.0779,
+      "step": 2666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9209133074831056e-05,
+      "loss": 2.0708,
+      "step": 2667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920837482485345e-05,
+      "loss": 2.0696,
+      "step": 2668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920761622194557e-05,
+      "loss": 2.0566,
+      "step": 2669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9206857266127326e-05,
+      "loss": 2.0637,
+      "step": 2670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920609795741861e-05,
+      "loss": 2.0786,
+      "step": 2671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9205338295839332e-05,
+      "loss": 2.072,
+      "step": 2672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920457828140942e-05,
+      "loss": 2.0461,
+      "step": 2673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9203817914148802e-05,
+      "loss": 2.075,
+      "step": 2674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9203057194077417e-05,
+      "loss": 2.0725,
+      "step": 2675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920229612121521e-05,
+      "loss": 2.061,
+      "step": 2676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9201534695582146e-05,
+      "loss": 2.104,
+      "step": 2677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9200772917198192e-05,
+      "loss": 2.0913,
+      "step": 2678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.920001078608332e-05,
+      "loss": 2.0664,
+      "step": 2679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.919924830225752e-05,
+      "loss": 2.0742,
+      "step": 2680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9198485465740786e-05,
+      "loss": 2.0161,
+      "step": 2681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9197722276553123e-05,
+      "loss": 2.0459,
+      "step": 2682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.919695873471454e-05,
+      "loss": 2.0168,
+      "step": 2683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9196194840245066e-05,
+      "loss": 2.0474,
+      "step": 2684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9195430593164736e-05,
+      "loss": 2.0854,
+      "step": 2685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9194665993493585e-05,
+      "loss": 2.082,
+      "step": 2686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9193901041251664e-05,
+      "loss": 2.0791,
+      "step": 2687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9193135736459036e-05,
+      "loss": 2.0859,
+      "step": 2688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9192370079135767e-05,
+      "loss": 2.0496,
+      "step": 2689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9191604069301936e-05,
+      "loss": 2.0369,
+      "step": 2690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9190837706977635e-05,
+      "loss": 1.9761,
+      "step": 2691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9190070992182954e-05,
+      "loss": 2.0034,
+      "step": 2692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.918930392493801e-05,
+      "loss": 2.0281,
+      "step": 2693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.918853650526291e-05,
+      "loss": 2.0364,
+      "step": 2694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9187768733177778e-05,
+      "loss": 2.0681,
+      "step": 2695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9187000608702752e-05,
+      "loss": 2.0549,
+      "step": 2696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9186232131857974e-05,
+      "loss": 2.051,
+      "step": 2697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.918546330266359e-05,
+      "loss": 2.0181,
+      "step": 2698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9184694121139774e-05,
+      "loss": 2.0237,
+      "step": 2699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.918392458730669e-05,
+      "loss": 2.0217,
+      "step": 2700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9183154701184512e-05,
+      "loss": 2.0625,
+      "step": 2701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9182384462793445e-05,
+      "loss": 2.0381,
+      "step": 2702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9181613872153675e-05,
+      "loss": 2.0264,
+      "step": 2703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9180842929285408e-05,
+      "loss": 2.012,
+      "step": 2704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9180071634208875e-05,
+      "loss": 2.0051,
+      "step": 2705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.917929998694429e-05,
+      "loss": 2.05,
+      "step": 2706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9178527987511894e-05,
+      "loss": 2.0601,
+      "step": 2707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9177755635931927e-05,
+      "loss": 2.0305,
+      "step": 2708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9176982932224647e-05,
+      "loss": 2.0876,
+      "step": 2709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9176209876410315e-05,
+      "loss": 2.0669,
+      "step": 2710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.917543646850921e-05,
+      "loss": 2.0803,
+      "step": 2711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.917466270854161e-05,
+      "loss": 2.0708,
+      "step": 2712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9173888596527797e-05,
+      "loss": 2.0413,
+      "step": 2713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.917311413248808e-05,
+      "loss": 2.0107,
+      "step": 2714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9172339316442773e-05,
+      "loss": 2.0544,
+      "step": 2715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9171564148412185e-05,
+      "loss": 2.0505,
+      "step": 2716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.917078862841665e-05,
+      "loss": 2.0764,
+      "step": 2717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.91700127564765e-05,
+      "loss": 2.092,
+      "step": 2718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.916923653261208e-05,
+      "loss": 2.0566,
+      "step": 2719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9168459956843757e-05,
+      "loss": 2.0642,
+      "step": 2720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9167683029191888e-05,
+      "loss": 2.0884,
+      "step": 2721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9166905749676848e-05,
+      "loss": 2.072,
+      "step": 2722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9166128118319015e-05,
+      "loss": 2.0491,
+      "step": 2723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9165350135138788e-05,
+      "loss": 2.0732,
+      "step": 2724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.916457180015657e-05,
+      "loss": 2.0259,
+      "step": 2725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9163793113392766e-05,
+      "loss": 1.9963,
+      "step": 2726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9163014074867806e-05,
+      "loss": 1.9971,
+      "step": 2727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9162234684602104e-05,
+      "loss": 1.991,
+      "step": 2728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9161454942616112e-05,
+      "loss": 2.0791,
+      "step": 2729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9160674848930266e-05,
+      "loss": 2.0679,
+      "step": 2730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9159894403565038e-05,
+      "loss": 2.1016,
+      "step": 2731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9159113606540882e-05,
+      "loss": 2.0923,
+      "step": 2732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9158332457878282e-05,
+      "loss": 2.0576,
+      "step": 2733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9157550957597713e-05,
+      "loss": 2.0659,
+      "step": 2734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.915676910571968e-05,
+      "loss": 2.0967,
+      "step": 2735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9155986902264674e-05,
+      "loss": 2.0686,
+      "step": 2736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9155204347253218e-05,
+      "loss": 2.1155,
+      "step": 2737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.915442144070583e-05,
+      "loss": 2.1016,
+      "step": 2738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.915363818264304e-05,
+      "loss": 2.0618,
+      "step": 2739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9152854573085386e-05,
+      "loss": 2.0366,
+      "step": 2740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9152070612053424e-05,
+      "loss": 2.0391,
+      "step": 2741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9151286299567705e-05,
+      "loss": 2.0283,
+      "step": 2742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9150501635648802e-05,
+      "loss": 1.9954,
+      "step": 2743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9149716620317288e-05,
+      "loss": 2.0051,
+      "step": 2744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.914893125359375e-05,
+      "loss": 2.0056,
+      "step": 2745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9148145535498786e-05,
+      "loss": 2.0359,
+      "step": 2746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9147359466052994e-05,
+      "loss": 2.1252,
+      "step": 2747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9146573045276997e-05,
+      "loss": 2.125,
+      "step": 2748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9145786273191415e-05,
+      "loss": 2.0979,
+      "step": 2749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.914499914981688e-05,
+      "loss": 2.0908,
+      "step": 2750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9144211675174023e-05,
+      "loss": 2.1108,
+      "step": 2751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9143423849283513e-05,
+      "loss": 2.0791,
+      "step": 2752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9142635672165994e-05,
+      "loss": 1.9929,
+      "step": 2753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9141847143842144e-05,
+      "loss": 1.9866,
+      "step": 2754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.914105826433264e-05,
+      "loss": 2.0842,
+      "step": 2755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9140269033658167e-05,
+      "loss": 2.0869,
+      "step": 2756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9139479451839423e-05,
+      "loss": 2.061,
+      "step": 2757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9138689518897113e-05,
+      "loss": 2.0911,
+      "step": 2758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.913789923485195e-05,
+      "loss": 2.0967,
+      "step": 2759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9137108599724668e-05,
+      "loss": 2.0745,
+      "step": 2760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9136317613535987e-05,
+      "loss": 2.0852,
+      "step": 2761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.913552627630666e-05,
+      "loss": 2.0901,
+      "step": 2762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9134734588057432e-05,
+      "loss": 2.049,
+      "step": 2763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9133942548809073e-05,
+      "loss": 2.0713,
+      "step": 2764
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9133150158582343e-05,
+      "loss": 2.0598,
+      "step": 2765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9132357417398028e-05,
+      "loss": 2.0598,
+      "step": 2766
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.913156432527691e-05,
+      "loss": 2.1035,
+      "step": 2767
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9130770882239797e-05,
+      "loss": 2.1111,
+      "step": 2768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912997708830749e-05,
+      "loss": 2.0671,
+      "step": 2769
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9129182943500804e-05,
+      "loss": 2.0549,
+      "step": 2770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912838844784057e-05,
+      "loss": 2.0835,
+      "step": 2771
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912759360134762e-05,
+      "loss": 2.0847,
+      "step": 2772
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912679840404279e-05,
+      "loss": 2.0298,
+      "step": 2773
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912600285594695e-05,
+      "loss": 2.0159,
+      "step": 2774
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9125206957080945e-05,
+      "loss": 2.0605,
+      "step": 2775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9124410707465657e-05,
+      "loss": 2.053,
+      "step": 2776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9123614107121965e-05,
+      "loss": 2.0647,
+      "step": 2777
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.912281715607076e-05,
+      "loss": 2.0007,
+      "step": 2778
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9122019854332937e-05,
+      "loss": 2.103,
+      "step": 2779
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9121222201929404e-05,
+      "loss": 2.1008,
+      "step": 2780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9120424198881085e-05,
+      "loss": 2.0798,
+      "step": 2781
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.91196258452089e-05,
+      "loss": 2.0747,
+      "step": 2782
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9118827140933787e-05,
+      "loss": 2.0913,
+      "step": 2783
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9118028086076692e-05,
+      "loss": 2.0967,
+      "step": 2784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.911722868065857e-05,
+      "loss": 2.0554,
+      "step": 2785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.911642892470038e-05,
+      "loss": 2.0,
+      "step": 2786
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9115628818223103e-05,
+      "loss": 2.1465,
+      "step": 2787
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9114828361247715e-05,
+      "loss": 2.0833,
+      "step": 2788
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9114027553795203e-05,
+      "loss": 2.0261,
+      "step": 2789
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9113226395886576e-05,
+      "loss": 2.0298,
+      "step": 2790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9112424887542842e-05,
+      "loss": 2.0852,
+      "step": 2791
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9111623028785016e-05,
+      "loss": 2.1221,
+      "step": 2792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9110820819634126e-05,
+      "loss": 2.0681,
+      "step": 2793
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9110018260111208e-05,
+      "loss": 2.0654,
+      "step": 2794
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9109215350237316e-05,
+      "loss": 2.0959,
+      "step": 2795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9108412090033494e-05,
+      "loss": 2.1108,
+      "step": 2796
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9107608479520816e-05,
+      "loss": 2.0591,
+      "step": 2797
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9106804518720346e-05,
+      "loss": 2.0676,
+      "step": 2798
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.910600020765318e-05,
+      "loss": 2.0549,
+      "step": 2799
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9105195546340398e-05,
+      "loss": 2.0566,
+      "step": 2800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.910439053480311e-05,
+      "loss": 2.1082,
+      "step": 2801
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.910358517306242e-05,
+      "loss": 2.106,
+      "step": 2802
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9102779461139452e-05,
+      "loss": 2.0833,
+      "step": 2803
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9101973399055334e-05,
+      "loss": 2.0642,
+      "step": 2804
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9101166986831204e-05,
+      "loss": 2.0374,
+      "step": 2805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9100360224488204e-05,
+      "loss": 2.033,
+      "step": 2806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9099553112047497e-05,
+      "loss": 2.0879,
+      "step": 2807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9098745649530247e-05,
+      "loss": 2.0894,
+      "step": 2808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9097937836957632e-05,
+      "loss": 2.0669,
+      "step": 2809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.909712967435083e-05,
+      "loss": 2.0454,
+      "step": 2810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9096321161731033e-05,
+      "loss": 2.085,
+      "step": 2811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9095512299119452e-05,
+      "loss": 2.071,
+      "step": 2812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.909470308653729e-05,
+      "loss": 2.0747,
+      "step": 2813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9093893524005773e-05,
+      "loss": 2.0608,
+      "step": 2814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9093083611546127e-05,
+      "loss": 2.0762,
+      "step": 2815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9092273349179594e-05,
+      "loss": 2.0676,
+      "step": 2816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.909146273692742e-05,
+      "loss": 2.0818,
+      "step": 2817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9090651774810864e-05,
+      "loss": 2.0974,
+      "step": 2818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9089840462851193e-05,
+      "loss": 2.0608,
+      "step": 2819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.908902880106968e-05,
+      "loss": 2.074,
+      "step": 2820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.908821678948761e-05,
+      "loss": 2.0769,
+      "step": 2821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9087404428126284e-05,
+      "loss": 2.0979,
+      "step": 2822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9086591717006995e-05,
+      "loss": 2.0598,
+      "step": 2823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.908577865615106e-05,
+      "loss": 2.0874,
+      "step": 2824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.90849652455798e-05,
+      "loss": 2.0791,
+      "step": 2825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9084151485314552e-05,
+      "loss": 2.0547,
+      "step": 2826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9083337375376645e-05,
+      "loss": 2.0449,
+      "step": 2827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9082522915787437e-05,
+      "loss": 2.0288,
+      "step": 2828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.908170810656828e-05,
+      "loss": 2.0918,
+      "step": 2829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9080892947740546e-05,
+      "loss": 2.1123,
+      "step": 2830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9080077439325612e-05,
+      "loss": 2.1438,
+      "step": 2831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9079261581344856e-05,
+      "loss": 2.1299,
+      "step": 2832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.907844537381968e-05,
+      "loss": 2.0466,
+      "step": 2833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.907762881677149e-05,
+      "loss": 2.0781,
+      "step": 2834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9076811910221696e-05,
+      "loss": 2.0576,
+      "step": 2835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9075994654191715e-05,
+      "loss": 2.0527,
+      "step": 2836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9075177048702988e-05,
+      "loss": 2.0391,
+      "step": 2837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.907435909377695e-05,
+      "loss": 2.0309,
+      "step": 2838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9073540789435057e-05,
+      "loss": 2.0942,
+      "step": 2839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9072722135698755e-05,
+      "loss": 2.0923,
+      "step": 2840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.907190313258953e-05,
+      "loss": 2.0657,
+      "step": 2841
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9071083780128844e-05,
+      "loss": 2.0371,
+      "step": 2842
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9070264078338195e-05,
+      "loss": 2.0413,
+      "step": 2843
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.906944402723907e-05,
+      "loss": 2.0195,
+      "step": 2844
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9068623626852975e-05,
+      "loss": 2.0542,
+      "step": 2845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.906780287720143e-05,
+      "loss": 2.0884,
+      "step": 2846
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9066981778305954e-05,
+      "loss": 2.1113,
+      "step": 2847
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.906616033018808e-05,
+      "loss": 2.0911,
+      "step": 2848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9065338532869347e-05,
+      "loss": 2.0364,
+      "step": 2849
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9064516386371314e-05,
+      "loss": 2.0359,
+      "step": 2850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.906369389071553e-05,
+      "loss": 2.0452,
+      "step": 2851
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9062871045923575e-05,
+      "loss": 2.0581,
+      "step": 2852
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9062047852017014e-05,
+      "loss": 2.0657,
+      "step": 2853
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9061224309017444e-05,
+      "loss": 2.075,
+      "step": 2854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.906040041694646e-05,
+      "loss": 2.1118,
+      "step": 2855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905957617582567e-05,
+      "loss": 2.0938,
+      "step": 2856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905875158567668e-05,
+      "loss": 2.0332,
+      "step": 2857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9057926646521122e-05,
+      "loss": 2.0225,
+      "step": 2858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9057101358380625e-05,
+      "loss": 2.0754,
+      "step": 2859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9056275721276835e-05,
+      "loss": 2.0581,
+      "step": 2860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9055449735231402e-05,
+      "loss": 2.0608,
+      "step": 2861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905462340026598e-05,
+      "loss": 2.0229,
+      "step": 2862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905379671640225e-05,
+      "loss": 2.0898,
+      "step": 2863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9052969683661877e-05,
+      "loss": 2.0889,
+      "step": 2864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905214230206657e-05,
+      "loss": 2.0615,
+      "step": 2865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9051314571638005e-05,
+      "loss": 2.0312,
+      "step": 2866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.905048649239789e-05,
+      "loss": 2.0555,
+      "step": 2867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9049658064367957e-05,
+      "loss": 2.0493,
+      "step": 2868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9048829287569916e-05,
+      "loss": 2.1021,
+      "step": 2869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9048000162025504e-05,
+      "loss": 2.1172,
+      "step": 2870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9047170687756472e-05,
+      "loss": 2.0437,
+      "step": 2871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.904634086478456e-05,
+      "loss": 2.02,
+      "step": 2872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9045510693131532e-05,
+      "loss": 2.0002,
+      "step": 2873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9044680172819165e-05,
+      "loss": 2.0183,
+      "step": 2874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9043849303869233e-05,
+      "loss": 2.0396,
+      "step": 2875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9043018086303522e-05,
+      "loss": 2.0605,
+      "step": 2876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9042186520143834e-05,
+      "loss": 2.0464,
+      "step": 2877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9041354605411975e-05,
+      "loss": 2.0544,
+      "step": 2878
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.904052234212976e-05,
+      "loss": 2.0649,
+      "step": 2879
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903968973031902e-05,
+      "loss": 2.1042,
+      "step": 2880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903885677000158e-05,
+      "loss": 2.0669,
+      "step": 2881
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903802346119929e-05,
+      "loss": 2.0508,
+      "step": 2882
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9037189803933997e-05,
+      "loss": 2.0312,
+      "step": 2883
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903635579822757e-05,
+      "loss": 2.0437,
+      "step": 2884
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9035521444101876e-05,
+      "loss": 2.0811,
+      "step": 2885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9034686741578792e-05,
+      "loss": 2.0852,
+      "step": 2886
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903385169068021e-05,
+      "loss": 2.0789,
+      "step": 2887
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9033016291428027e-05,
+      "loss": 2.0457,
+      "step": 2888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9032180543844156e-05,
+      "loss": 2.135,
+      "step": 2889
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.903134444795051e-05,
+      "loss": 2.1101,
+      "step": 2890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9030508003769006e-05,
+      "loss": 2.073,
+      "step": 2891
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.902967121132159e-05,
+      "loss": 2.0693,
+      "step": 2892
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9028834070630207e-05,
+      "loss": 2.0886,
+      "step": 2893
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.90279965817168e-05,
+      "loss": 2.1101,
+      "step": 2894
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.902715874460334e-05,
+      "loss": 2.0317,
+      "step": 2895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9026320559311794e-05,
+      "loss": 2.0374,
+      "step": 2896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9025482025864144e-05,
+      "loss": 2.0562,
+      "step": 2897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.902464314428238e-05,
+      "loss": 2.0369,
+      "step": 2898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9023803914588493e-05,
+      "loss": 2.0413,
+      "step": 2899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9022964336804506e-05,
+      "loss": 2.0471,
+      "step": 2900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.902212441095242e-05,
+      "loss": 2.085,
+      "step": 2901
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9021284137054273e-05,
+      "loss": 2.0938,
+      "step": 2902
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9020443515132094e-05,
+      "loss": 2.0515,
+      "step": 2903
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.901960254520793e-05,
+      "loss": 2.0559,
+      "step": 2904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9018761227303833e-05,
+      "loss": 2.0977,
+      "step": 2905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9017919561441868e-05,
+      "loss": 2.084,
+      "step": 2906
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.90170775476441e-05,
+      "loss": 1.9658,
+      "step": 2907
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9016235185932616e-05,
+      "loss": 1.9744,
+      "step": 2908
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9015392476329508e-05,
+      "loss": 2.1021,
+      "step": 2909
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9014549418856873e-05,
+      "loss": 2.0938,
+      "step": 2910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9013706013536812e-05,
+      "loss": 1.9739,
+      "step": 2911
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.901286226039145e-05,
+      "loss": 2.0,
+      "step": 2912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9012018159442916e-05,
+      "loss": 2.1055,
+      "step": 2913
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9011173710713335e-05,
+      "loss": 2.0764,
+      "step": 2914
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9010328914224865e-05,
+      "loss": 1.9922,
+      "step": 2915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9009483769999645e-05,
+      "loss": 2.0164,
+      "step": 2916
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900863827805985e-05,
+      "loss": 2.0801,
+      "step": 2917
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9007792438427647e-05,
+      "loss": 2.0728,
+      "step": 2918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900694625112522e-05,
+      "loss": 2.0681,
+      "step": 2919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9006099716174753e-05,
+      "loss": 2.0686,
+      "step": 2920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9005252833598456e-05,
+      "loss": 2.0217,
+      "step": 2921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9004405603418523e-05,
+      "loss": 2.0034,
+      "step": 2922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900355802565718e-05,
+      "loss": 2.0327,
+      "step": 2923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900271010033666e-05,
+      "loss": 2.0178,
+      "step": 2924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900186182747919e-05,
+      "loss": 2.0701,
+      "step": 2925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900101320710701e-05,
+      "loss": 2.0679,
+      "step": 2926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.900016423924239e-05,
+      "loss": 2.0935,
+      "step": 2927
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8999314923907584e-05,
+      "loss": 2.0845,
+      "step": 2928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.899846526112486e-05,
+      "loss": 2.0381,
+      "step": 2929
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8997615250916512e-05,
+      "loss": 2.0278,
+      "step": 2930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8996764893304814e-05,
+      "loss": 2.0469,
+      "step": 2931
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.899591418831208e-05,
+      "loss": 2.0381,
+      "step": 2932
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.899506313596061e-05,
+      "loss": 2.0601,
+      "step": 2933
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8994211736272728e-05,
+      "loss": 2.0884,
+      "step": 2934
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.899335998927076e-05,
+      "loss": 2.0129,
+      "step": 2935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8992507894977035e-05,
+      "loss": 2.04,
+      "step": 2936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8991655453413903e-05,
+      "loss": 2.0005,
+      "step": 2937
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.899080266460372e-05,
+      "loss": 2.0288,
+      "step": 2938
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8989949528568846e-05,
+      "loss": 2.0867,
+      "step": 2939
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898909604533166e-05,
+      "loss": 2.0813,
+      "step": 2940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898824221491454e-05,
+      "loss": 2.0942,
+      "step": 2941
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898738803733987e-05,
+      "loss": 2.1025,
+      "step": 2942
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8986533512630057e-05,
+      "loss": 2.1143,
+      "step": 2943
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898567864080751e-05,
+      "loss": 2.0754,
+      "step": 2944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8984823421894646e-05,
+      "loss": 2.0425,
+      "step": 2945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898396785591389e-05,
+      "loss": 2.054,
+      "step": 2946
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8983111942887682e-05,
+      "loss": 2.0962,
+      "step": 2947
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8982255682838462e-05,
+      "loss": 2.1157,
+      "step": 2948
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898139907578869e-05,
+      "loss": 2.0872,
+      "step": 2949
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.898054212176083e-05,
+      "loss": 2.0544,
+      "step": 2950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8979684820777344e-05,
+      "loss": 2.0044,
+      "step": 2951
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8978827172860727e-05,
+      "loss": 2.0352,
+      "step": 2952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.897796917803347e-05,
+      "loss": 2.0737,
+      "step": 2953
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8977110836318056e-05,
+      "loss": 2.0781,
+      "step": 2954
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.897625214773701e-05,
+      "loss": 2.0898,
+      "step": 2955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8975393112312847e-05,
+      "loss": 2.093,
+      "step": 2956
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8974533730068087e-05,
+      "loss": 2.0126,
+      "step": 2957
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8973674001025278e-05,
+      "loss": 2.0579,
+      "step": 2958
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8972813925206957e-05,
+      "loss": 2.0457,
+      "step": 2959
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.897195350263568e-05,
+      "loss": 2.052,
+      "step": 2960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8971092733334013e-05,
+      "loss": 2.0872,
+      "step": 2961
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8970231617324528e-05,
+      "loss": 2.0991,
+      "step": 2962
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.89693701546298e-05,
+      "loss": 2.0103,
+      "step": 2963
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8968508345272433e-05,
+      "loss": 2.0876,
+      "step": 2964
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8967646189275017e-05,
+      "loss": 2.0549,
+      "step": 2965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.896678368666016e-05,
+      "loss": 2.0361,
+      "step": 2966
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8965920837450492e-05,
+      "loss": 2.0552,
+      "step": 2967
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8965057641668627e-05,
+      "loss": 2.0474,
+      "step": 2968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8964194099337205e-05,
+      "loss": 2.0366,
+      "step": 2969
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8963330210478875e-05,
+      "loss": 2.03,
+      "step": 2970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.896246597511629e-05,
+      "loss": 2.0004,
+      "step": 2971
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8961601393272105e-05,
+      "loss": 2.0168,
+      "step": 2972
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.896073646496901e-05,
+      "loss": 2.0332,
+      "step": 2973
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8959871190229672e-05,
+      "loss": 2.0396,
+      "step": 2974
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8959005569076787e-05,
+      "loss": 2.0781,
+      "step": 2975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8958139601533058e-05,
+      "loss": 2.0854,
+      "step": 2976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.895727328762119e-05,
+      "loss": 2.0618,
+      "step": 2977
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8956406627363903e-05,
+      "loss": 2.0249,
+      "step": 2978
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.895553962078392e-05,
+      "loss": 2.0701,
+      "step": 2979
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.895467226790398e-05,
+      "loss": 2.0669,
+      "step": 2980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8953804568746837e-05,
+      "loss": 2.0349,
+      "step": 2981
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8952936523335224e-05,
+      "loss": 2.0361,
+      "step": 2982
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8952068131691925e-05,
+      "loss": 2.1099,
+      "step": 2983
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8951199393839704e-05,
+      "loss": 2.1301,
+      "step": 2984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8950330309801347e-05,
+      "loss": 2.0601,
+      "step": 2985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8949460879599638e-05,
+      "loss": 2.0637,
+      "step": 2986
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.894859110325738e-05,
+      "loss": 2.1233,
+      "step": 2987
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8947720980797385e-05,
+      "loss": 2.1113,
+      "step": 2988
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8946850512242466e-05,
+      "loss": 2.1008,
+      "step": 2989
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.894597969761545e-05,
+      "loss": 2.0718,
+      "step": 2990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.894510853693918e-05,
+      "loss": 2.0701,
+      "step": 2991
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8944237030236495e-05,
+      "loss": 2.052,
+      "step": 2992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8943365177530248e-05,
+      "loss": 2.1072,
+      "step": 2993
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.894249297884331e-05,
+      "loss": 2.0835,
+      "step": 2994
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.894162043419854e-05,
+      "loss": 2.0417,
+      "step": 2995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8940747543618838e-05,
+      "loss": 2.0486,
+      "step": 2996
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.893987430712708e-05,
+      "loss": 2.0742,
+      "step": 2997
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.893900072474617e-05,
+      "loss": 2.0818,
+      "step": 2998
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8938126796499014e-05,
+      "loss": 2.0916,
+      "step": 2999
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8937252522408536e-05,
+      "loss": 2.061,
+      "step": 3000
+    },
+    {
+      "Bleu_1": 0.7715497798904677,
+      "Bleu_2": 0.6109713412545874,
+      "Bleu_3": 0.4695130520254875,
+      "Bleu_4": 0.356803687352363,
+      "CIDEr": 1.200772281669958,
+      "METEOR": 0.2781862864567016,
+      "ROUGE_L": 0.5575132553338945,
+      "epoch": 0.15,
+      "step": 3000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.893637790249766e-05,
+      "loss": 2.0496,
+      "step": 3001
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8935502936789315e-05,
+      "loss": 2.061,
+      "step": 3002
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8934627625306457e-05,
+      "loss": 2.0206,
+      "step": 3003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8933751968072034e-05,
+      "loss": 2.063,
+      "step": 3004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8932875965109015e-05,
+      "loss": 2.0376,
+      "step": 3005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.893199961644036e-05,
+      "loss": 2.0999,
+      "step": 3006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8931122922089067e-05,
+      "loss": 2.0044,
+      "step": 3007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.893024588207811e-05,
+      "loss": 2.0679,
+      "step": 3008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8929368496430496e-05,
+      "loss": 2.0398,
+      "step": 3009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.892849076516923e-05,
+      "loss": 2.0249,
+      "step": 3010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.892761268831734e-05,
+      "loss": 2.0085,
+      "step": 3011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8926734265897837e-05,
+      "loss": 2.0039,
+      "step": 3012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8925855497933767e-05,
+      "loss": 2.0613,
+      "step": 3013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8924976384448176e-05,
+      "loss": 2.0542,
+      "step": 3014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8924096925464107e-05,
+      "loss": 2.0916,
+      "step": 3015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8923217121004634e-05,
+      "loss": 2.0625,
+      "step": 3016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8922336971092826e-05,
+      "loss": 2.0583,
+      "step": 3017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8921456475751756e-05,
+      "loss": 2.0698,
+      "step": 3018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8920575635004528e-05,
+      "loss": 2.0542,
+      "step": 3019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8919694448874228e-05,
+      "loss": 2.0537,
+      "step": 3020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8918812917383975e-05,
+      "loss": 2.0217,
+      "step": 3021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.891793104055687e-05,
+      "loss": 2.0227,
+      "step": 3022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.891704881841606e-05,
+      "loss": 2.1111,
+      "step": 3023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8916166250984664e-05,
+      "loss": 2.0801,
+      "step": 3024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8915283338285837e-05,
+      "loss": 2.0757,
+      "step": 3025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8914400080342724e-05,
+      "loss": 2.0659,
+      "step": 3026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.891351647717849e-05,
+      "loss": 2.063,
+      "step": 3027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8912632528816312e-05,
+      "loss": 2.1133,
+      "step": 3028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8911748235279367e-05,
+      "loss": 2.0164,
+      "step": 3029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.891086359659084e-05,
+      "loss": 2.0027,
+      "step": 3030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8909978612773933e-05,
+      "loss": 2.0623,
+      "step": 3031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.890909328385186e-05,
+      "loss": 2.073,
+      "step": 3032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.890820760984782e-05,
+      "loss": 2.0436,
+      "step": 3033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8907321590785058e-05,
+      "loss": 2.0544,
+      "step": 3034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8906435226686804e-05,
+      "loss": 2.0813,
+      "step": 3035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8905548517576295e-05,
+      "loss": 2.114,
+      "step": 3036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8904661463476786e-05,
+      "loss": 2.0393,
+      "step": 3037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8903774064411545e-05,
+      "loss": 2.0198,
+      "step": 3038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.890288632040384e-05,
+      "loss": 2.083,
+      "step": 3039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8901998231476947e-05,
+      "loss": 2.0835,
+      "step": 3040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8901109797654157e-05,
+      "loss": 2.084,
+      "step": 3041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8900221018958775e-05,
+      "loss": 2.1067,
+      "step": 3042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8899331895414098e-05,
+      "loss": 2.085,
+      "step": 3043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8898442427043443e-05,
+      "loss": 2.0837,
+      "step": 3044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8897552613870145e-05,
+      "loss": 2.051,
+      "step": 3045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.889666245591753e-05,
+      "loss": 2.0513,
+      "step": 3046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8895771953208947e-05,
+      "loss": 2.1018,
+      "step": 3047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8894881105767742e-05,
+      "loss": 2.0974,
+      "step": 3048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.889398991361728e-05,
+      "loss": 2.0911,
+      "step": 3049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.889309837678093e-05,
+      "loss": 2.0803,
+      "step": 3050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8892206495282076e-05,
+      "loss": 2.0457,
+      "step": 3051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8891314269144103e-05,
+      "loss": 2.0662,
+      "step": 3052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8890421698390406e-05,
+      "loss": 2.0651,
+      "step": 3053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8889528783044394e-05,
+      "loss": 2.0608,
+      "step": 3054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8888635523129485e-05,
+      "loss": 2.1121,
+      "step": 3055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8887741918669102e-05,
+      "loss": 2.1228,
+      "step": 3056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8886847969686677e-05,
+      "loss": 2.0604,
+      "step": 3057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8885953676205656e-05,
+      "loss": 2.0718,
+      "step": 3058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8885059038249487e-05,
+      "loss": 2.0742,
+      "step": 3059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.888416405584163e-05,
+      "loss": 2.1421,
+      "step": 3060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8883268729005565e-05,
+      "loss": 2.0881,
+      "step": 3061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8882373057764756e-05,
+      "loss": 2.0808,
+      "step": 3062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8881477042142706e-05,
+      "loss": 2.0793,
+      "step": 3063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.88805806821629e-05,
+      "loss": 2.0781,
+      "step": 3064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887968397784885e-05,
+      "loss": 2.0913,
+      "step": 3065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8878786929224068e-05,
+      "loss": 2.1104,
+      "step": 3066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887788953631208e-05,
+      "loss": 2.0171,
+      "step": 3067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8876991799136423e-05,
+      "loss": 2.0374,
+      "step": 3068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887609371772063e-05,
+      "loss": 2.0203,
+      "step": 3069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887519529208826e-05,
+      "loss": 2.0803,
+      "step": 3070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8874296522262868e-05,
+      "loss": 2.1226,
+      "step": 3071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887339740826803e-05,
+      "loss": 2.0996,
+      "step": 3072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8872497950127313e-05,
+      "loss": 1.9893,
+      "step": 3073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887159814786431e-05,
+      "loss": 2.0378,
+      "step": 3074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.887069800150263e-05,
+      "loss": 2.0518,
+      "step": 3075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8869797511065854e-05,
+      "loss": 2.0417,
+      "step": 3076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8868896676577616e-05,
+      "loss": 2.0277,
+      "step": 3077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8867995498061528e-05,
+      "loss": 2.0627,
+      "step": 3078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8867093975541226e-05,
+      "loss": 2.1177,
+      "step": 3079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8866192109040353e-05,
+      "loss": 2.1331,
+      "step": 3080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.886528989858256e-05,
+      "loss": 2.0862,
+      "step": 3081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.88643873441915e-05,
+      "loss": 2.0745,
+      "step": 3082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.886348444589085e-05,
+      "loss": 2.0118,
+      "step": 3083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8862581203704283e-05,
+      "loss": 2.0254,
+      "step": 3084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8861677617655486e-05,
+      "loss": 2.0581,
+      "step": 3085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8860773687768154e-05,
+      "loss": 2.0691,
+      "step": 3086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8859869414065995e-05,
+      "loss": 2.0532,
+      "step": 3087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8858964796572712e-05,
+      "loss": 2.0562,
+      "step": 3088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8858059835312046e-05,
+      "loss": 2.0417,
+      "step": 3089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.885715453030771e-05,
+      "loss": 2.0359,
+      "step": 3090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8856248881583456e-05,
+      "loss": 2.0737,
+      "step": 3091
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8855342889163024e-05,
+      "loss": 2.0498,
+      "step": 3092
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8854436553070184e-05,
+      "loss": 2.0852,
+      "step": 3093
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8853529873328696e-05,
+      "loss": 2.0833,
+      "step": 3094
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.885262284996234e-05,
+      "loss": 2.0957,
+      "step": 3095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8851715482994898e-05,
+      "loss": 2.0723,
+      "step": 3096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.885080777245017e-05,
+      "loss": 2.0598,
+      "step": 3097
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8849899718351953e-05,
+      "loss": 2.0503,
+      "step": 3098
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.884899132072406e-05,
+      "loss": 2.0989,
+      "step": 3099
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8848082579590327e-05,
+      "loss": 2.0747,
+      "step": 3100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8847173494974565e-05,
+      "loss": 2.0371,
+      "step": 3101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8846264066900624e-05,
+      "loss": 2.0552,
+      "step": 3102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8845354295392353e-05,
+      "loss": 2.0491,
+      "step": 3103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.884444418047361e-05,
+      "loss": 2.0701,
+      "step": 3104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8843533722168252e-05,
+      "loss": 2.0693,
+      "step": 3105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8842622920500168e-05,
+      "loss": 2.0256,
+      "step": 3106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8841711775493234e-05,
+      "loss": 2.0737,
+      "step": 3107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.884080028717135e-05,
+      "loss": 2.0854,
+      "step": 3108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.883988845555841e-05,
+      "loss": 2.0913,
+      "step": 3109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8838976280678337e-05,
+      "loss": 2.1008,
+      "step": 3110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8838063762555038e-05,
+      "loss": 2.0955,
+      "step": 3111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8837150901212453e-05,
+      "loss": 2.1267,
+      "step": 3112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.883623769667452e-05,
+      "loss": 2.115,
+      "step": 3113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8835324148965188e-05,
+      "loss": 2.1052,
+      "step": 3114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.883441025810841e-05,
+      "loss": 2.05,
+      "step": 3115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8833496024128147e-05,
+      "loss": 2.0261,
+      "step": 3116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8832581447048384e-05,
+      "loss": 2.1045,
+      "step": 3117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.88316665268931e-05,
+      "loss": 2.0857,
+      "step": 3118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8830751263686284e-05,
+      "loss": 2.0796,
+      "step": 3119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8829835657451943e-05,
+      "loss": 2.0569,
+      "step": 3120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8828919708214083e-05,
+      "loss": 2.0234,
+      "step": 3121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8828003415996733e-05,
+      "loss": 2.053,
+      "step": 3122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.882708678082391e-05,
+      "loss": 1.9944,
+      "step": 3123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8826169802719664e-05,
+      "loss": 2.0117,
+      "step": 3124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8825252481708025e-05,
+      "loss": 2.0703,
+      "step": 3125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8824334817813066e-05,
+      "loss": 2.0208,
+      "step": 3126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8823416811058838e-05,
+      "loss": 2.0195,
+      "step": 3127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8822498461469426e-05,
+      "loss": 2.0532,
+      "step": 3128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8821579769068905e-05,
+      "loss": 2.0776,
+      "step": 3129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.882066073388137e-05,
+      "loss": 2.0745,
+      "step": 3130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8819741355930922e-05,
+      "loss": 2.0356,
+      "step": 3131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.881882163524167e-05,
+      "loss": 2.0857,
+      "step": 3132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8817901571837723e-05,
+      "loss": 2.0005,
+      "step": 3133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.881698116574323e-05,
+      "loss": 2.0051,
+      "step": 3134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.881606041698231e-05,
+      "loss": 2.1299,
+      "step": 3135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.881513932557911e-05,
+      "loss": 2.1426,
+      "step": 3136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8814217891557796e-05,
+      "loss": 2.0195,
+      "step": 3137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8813296114942513e-05,
+      "loss": 2.0217,
+      "step": 3138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.881237399575745e-05,
+      "loss": 2.053,
+      "step": 3139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8811451534026788e-05,
+      "loss": 2.0566,
+      "step": 3140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8810528729774705e-05,
+      "loss": 2.0691,
+      "step": 3141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8809605583025412e-05,
+      "loss": 2.064,
+      "step": 3142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.880868209380311e-05,
+      "loss": 2.0808,
+      "step": 3143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8807758262132022e-05,
+      "loss": 2.0544,
+      "step": 3144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8806834088036372e-05,
+      "loss": 2.0505,
+      "step": 3145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8805909571540393e-05,
+      "loss": 2.0381,
+      "step": 3146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8804984712668334e-05,
+      "loss": 2.0208,
+      "step": 3147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8804059511444445e-05,
+      "loss": 2.0564,
+      "step": 3148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.880313396789299e-05,
+      "loss": 2.0818,
+      "step": 3149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.880220808203824e-05,
+      "loss": 2.0759,
+      "step": 3150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8801281853904474e-05,
+      "loss": 2.0471,
+      "step": 3151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8800355283515985e-05,
+      "loss": 2.0608,
+      "step": 3152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8799428370897066e-05,
+      "loss": 2.0488,
+      "step": 3153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8798501116072026e-05,
+      "loss": 2.053,
+      "step": 3154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8797573519065185e-05,
+      "loss": 2.0972,
+      "step": 3155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8796645579900864e-05,
+      "loss": 2.125,
+      "step": 3156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.87957172986034e-05,
+      "loss": 2.0972,
+      "step": 3157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.879478867519713e-05,
+      "loss": 2.0837,
+      "step": 3158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8793859709706413e-05,
+      "loss": 1.9958,
+      "step": 3159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8792930402155606e-05,
+      "loss": 2.0093,
+      "step": 3160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8792000752569083e-05,
+      "loss": 2.0969,
+      "step": 3161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.879107076097122e-05,
+      "loss": 2.0771,
+      "step": 3162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8790140427386408e-05,
+      "loss": 2.0352,
+      "step": 3163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8789209751839034e-05,
+      "loss": 2.0029,
+      "step": 3164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8788278734353516e-05,
+      "loss": 2.0642,
+      "step": 3165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8787347374954262e-05,
+      "loss": 2.0488,
+      "step": 3166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8786415673665703e-05,
+      "loss": 2.051,
+      "step": 3167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.878548363051226e-05,
+      "loss": 2.0378,
+      "step": 3168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8784551245518386e-05,
+      "loss": 2.0488,
+      "step": 3169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8783618518708526e-05,
+      "loss": 2.0776,
+      "step": 3170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.878268545010714e-05,
+      "loss": 2.0647,
+      "step": 3171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8781752039738698e-05,
+      "loss": 2.0952,
+      "step": 3172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8780818287627676e-05,
+      "loss": 2.0259,
+      "step": 3173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8779884193798565e-05,
+      "loss": 2.0576,
+      "step": 3174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8778949758275855e-05,
+      "loss": 2.0332,
+      "step": 3175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.877801498108405e-05,
+      "loss": 2.0425,
+      "step": 3176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.877707986224767e-05,
+      "loss": 2.0859,
+      "step": 3177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.877614440179123e-05,
+      "loss": 2.1003,
+      "step": 3178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.877520859973927e-05,
+      "loss": 2.0244,
+      "step": 3179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8774272456116322e-05,
+      "loss": 2.0654,
+      "step": 3180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.877333597094694e-05,
+      "loss": 2.0894,
+      "step": 3181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8772399144255677e-05,
+      "loss": 2.1006,
+      "step": 3182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8771461976067104e-05,
+      "loss": 2.0437,
+      "step": 3183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.87705244664058e-05,
+      "loss": 2.0676,
+      "step": 3184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876958661529635e-05,
+      "loss": 2.0759,
+      "step": 3185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8768648422763338e-05,
+      "loss": 2.0747,
+      "step": 3186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876770988883138e-05,
+      "loss": 2.0684,
+      "step": 3187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876677101352508e-05,
+      "loss": 2.0889,
+      "step": 3188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876583179686906e-05,
+      "loss": 2.031,
+      "step": 3189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8764892238887954e-05,
+      "loss": 2.0142,
+      "step": 3190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876395233960639e-05,
+      "loss": 2.0876,
+      "step": 3191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8763012099049033e-05,
+      "loss": 2.1208,
+      "step": 3192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8762071517240528e-05,
+      "loss": 2.0476,
+      "step": 3193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876113059420554e-05,
+      "loss": 2.0879,
+      "step": 3194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.876018932996875e-05,
+      "loss": 1.9719,
+      "step": 3195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.875924772455484e-05,
+      "loss": 1.9968,
+      "step": 3196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8758305777988496e-05,
+      "loss": 2.0708,
+      "step": 3197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8757363490294422e-05,
+      "loss": 2.0991,
+      "step": 3198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8756420861497332e-05,
+      "loss": 2.0864,
+      "step": 3199
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8755477891621947e-05,
+      "loss": 2.126,
+      "step": 3200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8754534580692992e-05,
+      "loss": 2.0874,
+      "step": 3201
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8753590928735202e-05,
+      "loss": 2.0706,
+      "step": 3202
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.875264693577333e-05,
+      "loss": 2.0884,
+      "step": 3203
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.875170260183212e-05,
+      "loss": 2.1035,
+      "step": 3204
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8750757926936345e-05,
+      "loss": 2.0415,
+      "step": 3205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8749812911110776e-05,
+      "loss": 2.0781,
+      "step": 3206
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8748867554380196e-05,
+      "loss": 2.1248,
+      "step": 3207
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8747921856769395e-05,
+      "loss": 2.1394,
+      "step": 3208
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.874697581830317e-05,
+      "loss": 2.1189,
+      "step": 3209
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8746029439006336e-05,
+      "loss": 2.0723,
+      "step": 3210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.87450827189037e-05,
+      "loss": 2.0886,
+      "step": 3211
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8744135658020104e-05,
+      "loss": 2.0913,
+      "step": 3212
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.874318825638037e-05,
+      "loss": 2.0205,
+      "step": 3213
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8742240514009347e-05,
+      "loss": 2.0413,
+      "step": 3214
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8741292430931888e-05,
+      "loss": 2.0503,
+      "step": 3215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.874034400717286e-05,
+      "loss": 2.0522,
+      "step": 3216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.873939524275713e-05,
+      "loss": 2.0548,
+      "step": 3217
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.873844613770958e-05,
+      "loss": 2.0935,
+      "step": 3218
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.87374966920551e-05,
+      "loss": 2.0581,
+      "step": 3219
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.873654690581858e-05,
+      "loss": 2.0688,
+      "step": 3220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8735596779024935e-05,
+      "loss": 2.0613,
+      "step": 3221
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.873464631169908e-05,
+      "loss": 2.0591,
+      "step": 3222
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.873369550386594e-05,
+      "loss": 2.0815,
+      "step": 3223
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8732744355550445e-05,
+      "loss": 2.0676,
+      "step": 3224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8731792866777542e-05,
+      "loss": 2.0845,
+      "step": 3225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8730841037572188e-05,
+      "loss": 2.0864,
+      "step": 3226
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8729888867959328e-05,
+      "loss": 2.1172,
+      "step": 3227
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8728936357963944e-05,
+      "loss": 2.0789,
+      "step": 3228
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8727983507611013e-05,
+      "loss": 2.0818,
+      "step": 3229
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.872703031692552e-05,
+      "loss": 2.0781,
+      "step": 3230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8726076785932456e-05,
+      "loss": 2.0769,
+      "step": 3231
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8725122914656836e-05,
+      "loss": 2.1245,
+      "step": 3232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8724168703123673e-05,
+      "loss": 2.0918,
+      "step": 3233
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8723214151357986e-05,
+      "loss": 2.0774,
+      "step": 3234
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8722259259384804e-05,
+      "loss": 2.0789,
+      "step": 3235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8721304027229175e-05,
+      "loss": 2.0708,
+      "step": 3236
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8720348454916146e-05,
+      "loss": 2.0876,
+      "step": 3237
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.871939254247078e-05,
+      "loss": 2.0403,
+      "step": 3238
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8718436289918135e-05,
+      "loss": 2.0825,
+      "step": 3239
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8717479697283296e-05,
+      "loss": 2.05,
+      "step": 3240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8716522764591344e-05,
+      "loss": 2.0537,
+      "step": 3241
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8715565491867375e-05,
+      "loss": 2.0833,
+      "step": 3242
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8714607879136496e-05,
+      "loss": 2.0093,
+      "step": 3243
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8713649926423807e-05,
+      "loss": 2.04,
+      "step": 3244
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8712691633754448e-05,
+      "loss": 2.0415,
+      "step": 3245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.871173300115354e-05,
+      "loss": 2.0469,
+      "step": 3246
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8710774028646212e-05,
+      "loss": 2.0151,
+      "step": 3247
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8709814716257625e-05,
+      "loss": 2.0244,
+      "step": 3248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8708855064012932e-05,
+      "loss": 1.9871,
+      "step": 3249
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.87078950719373e-05,
+      "loss": 2.0051,
+      "step": 3250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8706934740055902e-05,
+      "loss": 2.0798,
+      "step": 3251
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.870597406839392e-05,
+      "loss": 2.071,
+      "step": 3252
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.870501305697655e-05,
+      "loss": 2.0745,
+      "step": 3253
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.870405170582899e-05,
+      "loss": 2.0723,
+      "step": 3254
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8703090014976456e-05,
+      "loss": 2.0696,
+      "step": 3255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.870212798444416e-05,
+      "loss": 2.0305,
+      "step": 3256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8701165614257336e-05,
+      "loss": 1.9758,
+      "step": 3257
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8700202904441217e-05,
+      "loss": 2.0198,
+      "step": 3258
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869923985502105e-05,
+      "loss": 2.0828,
+      "step": 3259
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869827646602209e-05,
+      "loss": 2.0811,
+      "step": 3260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8697312737469605e-05,
+      "loss": 2.0061,
+      "step": 3261
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869634866938886e-05,
+      "loss": 2.0442,
+      "step": 3262
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8695384261805144e-05,
+      "loss": 2.0452,
+      "step": 3263
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869441951474374e-05,
+      "loss": 2.0496,
+      "step": 3264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8693454428229954e-05,
+      "loss": 2.1128,
+      "step": 3265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869248900228909e-05,
+      "loss": 2.0801,
+      "step": 3266
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869152323694647e-05,
+      "loss": 2.1082,
+      "step": 3267
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.869055713222741e-05,
+      "loss": 2.0977,
+      "step": 3268
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8689590688157254e-05,
+      "loss": 1.9626,
+      "step": 3269
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8688623904761347e-05,
+      "loss": 1.9941,
+      "step": 3270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8687656782065035e-05,
+      "loss": 2.0447,
+      "step": 3271
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8686689320093684e-05,
+      "loss": 2.0378,
+      "step": 3272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8685721518872668e-05,
+      "loss": 2.0464,
+      "step": 3273
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8684753378427356e-05,
+      "loss": 2.0537,
+      "step": 3274
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8683784898783144e-05,
+      "loss": 2.0681,
+      "step": 3275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.868281607996543e-05,
+      "loss": 2.0752,
+      "step": 3276
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8681846921999618e-05,
+      "loss": 2.073,
+      "step": 3277
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.868087742491112e-05,
+      "loss": 2.0703,
+      "step": 3278
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.867990758872536e-05,
+      "loss": 2.0862,
+      "step": 3279
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8678937413467778e-05,
+      "loss": 2.0674,
+      "step": 3280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8677966899163813e-05,
+      "loss": 2.0872,
+      "step": 3281
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.867699604583891e-05,
+      "loss": 2.0703,
+      "step": 3282
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8676024853518536e-05,
+      "loss": 1.9885,
+      "step": 3283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8675053322228153e-05,
+      "loss": 2.02,
+      "step": 3284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8674081451993242e-05,
+      "loss": 2.0854,
+      "step": 3285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.867310924283929e-05,
+      "loss": 2.0696,
+      "step": 3286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.867213669479179e-05,
+      "loss": 2.0962,
+      "step": 3287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8671163807876243e-05,
+      "loss": 2.0869,
+      "step": 3288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8670190582118166e-05,
+      "loss": 2.0608,
+      "step": 3289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8669217017543082e-05,
+      "loss": 2.0527,
+      "step": 3290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.866824311417652e-05,
+      "loss": 1.9971,
+      "step": 3291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.866726887204402e-05,
+      "loss": 2.0325,
+      "step": 3292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8666294291171127e-05,
+      "loss": 2.0286,
+      "step": 3293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.86653193715834e-05,
+      "loss": 2.0054,
+      "step": 3294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8664344113306408e-05,
+      "loss": 2.1045,
+      "step": 3295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8663368516365725e-05,
+      "loss": 2.0986,
+      "step": 3296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8662392580786927e-05,
+      "loss": 2.0591,
+      "step": 3297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8661416306595617e-05,
+      "loss": 2.0757,
+      "step": 3298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.8660439693817397e-05,
+      "loss": 2.0242,
+      "step": 3299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.865946274247787e-05,
+      "loss": 1.9932,
+      "step": 3300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.865848545260266e-05,
+      "loss": 2.0315,
+      "step": 3301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8657507824217394e-05,
+      "loss": 2.0244,
+      "step": 3302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8656529857347706e-05,
+      "loss": 2.0568,
+      "step": 3303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8655551552019246e-05,
+      "loss": 2.0608,
+      "step": 3304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8654572908257673e-05,
+      "loss": 1.9753,
+      "step": 3305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8653593926088647e-05,
+      "loss": 2.02,
+      "step": 3306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8652614605537832e-05,
+      "loss": 2.0586,
+      "step": 3307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.865163494663092e-05,
+      "loss": 2.0667,
+      "step": 3308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8650654949393607e-05,
+      "loss": 2.0669,
+      "step": 3309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8649674613851577e-05,
+      "loss": 2.0735,
+      "step": 3310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8648693940030548e-05,
+      "loss": 2.053,
+      "step": 3311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8647712927956232e-05,
+      "loss": 2.0706,
+      "step": 3312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.864673157765436e-05,
+      "loss": 2.0806,
+      "step": 3313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.864574988915066e-05,
+      "loss": 2.1104,
+      "step": 3314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8644767862470885e-05,
+      "loss": 2.0576,
+      "step": 3315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8643785497640778e-05,
+      "loss": 2.0637,
+      "step": 3316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8642802794686106e-05,
+      "loss": 2.0818,
+      "step": 3317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.864181975363264e-05,
+      "loss": 2.0713,
+      "step": 3318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8640836374506155e-05,
+      "loss": 2.0105,
+      "step": 3319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863985265733244e-05,
+      "loss": 2.0242,
+      "step": 3320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8638868602137294e-05,
+      "loss": 2.0688,
+      "step": 3321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863788420894652e-05,
+      "loss": 2.0676,
+      "step": 3322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863689947778594e-05,
+      "loss": 2.0852,
+      "step": 3323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8635914408681365e-05,
+      "loss": 2.0588,
+      "step": 3324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863492900165863e-05,
+      "loss": 1.9941,
+      "step": 3325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863394325674359e-05,
+      "loss": 2.0527,
+      "step": 3326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863295717396208e-05,
+      "loss": 2.0742,
+      "step": 3327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863197075333996e-05,
+      "loss": 2.0808,
+      "step": 3328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.863098399490311e-05,
+      "loss": 2.092,
+      "step": 3329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8629996898677394e-05,
+      "loss": 2.1162,
+      "step": 3330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8629009464688697e-05,
+      "loss": 2.0813,
+      "step": 3331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8628021692962925e-05,
+      "loss": 2.0491,
+      "step": 3332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8627033583525974e-05,
+      "loss": 2.0835,
+      "step": 3333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.862604513640375e-05,
+      "loss": 2.1187,
+      "step": 3334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.862505635162218e-05,
+      "loss": 2.0232,
+      "step": 3335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.86240672292072e-05,
+      "loss": 2.0454,
+      "step": 3336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.862307776918474e-05,
+      "loss": 2.015,
+      "step": 3337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8622087971580746e-05,
+      "loss": 2.0579,
+      "step": 3338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8621097836421184e-05,
+      "loss": 2.063,
+      "step": 3339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.862010736373201e-05,
+      "loss": 2.075,
+      "step": 3340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.86191165535392e-05,
+      "loss": 2.0916,
+      "step": 3341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8618125405868743e-05,
+      "loss": 2.0569,
+      "step": 3342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8617133920746617e-05,
+      "loss": 2.0823,
+      "step": 3343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.861614209819884e-05,
+      "loss": 2.0818,
+      "step": 3344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8615149938251407e-05,
+      "loss": 2.0405,
+      "step": 3345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8614157440930346e-05,
+      "loss": 2.0693,
+      "step": 3346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8613164606261678e-05,
+      "loss": 2.0831,
+      "step": 3347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8612171434271438e-05,
+      "loss": 2.0815,
+      "step": 3348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.861117792498568e-05,
+      "loss": 2.0276,
+      "step": 3349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8610184078430445e-05,
+      "loss": 2.0476,
+      "step": 3350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8609189894631805e-05,
+      "loss": 2.0757,
+      "step": 3351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8608195373615833e-05,
+      "loss": 2.0674,
+      "step": 3352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8607200515408597e-05,
+      "loss": 2.061,
+      "step": 3353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8606205320036194e-05,
+      "loss": 2.0515,
+      "step": 3354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8605209787524722e-05,
+      "loss": 2.041,
+      "step": 3355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.860421391790029e-05,
+      "loss": 2.0889,
+      "step": 3356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8603217711189012e-05,
+      "loss": 2.0752,
+      "step": 3357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8602221167417003e-05,
+      "loss": 2.0847,
+      "step": 3358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8601224286610405e-05,
+      "loss": 2.072,
+      "step": 3359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8600227068795364e-05,
+      "loss": 2.0686,
+      "step": 3360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8599229513998022e-05,
+      "loss": 2.0443,
+      "step": 3361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8598231622244547e-05,
+      "loss": 2.0686,
+      "step": 3362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.85972333935611e-05,
+      "loss": 2.0637,
+      "step": 3363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.859623482797386e-05,
+      "loss": 2.0588,
+      "step": 3364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8595235925509018e-05,
+      "loss": 2.0579,
+      "step": 3365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.859423668619276e-05,
+      "loss": 2.0637,
+      "step": 3366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.85932371100513e-05,
+      "loss": 2.0184,
+      "step": 3367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.859223719711085e-05,
+      "loss": 2.0452,
+      "step": 3368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8591236947397624e-05,
+      "loss": 2.0002,
+      "step": 3369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8590236360937854e-05,
+      "loss": 2.0295,
+      "step": 3370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8589235437757787e-05,
+      "loss": 2.0918,
+      "step": 3371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.858823417788366e-05,
+      "loss": 2.1084,
+      "step": 3372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8587232581341738e-05,
+      "loss": 2.083,
+      "step": 3373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8586230648158284e-05,
+      "loss": 2.0166,
+      "step": 3374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8585228378359574e-05,
+      "loss": 2.0625,
+      "step": 3375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8584225771971888e-05,
+      "loss": 2.0667,
+      "step": 3376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8583222829021525e-05,
+      "loss": 2.0996,
+      "step": 3377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8582219549534777e-05,
+      "loss": 2.085,
+      "step": 3378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.858121593353796e-05,
+      "loss": 2.0662,
+      "step": 3379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8580211981057393e-05,
+      "loss": 2.0759,
+      "step": 3380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8579207692119395e-05,
+      "loss": 2.0745,
+      "step": 3381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.857820306675031e-05,
+      "loss": 2.062,
+      "step": 3382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8577198104976485e-05,
+      "loss": 2.0072,
+      "step": 3383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.857619280682427e-05,
+      "loss": 2.0559,
+      "step": 3384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.857518717232003e-05,
+      "loss": 1.9954,
+      "step": 3385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.857418120149013e-05,
+      "loss": 2.0249,
+      "step": 3386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.857317489436096e-05,
+      "loss": 2.0427,
+      "step": 3387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8572168250958904e-05,
+      "loss": 2.0515,
+      "step": 3388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8571161271310357e-05,
+      "loss": 2.0332,
+      "step": 3389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8570153955441734e-05,
+      "loss": 2.0862,
+      "step": 3390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8569146303379443e-05,
+      "loss": 2.0935,
+      "step": 3391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8568138315149913e-05,
+      "loss": 2.1138,
+      "step": 3392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.856712999077958e-05,
+      "loss": 2.0786,
+      "step": 3393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8566121330294875e-05,
+      "loss": 2.0916,
+      "step": 3394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.856511233372226e-05,
+      "loss": 2.0132,
+      "step": 3395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8564103001088195e-05,
+      "loss": 2.0417,
+      "step": 3396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.856309333241914e-05,
+      "loss": 1.9802,
+      "step": 3397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.856208332774158e-05,
+      "loss": 1.9741,
+      "step": 3398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8561072987081995e-05,
+      "loss": 2.0508,
+      "step": 3399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8560062310466886e-05,
+      "loss": 2.0664,
+      "step": 3400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.855905129792275e-05,
+      "loss": 2.0969,
+      "step": 3401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8558039949476107e-05,
+      "loss": 2.0801,
+      "step": 3402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8557028265153477e-05,
+      "loss": 2.0389,
+      "step": 3403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8556016244981382e-05,
+      "loss": 2.0479,
+      "step": 3404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8555003888986372e-05,
+      "loss": 2.0947,
+      "step": 3405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.855399119719499e-05,
+      "loss": 2.0652,
+      "step": 3406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8552978169633793e-05,
+      "loss": 1.9744,
+      "step": 3407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8551964806329347e-05,
+      "loss": 1.9792,
+      "step": 3408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8550951107308223e-05,
+      "loss": 2.024,
+      "step": 3409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.854993707259701e-05,
+      "loss": 2.0833,
+      "step": 3410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.854892270222229e-05,
+      "loss": 2.085,
+      "step": 3411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8547907996210676e-05,
+      "loss": 2.1331,
+      "step": 3412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.854689295458877e-05,
+      "loss": 2.0112,
+      "step": 3413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8545877577383198e-05,
+      "loss": 2.0283,
+      "step": 3414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8544861864620572e-05,
+      "loss": 2.0334,
+      "step": 3415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8543845816327545e-05,
+      "loss": 2.0317,
+      "step": 3416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8542829432530754e-05,
+      "loss": 2.0381,
+      "step": 3417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8541812713256847e-05,
+      "loss": 2.0325,
+      "step": 3418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8540795658532493e-05,
+      "loss": 1.9995,
+      "step": 3419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853977826838436e-05,
+      "loss": 2.0144,
+      "step": 3420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853876054283913e-05,
+      "loss": 2.1206,
+      "step": 3421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8537742481923493e-05,
+      "loss": 2.1384,
+      "step": 3422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853672408566414e-05,
+      "loss": 2.054,
+      "step": 3423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853570535408779e-05,
+      "loss": 2.0686,
+      "step": 3424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853468628722115e-05,
+      "loss": 2.073,
+      "step": 3425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853366688509094e-05,
+      "loss": 2.0435,
+      "step": 3426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853264714772389e-05,
+      "loss": 2.0767,
+      "step": 3427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8531627075146758e-05,
+      "loss": 2.0688,
+      "step": 3428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.853060666738628e-05,
+      "loss": 2.0229,
+      "step": 3429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.852958592446922e-05,
+      "loss": 2.0293,
+      "step": 3430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8528564846422346e-05,
+      "loss": 2.072,
+      "step": 3431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8527543433272428e-05,
+      "loss": 2.1018,
+      "step": 3432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.852652168504626e-05,
+      "loss": 2.1091,
+      "step": 3433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8525499601770638e-05,
+      "loss": 2.0989,
+      "step": 3434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8524477183472352e-05,
+      "loss": 2.0833,
+      "step": 3435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8523454430178225e-05,
+      "loss": 2.051,
+      "step": 3436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8522431341915076e-05,
+      "loss": 2.0742,
+      "step": 3437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8521407918709733e-05,
+      "loss": 2.095,
+      "step": 3438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.852038416058903e-05,
+      "loss": 2.1086,
+      "step": 3439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8519360067579816e-05,
+      "loss": 2.1008,
+      "step": 3440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8518335639708952e-05,
+      "loss": 2.0994,
+      "step": 3441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8517310877003297e-05,
+      "loss": 2.126,
+      "step": 3442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.851628577948973e-05,
+      "loss": 2.0178,
+      "step": 3443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8515260347195127e-05,
+      "loss": 2.0645,
+      "step": 3444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.851423458014638e-05,
+      "loss": 2.0549,
+      "step": 3445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8513208478370383e-05,
+      "loss": 2.052,
+      "step": 3446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.851218204189406e-05,
+      "loss": 1.9802,
+      "step": 3447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8511155270744313e-05,
+      "loss": 2.0273,
+      "step": 3448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8510128164948072e-05,
+      "loss": 2.0178,
+      "step": 3449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.850910072453228e-05,
+      "loss": 2.0271,
+      "step": 3450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8508072949523865e-05,
+      "loss": 2.0823,
+      "step": 3451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.850704483994979e-05,
+      "loss": 2.0759,
+      "step": 3452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.850601639583702e-05,
+      "loss": 2.0696,
+      "step": 3453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.850498761721251e-05,
+      "loss": 2.0505,
+      "step": 3454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8503958504103255e-05,
+      "loss": 1.9836,
+      "step": 3455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8502929056536228e-05,
+      "loss": 1.9954,
+      "step": 3456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8501899274538435e-05,
+      "loss": 2.0627,
+      "step": 3457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8500869158136876e-05,
+      "loss": 2.0896,
+      "step": 3458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8499838707358564e-05,
+      "loss": 1.9839,
+      "step": 3459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8498807922230528e-05,
+      "loss": 2.031,
+      "step": 3460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8497776802779787e-05,
+      "loss": 2.0486,
+      "step": 3461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.849674534903339e-05,
+      "loss": 2.0774,
+      "step": 3462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.849571356101839e-05,
+      "loss": 2.0527,
+      "step": 3463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8494681438761835e-05,
+      "loss": 2.022,
+      "step": 3464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8493648982290793e-05,
+      "loss": 2.0488,
+      "step": 3465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8492616191632344e-05,
+      "loss": 2.0654,
+      "step": 3466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8491583066813566e-05,
+      "loss": 2.0806,
+      "step": 3467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8490549607861548e-05,
+      "loss": 2.0908,
+      "step": 3468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8489515814803405e-05,
+      "loss": 2.1008,
+      "step": 3469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8488481687666237e-05,
+      "loss": 2.1353,
+      "step": 3470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8487447226477165e-05,
+      "loss": 2.0205,
+      "step": 3471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8486412431263317e-05,
+      "loss": 2.0281,
+      "step": 3472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8485377302051827e-05,
+      "loss": 2.0564,
+      "step": 3473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8484341838869843e-05,
+      "loss": 2.0374,
+      "step": 3474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8483306041744513e-05,
+      "loss": 2.0632,
+      "step": 3475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.848226991070301e-05,
+      "loss": 2.1023,
+      "step": 3476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.84812334457725e-05,
+      "loss": 2.0703,
+      "step": 3477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8480196646980158e-05,
+      "loss": 2.0332,
+      "step": 3478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8479159514353183e-05,
+      "loss": 2.0203,
+      "step": 3479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.847812204791876e-05,
+      "loss": 2.0273,
+      "step": 3480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8477084247704108e-05,
+      "loss": 2.0703,
+      "step": 3481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8476046113736432e-05,
+      "loss": 2.0635,
+      "step": 3482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8475007646042966e-05,
+      "loss": 2.0896,
+      "step": 3483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8473968844650933e-05,
+      "loss": 2.1245,
+      "step": 3484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.847292970958758e-05,
+      "loss": 2.0833,
+      "step": 3485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8471890240880148e-05,
+      "loss": 2.0776,
+      "step": 3486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.847085043855591e-05,
+      "loss": 2.0388,
+      "step": 3487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8469810302642128e-05,
+      "loss": 2.0786,
+      "step": 3488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.846876983316608e-05,
+      "loss": 2.0925,
+      "step": 3489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.846772903015504e-05,
+      "loss": 2.0891,
+      "step": 3490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8466687893636315e-05,
+      "loss": 2.0651,
+      "step": 3491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8465646423637203e-05,
+      "loss": 2.1023,
+      "step": 3492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8464604620185014e-05,
+      "loss": 2.0784,
+      "step": 3493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8463562483307068e-05,
+      "loss": 2.0588,
+      "step": 3494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8462520013030696e-05,
+      "loss": 2.0918,
+      "step": 3495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8461477209383233e-05,
+      "loss": 2.0977,
+      "step": 3496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8460434072392036e-05,
+      "loss": 2.0266,
+      "step": 3497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.845939060208444e-05,
+      "loss": 2.0623,
+      "step": 3498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8458346798487823e-05,
+      "loss": 2.0314,
+      "step": 3499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.8457302661629556e-05,
+      "loss": 2.0776,
+      "step": 3500
+    },
+    {
+      "Bleu_1": 0.7695709856532097,
+      "Bleu_2": 0.609034090144051,
+      "Bleu_3": 0.4670137230848093,
+      "Bleu_4": 0.3543822976925825,
+      "CIDEr": 1.1942387455770136,
+      "METEOR": 0.2766686961785053,
+      "ROUGE_L": 0.5558572137143918,
+      "epoch": 0.17,
+      "step": 3500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8456258191537018e-05,
+      "loss": 2.0833,
+      "step": 3501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.84552133882376e-05,
+      "loss": 2.094,
+      "step": 3502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.84541682517587e-05,
+      "loss": 2.0461,
+      "step": 3503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8453122782127723e-05,
+      "loss": 2.0774,
+      "step": 3504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8452076979372093e-05,
+      "loss": 2.0177,
+      "step": 3505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.845103084351923e-05,
+      "loss": 2.0632,
+      "step": 3506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8449984374596563e-05,
+      "loss": 2.0327,
+      "step": 3507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8448937572631544e-05,
+      "loss": 2.0432,
+      "step": 3508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.844789043765162e-05,
+      "loss": 2.1279,
+      "step": 3509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.844684296968424e-05,
+      "loss": 2.1243,
+      "step": 3510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8445795168756892e-05,
+      "loss": 2.0232,
+      "step": 3511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8444747034897043e-05,
+      "loss": 2.0454,
+      "step": 3512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.844369856813218e-05,
+      "loss": 2.0752,
+      "step": 3513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.84426497684898e-05,
+      "loss": 2.1045,
+      "step": 3514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8441600635997405e-05,
+      "loss": 2.0859,
+      "step": 3515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.84405511706825e-05,
+      "loss": 2.092,
+      "step": 3516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.843950137257262e-05,
+      "loss": 2.0171,
+      "step": 3517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8438451241695286e-05,
+      "loss": 2.0532,
+      "step": 3518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8437400778078038e-05,
+      "loss": 2.0828,
+      "step": 3519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8436349981748423e-05,
+      "loss": 2.1072,
+      "step": 3520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8435298852734e-05,
+      "loss": 2.0691,
+      "step": 3521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.843424739106233e-05,
+      "loss": 2.0486,
+      "step": 3522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.843319559676098e-05,
+      "loss": 2.0625,
+      "step": 3523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8432143469857547e-05,
+      "loss": 2.0535,
+      "step": 3524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8431091010379615e-05,
+      "loss": 2.0803,
+      "step": 3525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.843003821835478e-05,
+      "loss": 2.1077,
+      "step": 3526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8428985093810655e-05,
+      "loss": 2.0723,
+      "step": 3527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8427931636774854e-05,
+      "loss": 2.0549,
+      "step": 3528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8426877847275e-05,
+      "loss": 1.97,
+      "step": 3529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8425823725338735e-05,
+      "loss": 1.9827,
+      "step": 3530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.84247692709937e-05,
+      "loss": 2.0234,
+      "step": 3531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8423714484267542e-05,
+      "loss": 2.0571,
+      "step": 3532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8422659365187924e-05,
+      "loss": 2.0271,
+      "step": 3533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8421603913782515e-05,
+      "loss": 2.0466,
+      "step": 3534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8420548130079e-05,
+      "loss": 2.0623,
+      "step": 3535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8419492014105053e-05,
+      "loss": 2.0718,
+      "step": 3536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8418435565888378e-05,
+      "loss": 2.071,
+      "step": 3537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8417378785456677e-05,
+      "loss": 2.0967,
+      "step": 3538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8416321672837662e-05,
+      "loss": 2.0359,
+      "step": 3539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.841526422805906e-05,
+      "loss": 2.0386,
+      "step": 3540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8414206451148592e-05,
+      "loss": 2.0588,
+      "step": 3541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8413148342134004e-05,
+      "loss": 2.0806,
+      "step": 3542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8412089901043043e-05,
+      "loss": 2.0245,
+      "step": 3543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.841103112790346e-05,
+      "loss": 2.0361,
+      "step": 3544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8409972022743025e-05,
+      "loss": 2.0813,
+      "step": 3545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8408912585589514e-05,
+      "loss": 2.1086,
+      "step": 3546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8407852816470702e-05,
+      "loss": 2.116,
+      "step": 3547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8406792715414387e-05,
+      "loss": 2.1206,
+      "step": 3548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8405732282448363e-05,
+      "loss": 2.0132,
+      "step": 3549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8404671517600445e-05,
+      "loss": 2.0344,
+      "step": 3550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8403610420898448e-05,
+      "loss": 2.0623,
+      "step": 3551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8402548992370193e-05,
+      "loss": 2.0579,
+      "step": 3552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8401487232043524e-05,
+      "loss": 2.0747,
+      "step": 3553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8400425139946277e-05,
+      "loss": 2.0706,
+      "step": 3554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8399362716106306e-05,
+      "loss": 2.0732,
+      "step": 3555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8398299960551476e-05,
+      "loss": 2.0818,
+      "step": 3556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8397236873309652e-05,
+      "loss": 2.0781,
+      "step": 3557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8396173454408714e-05,
+      "loss": 2.0435,
+      "step": 3558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8395109703876542e-05,
+      "loss": 2.1055,
+      "step": 3559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8394045621741044e-05,
+      "loss": 2.0994,
+      "step": 3560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8392981208030115e-05,
+      "loss": 2.0627,
+      "step": 3561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.839191646277167e-05,
+      "loss": 2.0562,
+      "step": 3562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8390851385993638e-05,
+      "loss": 2.0852,
+      "step": 3563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8389785977723937e-05,
+      "loss": 2.0857,
+      "step": 3564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8388720237990515e-05,
+      "loss": 2.0305,
+      "step": 3565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8387654166821316e-05,
+      "loss": 2.0188,
+      "step": 3566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.83865877642443e-05,
+      "loss": 2.0623,
+      "step": 3567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8385521030287427e-05,
+      "loss": 2.0632,
+      "step": 3568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8384453964978674e-05,
+      "loss": 2.0586,
+      "step": 3569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8383386568346024e-05,
+      "loss": 2.0811,
+      "step": 3570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8382318840417468e-05,
+      "loss": 2.0376,
+      "step": 3571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8381250781221003e-05,
+      "loss": 2.0881,
+      "step": 3572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.838018239078464e-05,
+      "loss": 2.0291,
+      "step": 3573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.83791136691364e-05,
+      "loss": 2.0483,
+      "step": 3574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8378044616304305e-05,
+      "loss": 2.1038,
+      "step": 3575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8376975232316385e-05,
+      "loss": 2.1118,
+      "step": 3576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.837590551720069e-05,
+      "loss": 2.073,
+      "step": 3577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8374835470985275e-05,
+      "loss": 2.0886,
+      "step": 3578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.837376509369819e-05,
+      "loss": 2.0515,
+      "step": 3579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8372694385367513e-05,
+      "loss": 2.0325,
+      "step": 3580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.837162334602132e-05,
+      "loss": 2.1091,
+      "step": 3581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8370551975687696e-05,
+      "loss": 2.1326,
+      "step": 3582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.836948027439474e-05,
+      "loss": 2.0898,
+      "step": 3583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8368408242170548e-05,
+      "loss": 2.1152,
+      "step": 3584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8367335879043245e-05,
+      "loss": 2.1021,
+      "step": 3585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8366263185040946e-05,
+      "loss": 2.0999,
+      "step": 3586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.836519016019178e-05,
+      "loss": 2.0493,
+      "step": 3587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8364116804523883e-05,
+      "loss": 2.0298,
+      "step": 3588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8363043118065412e-05,
+      "loss": 1.9636,
+      "step": 3589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8361969100844513e-05,
+      "loss": 1.9773,
+      "step": 3590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.836089475288936e-05,
+      "loss": 2.1108,
+      "step": 3591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835982007422812e-05,
+      "loss": 2.1079,
+      "step": 3592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835874506488898e-05,
+      "loss": 2.0718,
+      "step": 3593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8357669724900127e-05,
+      "loss": 2.0613,
+      "step": 3594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835659405428976e-05,
+      "loss": 2.0725,
+      "step": 3595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835551805308609e-05,
+      "loss": 2.0891,
+      "step": 3596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8354441721317333e-05,
+      "loss": 2.0969,
+      "step": 3597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835336505901171e-05,
+      "loss": 2.1062,
+      "step": 3598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8352288066197468e-05,
+      "loss": 2.1003,
+      "step": 3599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8351210742902838e-05,
+      "loss": 2.1484,
+      "step": 3600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.835013308915607e-05,
+      "loss": 2.0105,
+      "step": 3601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8349055104985437e-05,
+      "loss": 2.0247,
+      "step": 3602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8347976790419198e-05,
+      "loss": 2.054,
+      "step": 3603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8346898145485633e-05,
+      "loss": 2.0391,
+      "step": 3604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8345819170213027e-05,
+      "loss": 2.0488,
+      "step": 3605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8344739864629674e-05,
+      "loss": 2.0676,
+      "step": 3606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8343660228763885e-05,
+      "loss": 2.1116,
+      "step": 3607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.834258026264396e-05,
+      "loss": 2.0857,
+      "step": 3608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.834149996629823e-05,
+      "loss": 2.0703,
+      "step": 3609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8340419339755018e-05,
+      "loss": 2.0349,
+      "step": 3610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.833933838304267e-05,
+      "loss": 2.0803,
+      "step": 3611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.833825709618952e-05,
+      "loss": 2.0859,
+      "step": 3612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8337175479223942e-05,
+      "loss": 2.0542,
+      "step": 3613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8336093532174282e-05,
+      "loss": 2.0222,
+      "step": 3614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8335011255068924e-05,
+      "loss": 2.082,
+      "step": 3615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8333928647936243e-05,
+      "loss": 2.0723,
+      "step": 3616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.833284571080464e-05,
+      "loss": 1.9695,
+      "step": 3617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.833176244370249e-05,
+      "loss": 1.9897,
+      "step": 3618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.833067884665823e-05,
+      "loss": 2.0657,
+      "step": 3619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8329594919700257e-05,
+      "loss": 2.0649,
+      "step": 3620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8328510662857006e-05,
+      "loss": 1.9692,
+      "step": 3621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8327426076156902e-05,
+      "loss": 1.9946,
+      "step": 3622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8326341159628386e-05,
+      "loss": 2.0337,
+      "step": 3623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8325255913299918e-05,
+      "loss": 2.0562,
+      "step": 3624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.832417033719995e-05,
+      "loss": 2.1052,
+      "step": 3625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8323084431356952e-05,
+      "loss": 2.1238,
+      "step": 3626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8321998195799403e-05,
+      "loss": 2.0518,
+      "step": 3627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8320911630555787e-05,
+      "loss": 2.1074,
+      "step": 3628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8319824735654595e-05,
+      "loss": 2.0583,
+      "step": 3629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.831873751112433e-05,
+      "loss": 2.0271,
+      "step": 3630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8317649956993513e-05,
+      "loss": 1.9883,
+      "step": 3631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.831656207329065e-05,
+      "loss": 2.0076,
+      "step": 3632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8315473860044278e-05,
+      "loss": 2.0842,
+      "step": 3633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8314385317282926e-05,
+      "loss": 2.1162,
+      "step": 3634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8313296445035147e-05,
+      "loss": 2.0769,
+      "step": 3635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8312207243329495e-05,
+      "loss": 2.0601,
+      "step": 3636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8311117712194532e-05,
+      "loss": 2.1172,
+      "step": 3637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8310027851658827e-05,
+      "loss": 2.1038,
+      "step": 3638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8308937661750965e-05,
+      "loss": 2.0825,
+      "step": 3639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8307847142499533e-05,
+      "loss": 2.0955,
+      "step": 3640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8306756293933126e-05,
+      "loss": 2.0713,
+      "step": 3641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8305665116080356e-05,
+      "loss": 2.0632,
+      "step": 3642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.830457360896983e-05,
+      "loss": 2.0522,
+      "step": 3643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.830348177263018e-05,
+      "loss": 2.0427,
+      "step": 3644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.830238960709003e-05,
+      "loss": 2.0776,
+      "step": 3645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8301297112378028e-05,
+      "loss": 2.0813,
+      "step": 3646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.830020428852282e-05,
+      "loss": 2.0103,
+      "step": 3647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8299111135553063e-05,
+      "loss": 2.0056,
+      "step": 3648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8298017653497427e-05,
+      "loss": 2.0254,
+      "step": 3649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8296923842384583e-05,
+      "loss": 2.0129,
+      "step": 3650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8295829702243213e-05,
+      "loss": 2.1172,
+      "step": 3651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8294735233102018e-05,
+      "loss": 2.0701,
+      "step": 3652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8293640434989697e-05,
+      "loss": 2.0464,
+      "step": 3653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8292545307934952e-05,
+      "loss": 2.0469,
+      "step": 3654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8291449851966513e-05,
+      "loss": 2.0801,
+      "step": 3655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8290354067113097e-05,
+      "loss": 2.0764,
+      "step": 3656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8289257953403446e-05,
+      "loss": 1.9722,
+      "step": 3657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8288161510866304e-05,
+      "loss": 1.9746,
+      "step": 3658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.828706473953042e-05,
+      "loss": 2.105,
+      "step": 3659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8285967639424558e-05,
+      "loss": 2.0867,
+      "step": 3660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8284870210577485e-05,
+      "loss": 2.0767,
+      "step": 3661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8283772453017983e-05,
+      "loss": 2.0708,
+      "step": 3662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.828267436677484e-05,
+      "loss": 2.0159,
+      "step": 3663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.828157595187685e-05,
+      "loss": 2.0725,
+      "step": 3664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8280477208352824e-05,
+      "loss": 2.0371,
+      "step": 3665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8279378136231564e-05,
+      "loss": 2.0767,
+      "step": 3666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.82782787355419e-05,
+      "loss": 2.0898,
+      "step": 3667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8277179006312662e-05,
+      "loss": 2.0896,
+      "step": 3668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8276078948572687e-05,
+      "loss": 2.0786,
+      "step": 3669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.827497856235082e-05,
+      "loss": 2.0786,
+      "step": 3670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8273877847675924e-05,
+      "loss": 2.072,
+      "step": 3671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.827277680457686e-05,
+      "loss": 2.0664,
+      "step": 3672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8271675433082502e-05,
+      "loss": 2.042,
+      "step": 3673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.827057373322173e-05,
+      "loss": 2.0178,
+      "step": 3674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.826947170502344e-05,
+      "loss": 2.0764,
+      "step": 3675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8268369348516528e-05,
+      "loss": 2.1238,
+      "step": 3676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.82672666637299e-05,
+      "loss": 2.062,
+      "step": 3677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.826616365069248e-05,
+      "loss": 2.0801,
+      "step": 3678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8265060309433186e-05,
+      "loss": 2.0554,
+      "step": 3679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8263956639980956e-05,
+      "loss": 1.9924,
+      "step": 3680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8262852642364727e-05,
+      "loss": 2.0847,
+      "step": 3681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.826174831661346e-05,
+      "loss": 2.0828,
+      "step": 3682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.82606436627561e-05,
+      "loss": 2.0757,
+      "step": 3683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8259538680821627e-05,
+      "loss": 2.0557,
+      "step": 3684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.825843337083902e-05,
+      "loss": 2.0361,
+      "step": 3685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8257327732837255e-05,
+      "loss": 2.021,
+      "step": 3686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8256221766845327e-05,
+      "loss": 2.0444,
+      "step": 3687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8255115472892244e-05,
+      "loss": 2.0518,
+      "step": 3688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8254008851007015e-05,
+      "loss": 2.0923,
+      "step": 3689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8252901901218663e-05,
+      "loss": 2.1064,
+      "step": 3690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8251794623556213e-05,
+      "loss": 2.0256,
+      "step": 3691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8250687018048698e-05,
+      "loss": 2.0254,
+      "step": 3692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8249579084725172e-05,
+      "loss": 1.9651,
+      "step": 3693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8248470823614682e-05,
+      "loss": 2.0366,
+      "step": 3694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.82473622347463e-05,
+      "loss": 2.0569,
+      "step": 3695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8246253318149084e-05,
+      "loss": 2.0571,
+      "step": 3696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.824514407385213e-05,
+      "loss": 2.0977,
+      "step": 3697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8244034501884513e-05,
+      "loss": 2.0991,
+      "step": 3698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8242924602275336e-05,
+      "loss": 2.1104,
+      "step": 3699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.8241814375053707e-05,
+      "loss": 2.0918,
+      "step": 3700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8240703820248736e-05,
+      "loss": 2.0559,
+      "step": 3701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8239592937889548e-05,
+      "loss": 2.0654,
+      "step": 3702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8238481728005274e-05,
+      "loss": 2.0825,
+      "step": 3703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8237370190625056e-05,
+      "loss": 2.1086,
+      "step": 3704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8236258325778044e-05,
+      "loss": 2.0593,
+      "step": 3705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8235146133493388e-05,
+      "loss": 2.0459,
+      "step": 3706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.823403361380026e-05,
+      "loss": 2.092,
+      "step": 3707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8232920766727837e-05,
+      "loss": 2.103,
+      "step": 3708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8231807592305297e-05,
+      "loss": 2.066,
+      "step": 3709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.823069409056183e-05,
+      "loss": 2.0566,
+      "step": 3710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8229580261526644e-05,
+      "loss": 2.0388,
+      "step": 3711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822846610522894e-05,
+      "loss": 2.053,
+      "step": 3712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822735162169794e-05,
+      "loss": 2.0408,
+      "step": 3713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822623681096287e-05,
+      "loss": 2.0193,
+      "step": 3714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822512167305296e-05,
+      "loss": 2.1028,
+      "step": 3715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822400620799746e-05,
+      "loss": 2.1008,
+      "step": 3716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8222890415825614e-05,
+      "loss": 2.0833,
+      "step": 3717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8221774296566693e-05,
+      "loss": 2.0635,
+      "step": 3718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.822065785024996e-05,
+      "loss": 2.1096,
+      "step": 3719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8219541076904685e-05,
+      "loss": 2.0957,
+      "step": 3720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8218423976560165e-05,
+      "loss": 2.0828,
+      "step": 3721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8217306549245692e-05,
+      "loss": 2.0769,
+      "step": 3722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8216188794990572e-05,
+      "loss": 2.0415,
+      "step": 3723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8215070713824106e-05,
+      "loss": 2.0298,
+      "step": 3724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8213952305775624e-05,
+      "loss": 2.1028,
+      "step": 3725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8212833570874452e-05,
+      "loss": 2.1296,
+      "step": 3726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.821171450914993e-05,
+      "loss": 2.0793,
+      "step": 3727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.82105951206314e-05,
+      "loss": 2.0493,
+      "step": 3728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.820947540534822e-05,
+      "loss": 2.0365,
+      "step": 3729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.820835536332975e-05,
+      "loss": 2.0671,
+      "step": 3730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8207234994605368e-05,
+      "loss": 2.0037,
+      "step": 3731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8206114299204447e-05,
+      "loss": 1.97,
+      "step": 3732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8204993277156376e-05,
+      "loss": 2.0874,
+      "step": 3733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.820387192849056e-05,
+      "loss": 2.0964,
+      "step": 3734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8202750253236397e-05,
+      "loss": 2.0354,
+      "step": 3735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.820162825142331e-05,
+      "loss": 2.0769,
+      "step": 3736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8200505923080716e-05,
+      "loss": 2.0667,
+      "step": 3737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8199383268238045e-05,
+      "loss": 2.0461,
+      "step": 3738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8198260286924738e-05,
+      "loss": 2.0907,
+      "step": 3739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8197136979170253e-05,
+      "loss": 2.0754,
+      "step": 3740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8196013345004042e-05,
+      "loss": 2.0408,
+      "step": 3741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8194889384455566e-05,
+      "loss": 2.0457,
+      "step": 3742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81937650975543e-05,
+      "loss": 2.0405,
+      "step": 3743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.819264048432974e-05,
+      "loss": 2.0303,
+      "step": 3744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.819151554481136e-05,
+      "loss": 2.0486,
+      "step": 3745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.819039027902867e-05,
+      "loss": 2.0876,
+      "step": 3746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.818926468701118e-05,
+      "loss": 2.1067,
+      "step": 3747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81881387687884e-05,
+      "loss": 2.0908,
+      "step": 3748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8187012524389867e-05,
+      "loss": 2.0737,
+      "step": 3749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81858859538451e-05,
+      "loss": 2.0767,
+      "step": 3750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8184759057183658e-05,
+      "loss": 2.0044,
+      "step": 3751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8183631834435088e-05,
+      "loss": 2.0083,
+      "step": 3752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.818250428562894e-05,
+      "loss": 2.0205,
+      "step": 3753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8181376410794796e-05,
+      "loss": 2.0107,
+      "step": 3754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8180248209962218e-05,
+      "loss": 2.0684,
+      "step": 3755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.817911968316081e-05,
+      "loss": 2.0613,
+      "step": 3756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8177990830420157e-05,
+      "loss": 2.0095,
+      "step": 3757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.817686165176986e-05,
+      "loss": 2.0198,
+      "step": 3758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8175732147239533e-05,
+      "loss": 2.075,
+      "step": 3759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8174602316858798e-05,
+      "loss": 2.0649,
+      "step": 3760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.817347216065728e-05,
+      "loss": 2.0581,
+      "step": 3761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8172341678664617e-05,
+      "loss": 2.0618,
+      "step": 3762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8171210870910453e-05,
+      "loss": 2.1116,
+      "step": 3763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8170079737424443e-05,
+      "loss": 2.0967,
+      "step": 3764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.816894827823625e-05,
+      "loss": 2.0764,
+      "step": 3765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8167816493375552e-05,
+      "loss": 2.1096,
+      "step": 3766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.816668438287202e-05,
+      "loss": 2.0662,
+      "step": 3767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8165551946755345e-05,
+      "loss": 2.0396,
+      "step": 3768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.816441918505522e-05,
+      "loss": 2.0745,
+      "step": 3769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.816328609780136e-05,
+      "loss": 2.0913,
+      "step": 3770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8162152685023468e-05,
+      "loss": 2.0559,
+      "step": 3771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8161018946751273e-05,
+      "loss": 2.0635,
+      "step": 3772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8159884883014508e-05,
+      "loss": 2.1167,
+      "step": 3773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81587504938429e-05,
+      "loss": 2.0984,
+      "step": 3774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8157615779266216e-05,
+      "loss": 2.0719,
+      "step": 3775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8156480739314194e-05,
+      "loss": 2.1018,
+      "step": 3776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.815534537401661e-05,
+      "loss": 2.1213,
+      "step": 3777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8154209683403238e-05,
+      "loss": 2.1128,
+      "step": 3778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8153073667503856e-05,
+      "loss": 2.0913,
+      "step": 3779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8151937326348257e-05,
+      "loss": 2.0332,
+      "step": 3780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8150800659966234e-05,
+      "loss": 2.0601,
+      "step": 3781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8149663668387605e-05,
+      "loss": 2.0237,
+      "step": 3782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.814852635164218e-05,
+      "loss": 2.0544,
+      "step": 3783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8147388709759784e-05,
+      "loss": 2.0859,
+      "step": 3784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.814625074277025e-05,
+      "loss": 2.1094,
+      "step": 3785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8145112450703428e-05,
+      "loss": 2.0874,
+      "step": 3786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.814397383358916e-05,
+      "loss": 2.0737,
+      "step": 3787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8142834891457302e-05,
+      "loss": 2.0789,
+      "step": 3788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.814169562433773e-05,
+      "loss": 1.9756,
+      "step": 3789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8140556032260313e-05,
+      "loss": 1.9553,
+      "step": 3790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.813941611525494e-05,
+      "loss": 2.0613,
+      "step": 3791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81382758733515e-05,
+      "loss": 2.0703,
+      "step": 3792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8137135306579904e-05,
+      "loss": 2.0681,
+      "step": 3793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8135994414970047e-05,
+      "loss": 2.054,
+      "step": 3794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8134853198551864e-05,
+      "loss": 2.0957,
+      "step": 3795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8133711657355265e-05,
+      "loss": 2.0718,
+      "step": 3796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8132569791410202e-05,
+      "loss": 2.0884,
+      "step": 3797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8131427600746605e-05,
+      "loss": 2.0503,
+      "step": 3798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.813028508539444e-05,
+      "loss": 2.0872,
+      "step": 3799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8129142245383652e-05,
+      "loss": 2.0686,
+      "step": 3800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8127999080744225e-05,
+      "loss": 2.0994,
+      "step": 3801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.812685559150613e-05,
+      "loss": 2.0889,
+      "step": 3802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.812571177769936e-05,
+      "loss": 2.0427,
+      "step": 3803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81245676393539e-05,
+      "loss": 2.0623,
+      "step": 3804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8123423176499767e-05,
+      "loss": 2.0554,
+      "step": 3805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.812227838916696e-05,
+      "loss": 2.0935,
+      "step": 3806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.812113327738551e-05,
+      "loss": 2.0083,
+      "step": 3807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8119987841185438e-05,
+      "loss": 2.0566,
+      "step": 3808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8118842080596788e-05,
+      "loss": 2.1167,
+      "step": 3809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.81176959956496e-05,
+      "loss": 2.0969,
+      "step": 3810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8116549586373935e-05,
+      "loss": 2.0264,
+      "step": 3811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8115402852799853e-05,
+      "loss": 2.0281,
+      "step": 3812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8114255794957426e-05,
+      "loss": 1.9973,
+      "step": 3813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8113108412876736e-05,
+      "loss": 2.0146,
+      "step": 3814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8111960706587873e-05,
+      "loss": 2.062,
+      "step": 3815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8110812676120928e-05,
+      "loss": 2.0383,
+      "step": 3816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8109664321506007e-05,
+      "loss": 2.0769,
+      "step": 3817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8108515642773234e-05,
+      "loss": 2.0759,
+      "step": 3818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.810736663995272e-05,
+      "loss": 2.043,
+      "step": 3819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8106217313074607e-05,
+      "loss": 2.0371,
+      "step": 3820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.810506766216903e-05,
+      "loss": 2.0542,
+      "step": 3821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8103917687266135e-05,
+      "loss": 2.0676,
+      "step": 3822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8102767388396076e-05,
+      "loss": 2.116,
+      "step": 3823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.810161676558903e-05,
+      "loss": 2.0908,
+      "step": 3824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8100465818875164e-05,
+      "loss": 2.0608,
+      "step": 3825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8099314548284656e-05,
+      "loss": 2.0723,
+      "step": 3826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.80981629538477e-05,
+      "loss": 2.0708,
+      "step": 3827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8097011035594495e-05,
+      "loss": 2.085,
+      "step": 3828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8095858793555253e-05,
+      "loss": 2.0435,
+      "step": 3829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8094706227760184e-05,
+      "loss": 2.0464,
+      "step": 3830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.809355333823952e-05,
+      "loss": 2.095,
+      "step": 3831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8092400125023485e-05,
+      "loss": 2.0876,
+      "step": 3832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8091246588142327e-05,
+      "loss": 2.0547,
+      "step": 3833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8090092727626295e-05,
+      "loss": 2.0823,
+      "step": 3834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8088938543505645e-05,
+      "loss": 1.9911,
+      "step": 3835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8087784035810647e-05,
+      "loss": 2.0054,
+      "step": 3836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8086629204571578e-05,
+      "loss": 2.0684,
+      "step": 3837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8085474049818717e-05,
+      "loss": 2.0613,
+      "step": 3838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.808431857158236e-05,
+      "loss": 2.0886,
+      "step": 3839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8083162769892808e-05,
+      "loss": 2.0874,
+      "step": 3840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.808200664478037e-05,
+      "loss": 2.0811,
+      "step": 3841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8080850196275366e-05,
+      "loss": 2.0776,
+      "step": 3842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8079693424408124e-05,
+      "loss": 2.0845,
+      "step": 3843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.807853632920897e-05,
+      "loss": 2.0549,
+      "step": 3844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8077378910708256e-05,
+      "loss": 2.0459,
+      "step": 3845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8076221168936332e-05,
+      "loss": 2.0603,
+      "step": 3846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8075063103923554e-05,
+      "loss": 2.1294,
+      "step": 3847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8073904715700296e-05,
+      "loss": 2.1025,
+      "step": 3848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8072746004296936e-05,
+      "loss": 1.9893,
+      "step": 3849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8071586969743856e-05,
+      "loss": 2.0381,
+      "step": 3850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.807042761207145e-05,
+      "loss": 2.0513,
+      "step": 3851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.806926793131013e-05,
+      "loss": 2.0615,
+      "step": 3852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8068107927490292e-05,
+      "loss": 1.9127,
+      "step": 3853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8066947600642368e-05,
+      "loss": 1.9861,
+      "step": 3854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.806578695079678e-05,
+      "loss": 2.0691,
+      "step": 3855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8064625977983974e-05,
+      "loss": 2.0796,
+      "step": 3856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8063464682234378e-05,
+      "loss": 2.0542,
+      "step": 3857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8062303063578468e-05,
+      "loss": 2.0349,
+      "step": 3858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8061141122046684e-05,
+      "loss": 2.0381,
+      "step": 3859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8059978857669508e-05,
+      "loss": 2.0461,
+      "step": 3860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8058816270477417e-05,
+      "loss": 2.1069,
+      "step": 3861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8057653360500904e-05,
+      "loss": 2.1123,
+      "step": 3862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8056490127770455e-05,
+      "loss": 2.0847,
+      "step": 3863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8055326572316578e-05,
+      "loss": 2.1042,
+      "step": 3864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.805416269416979e-05,
+      "loss": 2.0781,
+      "step": 3865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8052998493360613e-05,
+      "loss": 2.082,
+      "step": 3866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8051833969919567e-05,
+      "loss": 2.0825,
+      "step": 3867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.80506691238772e-05,
+      "loss": 2.0681,
+      "step": 3868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8049503955264054e-05,
+      "loss": 2.0293,
+      "step": 3869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8048338464110688e-05,
+      "loss": 2.0239,
+      "step": 3870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.804717265044766e-05,
+      "loss": 2.063,
+      "step": 3871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.804600651430555e-05,
+      "loss": 2.075,
+      "step": 3872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.804484005571493e-05,
+      "loss": 2.0657,
+      "step": 3873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8043673274706395e-05,
+      "loss": 2.0916,
+      "step": 3874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8042506171310534e-05,
+      "loss": 2.0911,
+      "step": 3875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.804133874555797e-05,
+      "loss": 2.0134,
+      "step": 3876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8040170997479298e-05,
+      "loss": 2.0511,
+      "step": 3877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8039002927105156e-05,
+      "loss": 2.0576,
+      "step": 3878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8037834534466166e-05,
+      "loss": 2.0378,
+      "step": 3879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.803666581959297e-05,
+      "loss": 2.0618,
+      "step": 3880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8035496782516218e-05,
+      "loss": 2.0613,
+      "step": 3881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8034327423266562e-05,
+      "loss": 2.0425,
+      "step": 3882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8033157741874678e-05,
+      "loss": 2.0315,
+      "step": 3883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8031987738371224e-05,
+      "loss": 2.0183,
+      "step": 3884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8030817412786896e-05,
+      "loss": 2.0635,
+      "step": 3885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.802964676515238e-05,
+      "loss": 2.106,
+      "step": 3886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8028475795498364e-05,
+      "loss": 2.1033,
+      "step": 3887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8027304503855573e-05,
+      "loss": 2.1104,
+      "step": 3888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.802613289025471e-05,
+      "loss": 2.0295,
+      "step": 3889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8024960954726502e-05,
+      "loss": 2.0593,
+      "step": 3890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.802378869730169e-05,
+      "loss": 2.0933,
+      "step": 3891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8022616118011004e-05,
+      "loss": 2.1057,
+      "step": 3892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.80214432168852e-05,
+      "loss": 2.0959,
+      "step": 3893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8020269993955032e-05,
+      "loss": 2.1084,
+      "step": 3894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8019096449251272e-05,
+      "loss": 2.0669,
+      "step": 3895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8017922582804684e-05,
+      "loss": 2.0723,
+      "step": 3896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8016748394646064e-05,
+      "loss": 2.1113,
+      "step": 3897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8015573884806195e-05,
+      "loss": 2.0852,
+      "step": 3898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.8014399053315884e-05,
+      "loss": 2.0508,
+      "step": 3899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8013223900205936e-05,
+      "loss": 2.0764,
+      "step": 3900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8012048425507164e-05,
+      "loss": 2.0425,
+      "step": 3901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.80108726292504e-05,
+      "loss": 2.0317,
+      "step": 3902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8009696511466476e-05,
+      "loss": 2.0642,
+      "step": 3903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8008520072186234e-05,
+      "loss": 2.0798,
+      "step": 3904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8007343311440523e-05,
+      "loss": 2.0432,
+      "step": 3905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8006166229260204e-05,
+      "loss": 2.0564,
+      "step": 3906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.800498882567614e-05,
+      "loss": 2.1118,
+      "step": 3907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8003811100719215e-05,
+      "loss": 2.0859,
+      "step": 3908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8002633054420314e-05,
+      "loss": 2.0718,
+      "step": 3909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.8001454686810316e-05,
+      "loss": 2.052,
+      "step": 3910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.800027599792014e-05,
+      "loss": 2.0984,
+      "step": 3911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7999096987780685e-05,
+      "loss": 2.0796,
+      "step": 3912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.799791765642287e-05,
+      "loss": 2.0621,
+      "step": 3913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7996738003877625e-05,
+      "loss": 2.0857,
+      "step": 3914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7995558030175883e-05,
+      "loss": 2.093,
+      "step": 3915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7994377735348588e-05,
+      "loss": 2.1091,
+      "step": 3916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7993197119426696e-05,
+      "loss": 2.0276,
+      "step": 3917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7992016182441155e-05,
+      "loss": 2.041,
+      "step": 3918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.799083492442295e-05,
+      "loss": 2.092,
+      "step": 3919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7989653345403043e-05,
+      "loss": 2.0837,
+      "step": 3920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.798847144541243e-05,
+      "loss": 2.05,
+      "step": 3921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.79872892244821e-05,
+      "loss": 2.0588,
+      "step": 3922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.798610668264306e-05,
+      "loss": 2.0742,
+      "step": 3923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7984923819926314e-05,
+      "loss": 2.0671,
+      "step": 3924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7983740636362884e-05,
+      "loss": 2.0327,
+      "step": 3925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.79825571319838e-05,
+      "loss": 2.0659,
+      "step": 3926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.79813733068201e-05,
+      "loss": 2.0779,
+      "step": 3927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.798018916090282e-05,
+      "loss": 2.0662,
+      "step": 3928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.797900469426302e-05,
+      "loss": 2.0754,
+      "step": 3929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7977819906931757e-05,
+      "loss": 2.1133,
+      "step": 3930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7976634798940103e-05,
+      "loss": 2.1013,
+      "step": 3931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7975449370319132e-05,
+      "loss": 2.075,
+      "step": 3932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.797426362109994e-05,
+      "loss": 2.0129,
+      "step": 3933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7973077551313615e-05,
+      "loss": 2.0105,
+      "step": 3934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7971891160991257e-05,
+      "loss": 2.1003,
+      "step": 3935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.797070445016399e-05,
+      "loss": 2.1282,
+      "step": 3936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.796951741886292e-05,
+      "loss": 2.0281,
+      "step": 3937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.796833006711919e-05,
+      "loss": 2.0544,
+      "step": 3938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7967142394963918e-05,
+      "loss": 2.0117,
+      "step": 3939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7965954402428267e-05,
+      "loss": 2.0491,
+      "step": 3940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.796476608954338e-05,
+      "loss": 2.0466,
+      "step": 3941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7963577456340425e-05,
+      "loss": 2.0442,
+      "step": 3942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7962388502850568e-05,
+      "loss": 2.0256,
+      "step": 3943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7961199229104994e-05,
+      "loss": 2.0068,
+      "step": 3944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7960009635134884e-05,
+      "loss": 2.0901,
+      "step": 3945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7958819720971435e-05,
+      "loss": 2.0635,
+      "step": 3946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.795762948664585e-05,
+      "loss": 2.0273,
+      "step": 3947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7956438932189346e-05,
+      "loss": 2.0342,
+      "step": 3948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7955248057633146e-05,
+      "loss": 2.0579,
+      "step": 3949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7954056863008473e-05,
+      "loss": 2.0803,
+      "step": 3950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7952865348346564e-05,
+      "loss": 2.1306,
+      "step": 3951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7951673513678674e-05,
+      "loss": 2.0859,
+      "step": 3952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7950481359036047e-05,
+      "loss": 2.0796,
+      "step": 3953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.794928888444995e-05,
+      "loss": 2.0681,
+      "step": 3954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.794809608995166e-05,
+      "loss": 2.0571,
+      "step": 3955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7946902975572447e-05,
+      "loss": 2.0369,
+      "step": 3956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7945709541343602e-05,
+      "loss": 2.0339,
+      "step": 3957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7944515787296424e-05,
+      "loss": 2.0679,
+      "step": 3958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.794332171346222e-05,
+      "loss": 2.0505,
+      "step": 3959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7942127319872294e-05,
+      "loss": 2.0471,
+      "step": 3960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7940932606557976e-05,
+      "loss": 2.0911,
+      "step": 3961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7939737573550594e-05,
+      "loss": 2.1082,
+      "step": 3962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.793854222088148e-05,
+      "loss": 2.1331,
+      "step": 3963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7937346548581993e-05,
+      "loss": 2.103,
+      "step": 3964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7936150556683482e-05,
+      "loss": 2.0525,
+      "step": 3965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7934954245217306e-05,
+      "loss": 2.0645,
+      "step": 3966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7933757614214838e-05,
+      "loss": 2.0281,
+      "step": 3967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.793256066370747e-05,
+      "loss": 2.0608,
+      "step": 3968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7931363393726573e-05,
+      "loss": 2.0371,
+      "step": 3969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7930165804303564e-05,
+      "loss": 2.0457,
+      "step": 3970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7928967895469824e-05,
+      "loss": 2.0906,
+      "step": 3971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7927769667256786e-05,
+      "loss": 2.0884,
+      "step": 3972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7926571119695864e-05,
+      "loss": 2.0474,
+      "step": 3973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7925372252818493e-05,
+      "loss": 2.1279,
+      "step": 3974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.792417306665611e-05,
+      "loss": 2.0957,
+      "step": 3975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.792297356124016e-05,
+      "loss": 2.0952,
+      "step": 3976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7921773736602105e-05,
+      "loss": 2.1292,
+      "step": 3977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.79205735927734e-05,
+      "loss": 2.1357,
+      "step": 3978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.791937312978552e-05,
+      "loss": 1.968,
+      "step": 3979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7918172347669952e-05,
+      "loss": 2.0378,
+      "step": 3980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.791697124645818e-05,
+      "loss": 2.0067,
+      "step": 3981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.79157698261817e-05,
+      "loss": 2.0686,
+      "step": 3982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7914568086872022e-05,
+      "loss": 2.1121,
+      "step": 3983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7913366028560656e-05,
+      "loss": 2.0901,
+      "step": 3984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7912163651279126e-05,
+      "loss": 2.0168,
+      "step": 3985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7910960955058964e-05,
+      "loss": 2.0464,
+      "step": 3986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.790975793993171e-05,
+      "loss": 2.0627,
+      "step": 3987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7908554605928912e-05,
+      "loss": 2.0564,
+      "step": 3988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7907350953082126e-05,
+      "loss": 2.02,
+      "step": 3989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.790614698142291e-05,
+      "loss": 2.0093,
+      "step": 3990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7904942690982847e-05,
+      "loss": 2.0425,
+      "step": 3991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7903738081793512e-05,
+      "loss": 2.0476,
+      "step": 3992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7902533153886496e-05,
+      "loss": 2.0366,
+      "step": 3993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7901327907293392e-05,
+      "loss": 2.0422,
+      "step": 3994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7900122342045813e-05,
+      "loss": 2.1345,
+      "step": 3995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7898916458175373e-05,
+      "loss": 2.1089,
+      "step": 3996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7897710255713693e-05,
+      "loss": 2.0366,
+      "step": 3997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7896503734692405e-05,
+      "loss": 2.0347,
+      "step": 3998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7895296895143144e-05,
+      "loss": 2.0762,
+      "step": 3999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7894089737097565e-05,
+      "loss": 2.1096,
+      "step": 4000
+    },
+    {
+      "Bleu_1": 0.7706250182795276,
+      "Bleu_2": 0.6108194311288114,
+      "Bleu_3": 0.46926675212931823,
+      "Bleu_4": 0.3567697946684881,
+      "CIDEr": 1.2018765864979695,
+      "METEOR": 0.2777585193752336,
+      "ROUGE_L": 0.5572625922126528,
+      "epoch": 0.2,
+      "step": 4000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7892882260587317e-05,
+      "loss": 2.0681,
+      "step": 4001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7891674465644076e-05,
+      "loss": 2.074,
+      "step": 4002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.78904663522995e-05,
+      "loss": 2.0549,
+      "step": 4003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7889257920585283e-05,
+      "loss": 2.0745,
+      "step": 4004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7888049170533105e-05,
+      "loss": 2.0415,
+      "step": 4005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7886840102174672e-05,
+      "loss": 2.0867,
+      "step": 4006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7885630715541692e-05,
+      "loss": 2.0789,
+      "step": 4007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.788442101066587e-05,
+      "loss": 2.0764,
+      "step": 4008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7883210987578934e-05,
+      "loss": 2.1021,
+      "step": 4009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7882000646312622e-05,
+      "loss": 2.1094,
+      "step": 4010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7880789986898662e-05,
+      "loss": 2.1284,
+      "step": 4011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7879579009368807e-05,
+      "loss": 2.1116,
+      "step": 4012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7878367713754817e-05,
+      "loss": 2.1211,
+      "step": 4013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.787715610008845e-05,
+      "loss": 2.104,
+      "step": 4014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.787594416840149e-05,
+      "loss": 2.0969,
+      "step": 4015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7874731918725708e-05,
+      "loss": 2.0984,
+      "step": 4016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7873519351092892e-05,
+      "loss": 2.0251,
+      "step": 4017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7872306465534855e-05,
+      "loss": 2.0344,
+      "step": 4018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.787109326208339e-05,
+      "loss": 2.0444,
+      "step": 4019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7869879740770314e-05,
+      "loss": 2.0183,
+      "step": 4020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7868665901627454e-05,
+      "loss": 2.0532,
+      "step": 4021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.786745174468664e-05,
+      "loss": 2.0911,
+      "step": 4022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7866237269979713e-05,
+      "loss": 2.0828,
+      "step": 4023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.786502247753852e-05,
+      "loss": 2.0938,
+      "step": 4024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7863807367394915e-05,
+      "loss": 2.01,
+      "step": 4025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.786259193958077e-05,
+      "loss": 1.9622,
+      "step": 4026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7861376194127946e-05,
+      "loss": 2.0815,
+      "step": 4027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7860160131068335e-05,
+      "loss": 2.0635,
+      "step": 4028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7858943750433826e-05,
+      "loss": 2.113,
+      "step": 4029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.785772705225631e-05,
+      "loss": 2.1128,
+      "step": 4030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7856510036567704e-05,
+      "loss": 2.0618,
+      "step": 4031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7855292703399916e-05,
+      "loss": 2.0647,
+      "step": 4032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7854075052784867e-05,
+      "loss": 2.0654,
+      "step": 4033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7852857084754494e-05,
+      "loss": 2.0532,
+      "step": 4034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7851638799340734e-05,
+      "loss": 2.0255,
+      "step": 4035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.785042019657554e-05,
+      "loss": 2.0535,
+      "step": 4036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.784920127649086e-05,
+      "loss": 2.0508,
+      "step": 4037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7847982039118664e-05,
+      "loss": 2.0422,
+      "step": 4038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.784676248449092e-05,
+      "loss": 2.0647,
+      "step": 4039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.784554261263962e-05,
+      "loss": 2.0642,
+      "step": 4040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.784432242359674e-05,
+      "loss": 1.9706,
+      "step": 4041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.784310191739429e-05,
+      "loss": 2.0283,
+      "step": 4042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7841881094064272e-05,
+      "loss": 2.0557,
+      "step": 4043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.78406599536387e-05,
+      "loss": 2.0391,
+      "step": 4044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7839438496149597e-05,
+      "loss": 1.9944,
+      "step": 4045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7838216721628995e-05,
+      "loss": 1.9974,
+      "step": 4046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7836994630108926e-05,
+      "loss": 2.03,
+      "step": 4047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7835772221621453e-05,
+      "loss": 2.0327,
+      "step": 4048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7834549496198626e-05,
+      "loss": 2.0562,
+      "step": 4049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.78333264538725e-05,
+      "loss": 2.0652,
+      "step": 4050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7832103094675162e-05,
+      "loss": 2.0925,
+      "step": 4051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7830879418638686e-05,
+      "loss": 2.1028,
+      "step": 4052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7829655425795164e-05,
+      "loss": 2.0542,
+      "step": 4053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7828431116176692e-05,
+      "loss": 2.0371,
+      "step": 4054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7827206489815372e-05,
+      "loss": 2.0674,
+      "step": 4055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7825981546743324e-05,
+      "loss": 2.0627,
+      "step": 4056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7824756286992673e-05,
+      "loss": 2.0581,
+      "step": 4057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.782353071059554e-05,
+      "loss": 2.0513,
+      "step": 4058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.782230481758408e-05,
+      "loss": 2.0339,
+      "step": 4059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7821078607990424e-05,
+      "loss": 2.0583,
+      "step": 4060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7819852081846733e-05,
+      "loss": 2.0208,
+      "step": 4061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.781862523918518e-05,
+      "loss": 2.0425,
+      "step": 4062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7817398080037932e-05,
+      "loss": 2.0459,
+      "step": 4063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7816170604437166e-05,
+      "loss": 2.032,
+      "step": 4064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7814942812415075e-05,
+      "loss": 2.073,
+      "step": 4065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7813714704003854e-05,
+      "loss": 2.092,
+      "step": 4066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7812486279235707e-05,
+      "loss": 2.0972,
+      "step": 4067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7811257538142855e-05,
+      "loss": 2.0901,
+      "step": 4068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.781002848075752e-05,
+      "loss": 2.0757,
+      "step": 4069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7808799107111925e-05,
+      "loss": 2.0676,
+      "step": 4070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7807569417238313e-05,
+      "loss": 2.0837,
+      "step": 4071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7806339411168934e-05,
+      "loss": 2.0881,
+      "step": 4072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7805109088936037e-05,
+      "loss": 2.0386,
+      "step": 4073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.780387845057189e-05,
+      "loss": 2.0344,
+      "step": 4074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7802647496108765e-05,
+      "loss": 2.0491,
+      "step": 4075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7801416225578943e-05,
+      "loss": 2.0813,
+      "step": 4076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.780018463901471e-05,
+      "loss": 2.0276,
+      "step": 4077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7798952736448366e-05,
+      "loss": 2.0327,
+      "step": 4078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7797720517912215e-05,
+      "loss": 2.0608,
+      "step": 4079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.779648798343857e-05,
+      "loss": 2.0554,
+      "step": 4080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.779525513305975e-05,
+      "loss": 2.0254,
+      "step": 4081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.779402196680809e-05,
+      "loss": 2.0188,
+      "step": 4082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7792788484715925e-05,
+      "loss": 2.062,
+      "step": 4083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7791554686815606e-05,
+      "loss": 2.082,
+      "step": 4084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7790320573139482e-05,
+      "loss": 2.0679,
+      "step": 4085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7789086143719922e-05,
+      "loss": 2.0764,
+      "step": 4086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7787851398589295e-05,
+      "loss": 2.0532,
+      "step": 4087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7786616337779978e-05,
+      "loss": 2.0359,
+      "step": 4088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.778538096132436e-05,
+      "loss": 2.1116,
+      "step": 4089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7784145269254837e-05,
+      "loss": 2.1072,
+      "step": 4090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.778290926160382e-05,
+      "loss": 2.136,
+      "step": 4091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7781672938403715e-05,
+      "loss": 2.1111,
+      "step": 4092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.778043629968695e-05,
+      "loss": 1.9973,
+      "step": 4093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7779199345485942e-05,
+      "loss": 2.0129,
+      "step": 4094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7777962075833142e-05,
+      "loss": 2.0535,
+      "step": 4095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7776724490760986e-05,
+      "loss": 2.0374,
+      "step": 4096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7775486590301937e-05,
+      "loss": 2.0085,
+      "step": 4097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.777424837448845e-05,
+      "loss": 2.0505,
+      "step": 4098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7773009843353004e-05,
+      "loss": 2.1194,
+      "step": 4099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.7771770996928067e-05,
+      "loss": 2.1108,
+      "step": 4100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.777053183524613e-05,
+      "loss": 2.0654,
+      "step": 4101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7769292358339697e-05,
+      "loss": 2.05,
+      "step": 4102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7768052566241266e-05,
+      "loss": 2.0898,
+      "step": 4103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7766812458983345e-05,
+      "loss": 2.0686,
+      "step": 4104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7765572036598465e-05,
+      "loss": 2.0483,
+      "step": 4105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7764331299119137e-05,
+      "loss": 2.0764,
+      "step": 4106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7763090246577916e-05,
+      "loss": 2.0684,
+      "step": 4107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7761848879007333e-05,
+      "loss": 2.0476,
+      "step": 4108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7760607196439957e-05,
+      "loss": 2.0886,
+      "step": 4109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7759365198908335e-05,
+      "loss": 2.0667,
+      "step": 4110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7758122886445046e-05,
+      "loss": 2.0325,
+      "step": 4111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7756880259082666e-05,
+      "loss": 2.073,
+      "step": 4112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.775563731685378e-05,
+      "loss": 2.0779,
+      "step": 4113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.775439405979098e-05,
+      "loss": 2.0862,
+      "step": 4114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7753150487926875e-05,
+      "loss": 2.1155,
+      "step": 4115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7751906601294074e-05,
+      "loss": 2.1143,
+      "step": 4116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7750662399925194e-05,
+      "loss": 2.0637,
+      "step": 4117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7749417883852862e-05,
+      "loss": 2.0725,
+      "step": 4118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7748173053109717e-05,
+      "loss": 2.0815,
+      "step": 4119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.774692790772841e-05,
+      "loss": 2.0769,
+      "step": 4120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.774568244774158e-05,
+      "loss": 2.0769,
+      "step": 4121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.77444366731819e-05,
+      "loss": 2.0806,
+      "step": 4122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.774319058408203e-05,
+      "loss": 2.0967,
+      "step": 4123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7741944180474647e-05,
+      "loss": 2.1169,
+      "step": 4124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7740697462392445e-05,
+      "loss": 2.0281,
+      "step": 4125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.773945042986811e-05,
+      "loss": 2.054,
+      "step": 4126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7738203082934347e-05,
+      "loss": 2.0604,
+      "step": 4127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7736955421623866e-05,
+      "loss": 2.0872,
+      "step": 4128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7735707445969384e-05,
+      "loss": 2.095,
+      "step": 4129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7734459156003633e-05,
+      "loss": 2.021,
+      "step": 4130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7733210551759337e-05,
+      "loss": 2.071,
+      "step": 4131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7731961633269253e-05,
+      "loss": 2.0833,
+      "step": 4132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7730712400566124e-05,
+      "loss": 2.1211,
+      "step": 4133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7729462853682714e-05,
+      "loss": 2.0874,
+      "step": 4134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.772821299265178e-05,
+      "loss": 2.0518,
+      "step": 4135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7726962817506114e-05,
+      "loss": 2.0781,
+      "step": 4136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.772571232827849e-05,
+      "loss": 2.1052,
+      "step": 4137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7724461525001702e-05,
+      "loss": 2.0835,
+      "step": 4138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7723210407708555e-05,
+      "loss": 2.0627,
+      "step": 4139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7721958976431848e-05,
+      "loss": 2.064,
+      "step": 4140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7720707231204412e-05,
+      "loss": 2.0422,
+      "step": 4141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7719455172059068e-05,
+      "loss": 2.0544,
+      "step": 4142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7718202799028646e-05,
+      "loss": 2.0859,
+      "step": 4143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7716950112145992e-05,
+      "loss": 2.0972,
+      "step": 4144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.771569711144395e-05,
+      "loss": 2.0972,
+      "step": 4145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7714443796955385e-05,
+      "loss": 2.0916,
+      "step": 4146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7713190168713164e-05,
+      "loss": 2.0979,
+      "step": 4147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.771193622675015e-05,
+      "loss": 2.0894,
+      "step": 4148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7710681971099246e-05,
+      "loss": 2.113,
+      "step": 4149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.770942740179333e-05,
+      "loss": 2.1011,
+      "step": 4150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7708172518865306e-05,
+      "loss": 2.1191,
+      "step": 4151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7706917322348075e-05,
+      "loss": 2.1106,
+      "step": 4152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7705661812274562e-05,
+      "loss": 2.0723,
+      "step": 4153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.770440598867769e-05,
+      "loss": 2.0798,
+      "step": 4154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.770314985159039e-05,
+      "loss": 2.0452,
+      "step": 4155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7701893401045598e-05,
+      "loss": 2.0022,
+      "step": 4156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.770063663707627e-05,
+      "loss": 2.1113,
+      "step": 4157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.769937955971536e-05,
+      "loss": 2.1311,
+      "step": 4158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7698122168995835e-05,
+      "loss": 2.0359,
+      "step": 4159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7696864464950663e-05,
+      "loss": 2.0225,
+      "step": 4160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7695606447612832e-05,
+      "loss": 2.0173,
+      "step": 4161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.769434811701533e-05,
+      "loss": 2.0369,
+      "step": 4162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.769308947319116e-05,
+      "loss": 2.0974,
+      "step": 4163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7691830516173316e-05,
+      "loss": 2.0891,
+      "step": 4164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7690571245994824e-05,
+      "loss": 2.1289,
+      "step": 4165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7689311662688704e-05,
+      "loss": 2.1194,
+      "step": 4166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.768805176628799e-05,
+      "loss": 2.1118,
+      "step": 4167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7686791556825713e-05,
+      "loss": 2.1208,
+      "step": 4168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7685531034334927e-05,
+      "loss": 2.0742,
+      "step": 4169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7684270198848685e-05,
+      "loss": 2.0969,
+      "step": 4170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7683009050400053e-05,
+      "loss": 2.0193,
+      "step": 4171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7681747589022105e-05,
+      "loss": 2.0273,
+      "step": 4172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7680485814747913e-05,
+      "loss": 2.0901,
+      "step": 4173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7679223727610575e-05,
+      "loss": 2.0476,
+      "step": 4174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7677961327643188e-05,
+      "loss": 1.9795,
+      "step": 4175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7676698614878845e-05,
+      "loss": 2.0276,
+      "step": 4176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7675435589350673e-05,
+      "loss": 2.0605,
+      "step": 4177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7674172251091784e-05,
+      "loss": 2.04,
+      "step": 4178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7672908600135314e-05,
+      "loss": 2.1196,
+      "step": 4179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7671644636514392e-05,
+      "loss": 2.1221,
+      "step": 4180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7670380360262178e-05,
+      "loss": 2.0493,
+      "step": 4181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7669115771411814e-05,
+      "loss": 2.0112,
+      "step": 4182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7667850869996464e-05,
+      "loss": 2.1111,
+      "step": 4183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7666585656049304e-05,
+      "loss": 2.1138,
+      "step": 4184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.766532012960351e-05,
+      "loss": 2.0916,
+      "step": 4185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.766405429069227e-05,
+      "loss": 2.0603,
+      "step": 4186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7662788139348774e-05,
+      "loss": 2.032,
+      "step": 4187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7661521675606232e-05,
+      "loss": 2.0713,
+      "step": 4188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7660254899497854e-05,
+      "loss": 2.114,
+      "step": 4189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7658987811056854e-05,
+      "loss": 2.1133,
+      "step": 4190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7657720410316467e-05,
+      "loss": 2.0488,
+      "step": 4191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7656452697309927e-05,
+      "loss": 2.0317,
+      "step": 4192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7655184672070474e-05,
+      "loss": 2.02,
+      "step": 4193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7653916334631374e-05,
+      "loss": 1.9956,
+      "step": 4194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.765264768502587e-05,
+      "loss": 2.0288,
+      "step": 4195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.765137872328724e-05,
+      "loss": 2.0068,
+      "step": 4196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7650109449448763e-05,
+      "loss": 2.0608,
+      "step": 4197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.764883986354372e-05,
+      "loss": 2.0483,
+      "step": 4198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7647569965605398e-05,
+      "loss": 2.0676,
+      "step": 4199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7646299755667115e-05,
+      "loss": 2.0947,
+      "step": 4200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.764502923376217e-05,
+      "loss": 2.0339,
+      "step": 4201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.764375839992388e-05,
+      "loss": 2.0249,
+      "step": 4202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.764248725418558e-05,
+      "loss": 2.0498,
+      "step": 4203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7641215796580592e-05,
+      "loss": 2.0693,
+      "step": 4204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7639944027142273e-05,
+      "loss": 2.071,
+      "step": 4205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.763867194590396e-05,
+      "loss": 2.0466,
+      "step": 4206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7637399552899013e-05,
+      "loss": 2.0519,
+      "step": 4207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.763612684816081e-05,
+      "loss": 2.1169,
+      "step": 4208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.763485383172272e-05,
+      "loss": 2.012,
+      "step": 4209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7633580503618124e-05,
+      "loss": 2.0396,
+      "step": 4210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7632306863880415e-05,
+      "loss": 2.0725,
+      "step": 4211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.763103291254299e-05,
+      "loss": 2.0615,
+      "step": 4212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7629758649639264e-05,
+      "loss": 2.0396,
+      "step": 4213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7628484075202652e-05,
+      "loss": 1.9985,
+      "step": 4214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.762720918926657e-05,
+      "loss": 2.1279,
+      "step": 4215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.762593399186446e-05,
+      "loss": 2.1147,
+      "step": 4216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.762465848302976e-05,
+      "loss": 2.0952,
+      "step": 4217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7623382662795913e-05,
+      "loss": 2.1089,
+      "step": 4218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7622106531196385e-05,
+      "loss": 2.0325,
+      "step": 4219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.762083008826463e-05,
+      "loss": 2.0212,
+      "step": 4220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7619553334034128e-05,
+      "loss": 2.0466,
+      "step": 4221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7618276268538366e-05,
+      "loss": 2.0554,
+      "step": 4222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7616998891810823e-05,
+      "loss": 2.0515,
+      "step": 4223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7615721203885003e-05,
+      "loss": 2.0913,
+      "step": 4224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.761444320479441e-05,
+      "loss": 2.0667,
+      "step": 4225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7613164894572553e-05,
+      "loss": 2.041,
+      "step": 4226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7611886273252965e-05,
+      "loss": 1.9962,
+      "step": 4227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7610607340869167e-05,
+      "loss": 2.022,
+      "step": 4228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7609328097454705e-05,
+      "loss": 1.9775,
+      "step": 4229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.760804854304312e-05,
+      "loss": 1.9719,
+      "step": 4230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7606768677667968e-05,
+      "loss": 2.1218,
+      "step": 4231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7605488501362812e-05,
+      "loss": 2.1165,
+      "step": 4232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7604208014161226e-05,
+      "loss": 2.0686,
+      "step": 4233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7602927216096787e-05,
+      "loss": 2.021,
+      "step": 4234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7601646107203078e-05,
+      "loss": 2.0889,
+      "step": 4235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7600364687513705e-05,
+      "loss": 2.0859,
+      "step": 4236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7599082957062265e-05,
+      "loss": 2.0549,
+      "step": 4237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.759780091588237e-05,
+      "loss": 2.0657,
+      "step": 4238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7596518564007637e-05,
+      "loss": 2.0901,
+      "step": 4239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7595235901471703e-05,
+      "loss": 2.0847,
+      "step": 4240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7593952928308194e-05,
+      "loss": 1.996,
+      "step": 4241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7592669644550764e-05,
+      "loss": 2.03,
+      "step": 4242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7591386050233062e-05,
+      "loss": 2.1106,
+      "step": 4243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7590102145388744e-05,
+      "loss": 2.1018,
+      "step": 4244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7588817930051486e-05,
+      "loss": 2.0437,
+      "step": 4245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.758753340425496e-05,
+      "loss": 2.0415,
+      "step": 4246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7586248568032856e-05,
+      "loss": 2.0874,
+      "step": 4247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7584963421418867e-05,
+      "loss": 2.0671,
+      "step": 4248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7583677964446688e-05,
+      "loss": 2.0662,
+      "step": 4249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.758239219715003e-05,
+      "loss": 2.0886,
+      "step": 4250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.758110611956262e-05,
+      "loss": 2.0457,
+      "step": 4251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7579819731718175e-05,
+      "loss": 2.0293,
+      "step": 4252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7578533033650434e-05,
+      "loss": 1.9536,
+      "step": 4253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7577246025393133e-05,
+      "loss": 2.0242,
+      "step": 4254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7575958706980024e-05,
+      "loss": 2.0745,
+      "step": 4255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7574671078444873e-05,
+      "loss": 2.0557,
+      "step": 4256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.757338313982144e-05,
+      "loss": 2.0538,
+      "step": 4257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7572094891143496e-05,
+      "loss": 2.0833,
+      "step": 4258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.757080633244483e-05,
+      "loss": 2.1396,
+      "step": 4259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.756951746375923e-05,
+      "loss": 2.1028,
+      "step": 4260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7568228285120504e-05,
+      "loss": 2.0962,
+      "step": 4261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7566938796562444e-05,
+      "loss": 2.0789,
+      "step": 4262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7565648998118875e-05,
+      "loss": 1.9875,
+      "step": 4263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7564358889823624e-05,
+      "loss": 2.0232,
+      "step": 4264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.756306847171051e-05,
+      "loss": 2.031,
+      "step": 4265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.756177774381338e-05,
+      "loss": 2.0557,
+      "step": 4266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7560486706166086e-05,
+      "loss": 2.0435,
+      "step": 4267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7559195358802477e-05,
+      "loss": 2.0513,
+      "step": 4268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7557903701756423e-05,
+      "loss": 1.9917,
+      "step": 4269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7556611735061788e-05,
+      "loss": 1.9858,
+      "step": 4270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.755531945875246e-05,
+      "loss": 2.0901,
+      "step": 4271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7554026872862324e-05,
+      "loss": 2.0735,
+      "step": 4272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.755273397742528e-05,
+      "loss": 2.0261,
+      "step": 4273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7551440772475225e-05,
+      "loss": 2.0554,
+      "step": 4274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.755014725804608e-05,
+      "loss": 2.0515,
+      "step": 4275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7548853434171763e-05,
+      "loss": 2.0056,
+      "step": 4276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7547559300886203e-05,
+      "loss": 2.0623,
+      "step": 4277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7546264858223332e-05,
+      "loss": 2.0234,
+      "step": 4278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7544970106217104e-05,
+      "loss": 2.0537,
+      "step": 4279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7543675044901467e-05,
+      "loss": 2.0459,
+      "step": 4280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7542379674310388e-05,
+      "loss": 2.1072,
+      "step": 4281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7541083994477832e-05,
+      "loss": 2.0874,
+      "step": 4282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7539788005437773e-05,
+      "loss": 2.0596,
+      "step": 4283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7538491707224202e-05,
+      "loss": 2.0374,
+      "step": 4284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7537195099871115e-05,
+      "loss": 2.0652,
+      "step": 4285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7535898183412506e-05,
+      "loss": 2.0754,
+      "step": 4286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7534600957882396e-05,
+      "loss": 2.0916,
+      "step": 4287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7533303423314793e-05,
+      "loss": 2.0923,
+      "step": 4288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7532005579743725e-05,
+      "loss": 2.0449,
+      "step": 4289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7530707427203235e-05,
+      "loss": 2.0354,
+      "step": 4290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.752940896572736e-05,
+      "loss": 2.0649,
+      "step": 4291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7528110195350147e-05,
+      "loss": 2.0662,
+      "step": 4292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.752681111610566e-05,
+      "loss": 2.0015,
+      "step": 4293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7525511728027965e-05,
+      "loss": 2.0215,
+      "step": 4294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7524212031151134e-05,
+      "loss": 2.0659,
+      "step": 4295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7522912025509247e-05,
+      "loss": 2.0642,
+      "step": 4296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7521611711136405e-05,
+      "loss": 2.1023,
+      "step": 4297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7520311088066708e-05,
+      "loss": 2.0945,
+      "step": 4298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.751901015633425e-05,
+      "loss": 2.0664,
+      "step": 4299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.7517708915973156e-05,
+      "loss": 2.0469,
+      "step": 4300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7516407367017545e-05,
+      "loss": 2.0544,
+      "step": 4301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7515105509501557e-05,
+      "loss": 2.0347,
+      "step": 4302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7513803343459323e-05,
+      "loss": 2.0891,
+      "step": 4303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7512500868924993e-05,
+      "loss": 2.0789,
+      "step": 4304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7511198085932725e-05,
+      "loss": 1.9871,
+      "step": 4305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.750989499451668e-05,
+      "loss": 2.0139,
+      "step": 4306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.750859159471103e-05,
+      "loss": 2.0999,
+      "step": 4307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7507287886549963e-05,
+      "loss": 2.1155,
+      "step": 4308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7505983870067653e-05,
+      "loss": 2.0061,
+      "step": 4309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7504679545298307e-05,
+      "loss": 2.0208,
+      "step": 4310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7503374912276128e-05,
+      "loss": 2.0093,
+      "step": 4311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.750206997103533e-05,
+      "loss": 1.9756,
+      "step": 4312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7500764721610127e-05,
+      "loss": 2.0471,
+      "step": 4313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.749945916403475e-05,
+      "loss": 2.043,
+      "step": 4314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.749815329834344e-05,
+      "loss": 2.1287,
+      "step": 4315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7496847124570434e-05,
+      "loss": 2.1191,
+      "step": 4316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.749554064275e-05,
+      "loss": 2.0916,
+      "step": 4317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7494233852916383e-05,
+      "loss": 2.0898,
+      "step": 4318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7492926755103855e-05,
+      "loss": 1.9958,
+      "step": 4319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.74916193493467e-05,
+      "loss": 2.0166,
+      "step": 4320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7490311635679196e-05,
+      "loss": 2.0652,
+      "step": 4321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.748900361413564e-05,
+      "loss": 2.052,
+      "step": 4322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7487695284750336e-05,
+      "loss": 2.1548,
+      "step": 4323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.748638664755759e-05,
+      "loss": 2.1616,
+      "step": 4324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7485077702591715e-05,
+      "loss": 2.019,
+      "step": 4325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7483768449887044e-05,
+      "loss": 2.022,
+      "step": 4326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.748245888947791e-05,
+      "loss": 2.0566,
+      "step": 4327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.748114902139865e-05,
+      "loss": 2.0781,
+      "step": 4328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.747983884568362e-05,
+      "loss": 2.0535,
+      "step": 4329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.747852836236717e-05,
+      "loss": 2.0422,
+      "step": 4330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7477217571483668e-05,
+      "loss": 2.0304,
+      "step": 4331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7475906473067493e-05,
+      "loss": 2.0825,
+      "step": 4332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7474595067153025e-05,
+      "loss": 2.0898,
+      "step": 4333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7473283353774654e-05,
+      "loss": 2.0859,
+      "step": 4334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7471971332966776e-05,
+      "loss": 2.0161,
+      "step": 4335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7470659004763797e-05,
+      "loss": 2.0112,
+      "step": 4336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7469346369200134e-05,
+      "loss": 2.0532,
+      "step": 4337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7468033426310214e-05,
+      "loss": 2.0652,
+      "step": 4338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7466720176128457e-05,
+      "loss": 2.0598,
+      "step": 4339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7465406618689303e-05,
+      "loss": 2.0706,
+      "step": 4340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7464092754027207e-05,
+      "loss": 2.0513,
+      "step": 4341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.746277858217661e-05,
+      "loss": 2.0632,
+      "step": 4342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7461464103171994e-05,
+      "loss": 2.1143,
+      "step": 4343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7460149317047812e-05,
+      "loss": 2.1052,
+      "step": 4344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7458834223838548e-05,
+      "loss": 2.0632,
+      "step": 4345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7457518823578695e-05,
+      "loss": 2.0513,
+      "step": 4346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7456203116302738e-05,
+      "loss": 2.0955,
+      "step": 4347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.745488710204519e-05,
+      "loss": 2.0886,
+      "step": 4348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7453570780840553e-05,
+      "loss": 2.0835,
+      "step": 4349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.745225415272335e-05,
+      "loss": 2.0911,
+      "step": 4350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7450937217728108e-05,
+      "loss": 2.1006,
+      "step": 4351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7449619975889363e-05,
+      "loss": 2.0979,
+      "step": 4352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7448302427241655e-05,
+      "loss": 2.1038,
+      "step": 4353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.744698457181954e-05,
+      "loss": 2.1038,
+      "step": 4354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7445666409657572e-05,
+      "loss": 2.0786,
+      "step": 4355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7444347940790322e-05,
+      "loss": 2.0815,
+      "step": 4356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.744302916525236e-05,
+      "loss": 2.0198,
+      "step": 4357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7441710083078278e-05,
+      "loss": 2.0129,
+      "step": 4358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7440390694302666e-05,
+      "loss": 2.0938,
+      "step": 4359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7439070998960115e-05,
+      "loss": 2.0737,
+      "step": 4360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.743775099708524e-05,
+      "loss": 2.0225,
+      "step": 4361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.743643068871265e-05,
+      "loss": 2.0183,
+      "step": 4362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7435110073876975e-05,
+      "loss": 2.1016,
+      "step": 4363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7433789152612848e-05,
+      "loss": 2.0984,
+      "step": 4364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.74324679249549e-05,
+      "loss": 2.094,
+      "step": 4365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7431146390937786e-05,
+      "loss": 2.1018,
+      "step": 4366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7429824550596158e-05,
+      "loss": 2.0667,
+      "step": 4367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7428502403964682e-05,
+      "loss": 2.0803,
+      "step": 4368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.742717995107803e-05,
+      "loss": 2.1091,
+      "step": 4369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7425857191970883e-05,
+      "loss": 2.0891,
+      "step": 4370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7424534126677922e-05,
+      "loss": 2.0676,
+      "step": 4371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.742321075523385e-05,
+      "loss": 2.0886,
+      "step": 4372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7421887077673365e-05,
+      "loss": 2.0962,
+      "step": 4373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7420563094031183e-05,
+      "loss": 2.0835,
+      "step": 4374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7419238804342027e-05,
+      "loss": 2.0789,
+      "step": 4375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7417914208640616e-05,
+      "loss": 2.0942,
+      "step": 4376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7416589306961692e-05,
+      "loss": 2.0205,
+      "step": 4377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7415264099339997e-05,
+      "loss": 2.0193,
+      "step": 4378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7413938585810286e-05,
+      "loss": 2.0745,
+      "step": 4379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7412612766407322e-05,
+      "loss": 2.0764,
+      "step": 4380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.741128664116586e-05,
+      "loss": 2.0364,
+      "step": 4381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.740996021012069e-05,
+      "loss": 2.0156,
+      "step": 4382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7408633473306582e-05,
+      "loss": 1.9812,
+      "step": 4383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7407306430758345e-05,
+      "loss": 2.0071,
+      "step": 4384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7405979082510764e-05,
+      "loss": 2.0581,
+      "step": 4385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.740465142859866e-05,
+      "loss": 2.0806,
+      "step": 4386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7403323469056843e-05,
+      "loss": 2.032,
+      "step": 4387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7401995203920135e-05,
+      "loss": 2.0303,
+      "step": 4388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.740066663322337e-05,
+      "loss": 2.0959,
+      "step": 4389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7399337757001387e-05,
+      "loss": 2.0696,
+      "step": 4390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739800857528904e-05,
+      "loss": 2.0234,
+      "step": 4391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739667908812118e-05,
+      "loss": 2.0476,
+      "step": 4392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739534929553267e-05,
+      "loss": 2.0876,
+      "step": 4393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7394019197558388e-05,
+      "loss": 2.0391,
+      "step": 4394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739268879423321e-05,
+      "loss": 2.0562,
+      "step": 4395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739135808559202e-05,
+      "loss": 2.051,
+      "step": 4396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.739002707166973e-05,
+      "loss": 2.1108,
+      "step": 4397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7388695752501225e-05,
+      "loss": 2.0825,
+      "step": 4398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7387364128121427e-05,
+      "loss": 2.0522,
+      "step": 4399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7386032198565257e-05,
+      "loss": 2.0608,
+      "step": 4400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7384699963867643e-05,
+      "loss": 2.051,
+      "step": 4401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7383367424063517e-05,
+      "loss": 2.0654,
+      "step": 4402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.738203457918783e-05,
+      "loss": 2.0103,
+      "step": 4403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7380701429275528e-05,
+      "loss": 1.9983,
+      "step": 4404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7379367974361573e-05,
+      "loss": 2.0439,
+      "step": 4405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7378034214480937e-05,
+      "loss": 2.0383,
+      "step": 4406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7376700149668592e-05,
+      "loss": 2.0974,
+      "step": 4407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7375365779959525e-05,
+      "loss": 2.1125,
+      "step": 4408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7374031105388728e-05,
+      "loss": 2.1191,
+      "step": 4409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7372696125991196e-05,
+      "loss": 2.0959,
+      "step": 4410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7371360841801947e-05,
+      "loss": 2.0276,
+      "step": 4411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.737002525285599e-05,
+      "loss": 2.0222,
+      "step": 4412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.736868935918835e-05,
+      "loss": 2.0657,
+      "step": 4413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.736735316083406e-05,
+      "loss": 2.0581,
+      "step": 4414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7366016657828164e-05,
+      "loss": 2.0229,
+      "step": 4415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7364679850205707e-05,
+      "loss": 2.0457,
+      "step": 4416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7363342738001746e-05,
+      "loss": 2.0693,
+      "step": 4417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7362005321251337e-05,
+      "loss": 2.0864,
+      "step": 4418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7360667599989568e-05,
+      "loss": 2.0576,
+      "step": 4419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.735932957425151e-05,
+      "loss": 2.0564,
+      "step": 4420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.735799124407225e-05,
+      "loss": 2.0161,
+      "step": 4421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7356652609486884e-05,
+      "loss": 2.0352,
+      "step": 4422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7355313670530523e-05,
+      "loss": 2.0579,
+      "step": 4423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.735397442723827e-05,
+      "loss": 2.0764,
+      "step": 4424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7352634879645253e-05,
+      "loss": 2.0791,
+      "step": 4425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7351295027786594e-05,
+      "loss": 2.1001,
+      "step": 4426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.734995487169743e-05,
+      "loss": 2.0894,
+      "step": 4427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7348614411412914e-05,
+      "loss": 2.0872,
+      "step": 4428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7347273646968184e-05,
+      "loss": 2.084,
+      "step": 4429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7345932578398413e-05,
+      "loss": 2.0767,
+      "step": 4430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7344591205738754e-05,
+      "loss": 2.0449,
+      "step": 4431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7343249529024395e-05,
+      "loss": 2.042,
+      "step": 4432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.734190754829052e-05,
+      "loss": 1.9858,
+      "step": 4433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7340565263572318e-05,
+      "loss": 2.0051,
+      "step": 4434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7339222674904984e-05,
+      "loss": 2.0894,
+      "step": 4435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7337879782323734e-05,
+      "loss": 2.0955,
+      "step": 4436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7336536585863775e-05,
+      "loss": 2.1108,
+      "step": 4437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7335193085560343e-05,
+      "loss": 2.105,
+      "step": 4438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7333849281448655e-05,
+      "loss": 2.0513,
+      "step": 4439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7332505173563963e-05,
+      "loss": 2.0808,
+      "step": 4440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7331160761941506e-05,
+      "loss": 2.1025,
+      "step": 4441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.732981604661655e-05,
+      "loss": 2.0852,
+      "step": 4442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7328471027624348e-05,
+      "loss": 2.1216,
+      "step": 4443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.732712570500018e-05,
+      "loss": 2.0728,
+      "step": 4444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.732578007877931e-05,
+      "loss": 2.063,
+      "step": 4445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7324434148997044e-05,
+      "loss": 2.0852,
+      "step": 4446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.732308791568867e-05,
+      "loss": 2.1162,
+      "step": 4447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7321741378889494e-05,
+      "loss": 2.1108,
+      "step": 4448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7320394538634818e-05,
+      "loss": 2.0476,
+      "step": 4449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7319047394959973e-05,
+      "loss": 2.0625,
+      "step": 4450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7317699947900282e-05,
+      "loss": 2.0662,
+      "step": 4451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.731635219749108e-05,
+      "loss": 2.0569,
+      "step": 4452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7315004143767705e-05,
+      "loss": 2.0889,
+      "step": 4453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.731365578676552e-05,
+      "loss": 2.0933,
+      "step": 4454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7312307126519873e-05,
+      "loss": 2.0061,
+      "step": 4455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7310958163066133e-05,
+      "loss": 2.0173,
+      "step": 4456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.730960889643968e-05,
+      "loss": 2.0513,
+      "step": 4457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7308259326675897e-05,
+      "loss": 2.0459,
+      "step": 4458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.730690945381017e-05,
+      "loss": 2.0522,
+      "step": 4459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7305559277877896e-05,
+      "loss": 2.075,
+      "step": 4460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.730420879891449e-05,
+      "loss": 2.0449,
+      "step": 4461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7302858016955364e-05,
+      "loss": 2.0093,
+      "step": 4462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7301506932035935e-05,
+      "loss": 2.037,
+      "step": 4463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7300155544191638e-05,
+      "loss": 2.0405,
+      "step": 4464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.729880385345791e-05,
+      "loss": 2.1116,
+      "step": 4465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7297451859870206e-05,
+      "loss": 2.0886,
+      "step": 4466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7296099563463967e-05,
+      "loss": 2.0845,
+      "step": 4467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7294746964274664e-05,
+      "loss": 2.0764,
+      "step": 4468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7293394062337765e-05,
+      "loss": 2.0671,
+      "step": 4469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7292040857688745e-05,
+      "loss": 2.0613,
+      "step": 4470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.72906873503631e-05,
+      "loss": 2.104,
+      "step": 4471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.728933354039631e-05,
+      "loss": 2.0964,
+      "step": 4472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.728797942782389e-05,
+      "loss": 2.0784,
+      "step": 4473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7286625012681343e-05,
+      "loss": 2.0994,
+      "step": 4474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7285270295004193e-05,
+      "loss": 2.0681,
+      "step": 4475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7283915274827957e-05,
+      "loss": 2.0349,
+      "step": 4476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7282559952188174e-05,
+      "loss": 2.0916,
+      "step": 4477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.728120432712039e-05,
+      "loss": 2.0786,
+      "step": 4478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7279848399660147e-05,
+      "loss": 2.0195,
+      "step": 4479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7278492169843002e-05,
+      "loss": 2.0266,
+      "step": 4480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7277135637704525e-05,
+      "loss": 2.0764,
+      "step": 4481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7275778803280294e-05,
+      "loss": 2.0698,
+      "step": 4482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.727442166660588e-05,
+      "loss": 2.0334,
+      "step": 4483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.727306422771688e-05,
+      "loss": 2.0442,
+      "step": 4484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.727170648664889e-05,
+      "loss": 2.0349,
+      "step": 4485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7270348443437513e-05,
+      "loss": 2.073,
+      "step": 4486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.726899009811836e-05,
+      "loss": 2.0872,
+      "step": 4487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.726763145072706e-05,
+      "loss": 2.0586,
+      "step": 4488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7266272501299235e-05,
+      "loss": 2.0466,
+      "step": 4489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.726491324987052e-05,
+      "loss": 2.0813,
+      "step": 4490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7263553696476564e-05,
+      "loss": 2.116,
+      "step": 4491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7262193841153026e-05,
+      "loss": 2.1133,
+      "step": 4492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7260833683935556e-05,
+      "loss": 2.0144,
+      "step": 4493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7259473224859826e-05,
+      "loss": 2.0444,
+      "step": 4494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7258112463961513e-05,
+      "loss": 2.0378,
+      "step": 4495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7256751401276305e-05,
+      "loss": 2.032,
+      "step": 4496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.725539003683988e-05,
+      "loss": 2.0999,
+      "step": 4497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7254028370687957e-05,
+      "loss": 2.1155,
+      "step": 4498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7252666402856233e-05,
+      "loss": 2.0559,
+      "step": 4499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.725130413338043e-05,
+      "loss": 2.0691,
+      "step": 4500
+    },
+    {
+      "Bleu_1": 0.7729591668885013,
+      "Bleu_2": 0.6121898235413081,
+      "Bleu_3": 0.4701767388562149,
+      "Bleu_4": 0.35707768875379986,
+      "CIDEr": 1.2021732288629379,
+      "METEOR": 0.2778071772146534,
+      "ROUGE_L": 0.5581219470352456,
+      "epoch": 0.23,
+      "step": 4500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.724994156229627e-05,
+      "loss": 2.0754,
+      "step": 4501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7248578689639486e-05,
+      "loss": 2.0969,
+      "step": 4502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7247215515445806e-05,
+      "loss": 2.0647,
+      "step": 4503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7245852039750993e-05,
+      "loss": 2.052,
+      "step": 4504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7244488262590795e-05,
+      "loss": 2.0562,
+      "step": 4505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.724312418400098e-05,
+      "loss": 2.0356,
+      "step": 4506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.724175980401732e-05,
+      "loss": 2.0801,
+      "step": 4507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7240395122675582e-05,
+      "loss": 2.0005,
+      "step": 4508
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7239030140011568e-05,
+      "loss": 2.054,
+      "step": 4509
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7237664856061065e-05,
+      "loss": 2.0542,
+      "step": 4510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.723629927085988e-05,
+      "loss": 1.983,
+      "step": 4511
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7234933384443822e-05,
+      "loss": 2.041,
+      "step": 4512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7233567196848705e-05,
+      "loss": 2.0339,
+      "step": 4513
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7232200708110365e-05,
+      "loss": 2.0168,
+      "step": 4514
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7230833918264627e-05,
+      "loss": 2.0647,
+      "step": 4515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7229466827347344e-05,
+      "loss": 2.0559,
+      "step": 4516
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7228099435394357e-05,
+      "loss": 2.0947,
+      "step": 4517
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7226731742441532e-05,
+      "loss": 2.1028,
+      "step": 4518
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7225363748524734e-05,
+      "loss": 2.0406,
+      "step": 4519
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7223995453679827e-05,
+      "loss": 2.0996,
+      "step": 4520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7222626857942708e-05,
+      "loss": 2.0239,
+      "step": 4521
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7221257961349255e-05,
+      "loss": 2.0076,
+      "step": 4522
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7219888763935367e-05,
+      "loss": 2.1058,
+      "step": 4523
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7218519265736955e-05,
+      "loss": 2.105,
+      "step": 4524
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.721714946678993e-05,
+      "loss": 2.0664,
+      "step": 4525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7215779367130212e-05,
+      "loss": 2.0879,
+      "step": 4526
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7214408966793736e-05,
+      "loss": 2.1292,
+      "step": 4527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7213038265816434e-05,
+      "loss": 2.1375,
+      "step": 4528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7211667264234248e-05,
+      "loss": 2.0652,
+      "step": 4529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7210295962083138e-05,
+      "loss": 2.084,
+      "step": 4530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7208924359399063e-05,
+      "loss": 2.0798,
+      "step": 4531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.720755245621799e-05,
+      "loss": 2.0657,
+      "step": 4532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.720618025257589e-05,
+      "loss": 2.0239,
+      "step": 4533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7204807748508758e-05,
+      "loss": 2.0105,
+      "step": 4534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7203434944052583e-05,
+      "loss": 2.0659,
+      "step": 4535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7202061839243363e-05,
+      "loss": 2.0615,
+      "step": 4536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7200688434117104e-05,
+      "loss": 2.0625,
+      "step": 4537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.719931472870983e-05,
+      "loss": 2.0366,
+      "step": 4538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7197940723057555e-05,
+      "loss": 2.095,
+      "step": 4539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7196566417196317e-05,
+      "loss": 2.1042,
+      "step": 4540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7195191811162154e-05,
+      "loss": 2.0708,
+      "step": 4541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7193816904991117e-05,
+      "loss": 2.053,
+      "step": 4542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.719244169871925e-05,
+      "loss": 2.0786,
+      "step": 4543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7191066192382627e-05,
+      "loss": 2.0757,
+      "step": 4544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.718969038601732e-05,
+      "loss": 2.0295,
+      "step": 4545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.71883142796594e-05,
+      "loss": 2.0625,
+      "step": 4546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7186937873344962e-05,
+      "loss": 2.0525,
+      "step": 4547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7185561167110093e-05,
+      "loss": 2.053,
+      "step": 4548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7184184160990903e-05,
+      "loss": 2.094,
+      "step": 4549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7182806855023493e-05,
+      "loss": 2.0911,
+      "step": 4550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7181429249243988e-05,
+      "loss": 2.0964,
+      "step": 4551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.718005134368851e-05,
+      "loss": 2.0872,
+      "step": 4552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7178673138393197e-05,
+      "loss": 2.0574,
+      "step": 4553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7177294633394194e-05,
+      "loss": 2.0198,
+      "step": 4554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7175915828727643e-05,
+      "loss": 2.1055,
+      "step": 4555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7174536724429706e-05,
+      "loss": 2.0923,
+      "step": 4556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7173157320536547e-05,
+      "loss": 2.0581,
+      "step": 4557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7171777617084336e-05,
+      "loss": 2.0879,
+      "step": 4558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7170397614109263e-05,
+      "loss": 2.03,
+      "step": 4559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7169017311647508e-05,
+      "loss": 2.0134,
+      "step": 4560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7167636709735275e-05,
+      "loss": 2.0725,
+      "step": 4561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.716625580840876e-05,
+      "loss": 2.0864,
+      "step": 4562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7164874607704185e-05,
+      "loss": 2.0447,
+      "step": 4563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.716349310765776e-05,
+      "loss": 2.0515,
+      "step": 4564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7162111308305728e-05,
+      "loss": 2.0039,
+      "step": 4565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.716072920968431e-05,
+      "loss": 2.02,
+      "step": 4566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.715934681182976e-05,
+      "loss": 2.0474,
+      "step": 4567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.715796411477832e-05,
+      "loss": 2.0388,
+      "step": 4568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.715658111856626e-05,
+      "loss": 2.0588,
+      "step": 4569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7155197823229843e-05,
+      "loss": 2.0127,
+      "step": 4570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.715381422880534e-05,
+      "loss": 2.1052,
+      "step": 4571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7152430335329048e-05,
+      "loss": 2.1133,
+      "step": 4572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.715104614283724e-05,
+      "loss": 2.0867,
+      "step": 4573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7149661651366224e-05,
+      "loss": 2.0959,
+      "step": 4574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.714827686095231e-05,
+      "loss": 2.0537,
+      "step": 4575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7146891771631807e-05,
+      "loss": 2.0667,
+      "step": 4576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.714550638344104e-05,
+      "loss": 2.1091,
+      "step": 4577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7144120696416338e-05,
+      "loss": 2.0564,
+      "step": 4578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.714273471059404e-05,
+      "loss": 1.9875,
+      "step": 4579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7141348426010486e-05,
+      "loss": 1.9795,
+      "step": 4580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7139961842702042e-05,
+      "loss": 2.04,
+      "step": 4581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7138574960705057e-05,
+      "loss": 2.0308,
+      "step": 4582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7137187780055907e-05,
+      "loss": 1.9995,
+      "step": 4583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7135800300790966e-05,
+      "loss": 2.0063,
+      "step": 4584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7134412522946627e-05,
+      "loss": 2.0906,
+      "step": 4585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7133024446559275e-05,
+      "loss": 2.0928,
+      "step": 4586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7131636071665308e-05,
+      "loss": 2.0093,
+      "step": 4587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.713024739830114e-05,
+      "loss": 2.0276,
+      "step": 4588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7128858426503192e-05,
+      "loss": 2.0964,
+      "step": 4589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7127469156307878e-05,
+      "loss": 2.0962,
+      "step": 4590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7126079587751632e-05,
+      "loss": 2.115,
+      "step": 4591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.71246897208709e-05,
+      "loss": 2.1079,
+      "step": 4592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7123299555702127e-05,
+      "loss": 2.0442,
+      "step": 4593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7121909092281766e-05,
+      "loss": 2.0259,
+      "step": 4594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7120518330646282e-05,
+      "loss": 2.0378,
+      "step": 4595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7119127270832144e-05,
+      "loss": 2.0413,
+      "step": 4596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7117735912875836e-05,
+      "loss": 2.0906,
+      "step": 4597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7116344256813837e-05,
+      "loss": 2.0566,
+      "step": 4598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.711495230268265e-05,
+      "loss": 2.0272,
+      "step": 4599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7113560050518774e-05,
+      "loss": 2.0398,
+      "step": 4600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7112167500358714e-05,
+      "loss": 2.0664,
+      "step": 4601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7110774652239e-05,
+      "loss": 2.054,
+      "step": 4602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7109381506196145e-05,
+      "loss": 2.0125,
+      "step": 4603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7107988062266692e-05,
+      "loss": 2.0344,
+      "step": 4604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7106594320487173e-05,
+      "loss": 2.058,
+      "step": 4605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7105200280894146e-05,
+      "loss": 2.0759,
+      "step": 4606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7103805943524166e-05,
+      "loss": 2.0259,
+      "step": 4607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7102411308413793e-05,
+      "loss": 2.0378,
+      "step": 4608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.710101637559961e-05,
+      "loss": 1.9373,
+      "step": 4609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.709962114511819e-05,
+      "loss": 1.9695,
+      "step": 4610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.709822561700612e-05,
+      "loss": 2.092,
+      "step": 4611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7096829791299992e-05,
+      "loss": 2.1123,
+      "step": 4612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7095433668036425e-05,
+      "loss": 2.0657,
+      "step": 4613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7094037247252017e-05,
+      "loss": 2.0601,
+      "step": 4614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7092640528983397e-05,
+      "loss": 2.0205,
+      "step": 4615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7091243513267187e-05,
+      "loss": 2.0518,
+      "step": 4616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.708984620014002e-05,
+      "loss": 2.015,
+      "step": 4617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7088448589638545e-05,
+      "loss": 2.0627,
+      "step": 4618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7087050681799408e-05,
+      "loss": 2.061,
+      "step": 4619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.708565247665927e-05,
+      "loss": 2.0554,
+      "step": 4620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7084253974254794e-05,
+      "loss": 2.0417,
+      "step": 4621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.708285517462266e-05,
+      "loss": 2.0217,
+      "step": 4622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7081456077799547e-05,
+      "loss": 2.0945,
+      "step": 4623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.708005668382214e-05,
+      "loss": 2.1111,
+      "step": 4624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7078656992727146e-05,
+      "loss": 2.094,
+      "step": 4625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.707725700455126e-05,
+      "loss": 2.0764,
+      "step": 4626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.70758567193312e-05,
+      "loss": 2.0674,
+      "step": 4627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.707445613710369e-05,
+      "loss": 2.0752,
+      "step": 4628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7073055257905456e-05,
+      "loss": 2.0728,
+      "step": 4629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.707165408177323e-05,
+      "loss": 2.1011,
+      "step": 4630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7070252608743764e-05,
+      "loss": 2.0693,
+      "step": 4631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.70688508388538e-05,
+      "loss": 2.0759,
+      "step": 4632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7067448772140113e-05,
+      "loss": 2.0737,
+      "step": 4633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7066046408639455e-05,
+      "loss": 2.074,
+      "step": 4634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.706464374838861e-05,
+      "loss": 2.1011,
+      "step": 4635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7063240791424358e-05,
+      "loss": 2.0872,
+      "step": 4636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7061837537783492e-05,
+      "loss": 2.1306,
+      "step": 4637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.706043398750281e-05,
+      "loss": 2.0701,
+      "step": 4638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7059030140619116e-05,
+      "loss": 2.0764,
+      "step": 4639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7057625997169227e-05,
+      "loss": 2.1025,
+      "step": 4640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7056221557189963e-05,
+      "loss": 2.072,
+      "step": 4641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7054816820718155e-05,
+      "loss": 2.0959,
+      "step": 4642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7053411787790644e-05,
+      "loss": 2.1086,
+      "step": 4643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.705200645844427e-05,
+      "loss": 2.0984,
+      "step": 4644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7050600832715893e-05,
+      "loss": 2.0408,
+      "step": 4645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7049194910642366e-05,
+      "loss": 2.0237,
+      "step": 4646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.704778869226056e-05,
+      "loss": 2.103,
+      "step": 4647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.704638217760735e-05,
+      "loss": 2.0979,
+      "step": 4648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7044975366719626e-05,
+      "loss": 2.0603,
+      "step": 4649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7043568259634276e-05,
+      "loss": 2.0823,
+      "step": 4650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7042160856388196e-05,
+      "loss": 2.0635,
+      "step": 4651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.70407531570183e-05,
+      "loss": 2.041,
+      "step": 4652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7039345161561503e-05,
+      "loss": 2.0052,
+      "step": 4653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7037936870054726e-05,
+      "loss": 2.0786,
+      "step": 4654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7036528282534898e-05,
+      "loss": 1.9602,
+      "step": 4655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7035119399038963e-05,
+      "loss": 1.9973,
+      "step": 4656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7033710219603858e-05,
+      "loss": 2.0554,
+      "step": 4657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7032300744266548e-05,
+      "loss": 2.0652,
+      "step": 4658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7030890973063987e-05,
+      "loss": 2.1274,
+      "step": 4659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.702948090603315e-05,
+      "loss": 2.1123,
+      "step": 4660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.702807054321101e-05,
+      "loss": 2.0862,
+      "step": 4661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.702665988463455e-05,
+      "loss": 2.0906,
+      "step": 4662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7025248930340774e-05,
+      "loss": 2.0532,
+      "step": 4663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.702383768036667e-05,
+      "loss": 2.0886,
+      "step": 4664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7022426134749256e-05,
+      "loss": 2.1033,
+      "step": 4665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.702101429352554e-05,
+      "loss": 2.0798,
+      "step": 4666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.701960215673255e-05,
+      "loss": 2.0791,
+      "step": 4667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.701818972440732e-05,
+      "loss": 2.085,
+      "step": 4668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7016776996586886e-05,
+      "loss": 2.0959,
+      "step": 4669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7015363973308295e-05,
+      "loss": 2.0508,
+      "step": 4670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7013950654608602e-05,
+      "loss": 2.0498,
+      "step": 4671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7012537040524872e-05,
+      "loss": 2.1028,
+      "step": 4672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7011123131094177e-05,
+      "loss": 2.0791,
+      "step": 4673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.700970892635359e-05,
+      "loss": 2.0972,
+      "step": 4674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7008294426340197e-05,
+      "loss": 2.0767,
+      "step": 4675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7006879631091093e-05,
+      "loss": 2.0879,
+      "step": 4676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7005464540643382e-05,
+      "loss": 2.0444,
+      "step": 4677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.700404915503417e-05,
+      "loss": 2.0706,
+      "step": 4678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.700263347430058e-05,
+      "loss": 2.0925,
+      "step": 4679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7001217498479724e-05,
+      "loss": 2.0991,
+      "step": 4680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6999801227608745e-05,
+      "loss": 2.0703,
+      "step": 4681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.699838466172478e-05,
+      "loss": 2.064,
+      "step": 4682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.699696780086498e-05,
+      "loss": 2.1184,
+      "step": 4683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.699555064506649e-05,
+      "loss": 2.1062,
+      "step": 4684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6994133194366482e-05,
+      "loss": 2.0481,
+      "step": 4685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.699271544880213e-05,
+      "loss": 2.0383,
+      "step": 4686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6991297408410605e-05,
+      "loss": 2.0859,
+      "step": 4687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6989879073229095e-05,
+      "loss": 2.0906,
+      "step": 4688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6988460443294803e-05,
+      "loss": 2.0093,
+      "step": 4689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6987041518644914e-05,
+      "loss": 2.0327,
+      "step": 4690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6985622299316657e-05,
+      "loss": 2.0857,
+      "step": 4691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6984202785347235e-05,
+      "loss": 2.0784,
+      "step": 4692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6982782976773878e-05,
+      "loss": 2.0586,
+      "step": 4693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6981362873633815e-05,
+      "loss": 2.085,
+      "step": 4694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.69799424759643e-05,
+      "loss": 1.9861,
+      "step": 4695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6978521783802558e-05,
+      "loss": 2.0232,
+      "step": 4696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.697710079718587e-05,
+      "loss": 2.0442,
+      "step": 4697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6975679516151482e-05,
+      "loss": 2.0559,
+      "step": 4698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6974257940736678e-05,
+      "loss": 2.1189,
+      "step": 4699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.6972836070978727e-05,
+      "loss": 2.1042,
+      "step": 4700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.697141390691492e-05,
+      "loss": 2.0566,
+      "step": 4701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.696999144858255e-05,
+      "loss": 2.0522,
+      "step": 4702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6968568696018923e-05,
+      "loss": 2.0918,
+      "step": 4703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6967145649261347e-05,
+      "loss": 2.1001,
+      "step": 4704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6965722308347137e-05,
+      "loss": 2.0911,
+      "step": 4705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.696429867331362e-05,
+      "loss": 2.0474,
+      "step": 4706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6962874744198138e-05,
+      "loss": 2.0908,
+      "step": 4707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6961450521038015e-05,
+      "loss": 2.0779,
+      "step": 4708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6960026003870612e-05,
+      "loss": 2.0249,
+      "step": 4709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6958601192733286e-05,
+      "loss": 2.0151,
+      "step": 4710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.695717608766339e-05,
+      "loss": 2.0444,
+      "step": 4711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6955750688698305e-05,
+      "loss": 2.0574,
+      "step": 4712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.695432499587541e-05,
+      "loss": 2.0613,
+      "step": 4713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6952899009232088e-05,
+      "loss": 2.0537,
+      "step": 4714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.695147272880574e-05,
+      "loss": 2.0537,
+      "step": 4715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6950046154633763e-05,
+      "loss": 2.0918,
+      "step": 4716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.694861928675357e-05,
+      "loss": 2.0381,
+      "step": 4717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6947192125202572e-05,
+      "loss": 2.0686,
+      "step": 4718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6945764670018203e-05,
+      "loss": 2.1306,
+      "step": 4719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6944336921237896e-05,
+      "loss": 2.1138,
+      "step": 4720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.694290887889909e-05,
+      "loss": 2.0459,
+      "step": 4721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.694148054303923e-05,
+      "loss": 2.05,
+      "step": 4722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.694005191369578e-05,
+      "loss": 2.0518,
+      "step": 4723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.69386229909062e-05,
+      "loss": 2.0801,
+      "step": 4724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6937193774707966e-05,
+      "loss": 2.0835,
+      "step": 4725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6935764265138547e-05,
+      "loss": 2.0444,
+      "step": 4726
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.693433446223544e-05,
+      "loss": 2.0317,
+      "step": 4727
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.693290436603614e-05,
+      "loss": 2.0613,
+      "step": 4728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.693147397657814e-05,
+      "loss": 2.0518,
+      "step": 4729
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6930043293898958e-05,
+      "loss": 2.0286,
+      "step": 4730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.692861231803611e-05,
+      "loss": 2.0059,
+      "step": 4731
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6927181049027128e-05,
+      "loss": 1.9917,
+      "step": 4732
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6925749486909534e-05,
+      "loss": 1.9944,
+      "step": 4733
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6924317631720878e-05,
+      "loss": 2.0046,
+      "step": 4734
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6922885483498704e-05,
+      "loss": 2.0591,
+      "step": 4735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.692145304228057e-05,
+      "loss": 2.0625,
+      "step": 4736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6920020308104038e-05,
+      "loss": 2.0393,
+      "step": 4737
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6918587281006685e-05,
+      "loss": 2.0674,
+      "step": 4738
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6917153961026087e-05,
+      "loss": 1.9941,
+      "step": 4739
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6915720348199834e-05,
+      "loss": 2.0256,
+      "step": 4740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6914286442565515e-05,
+      "loss": 2.061,
+      "step": 4741
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6912852244160737e-05,
+      "loss": 2.0278,
+      "step": 4742
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.691141775302311e-05,
+      "loss": 2.0959,
+      "step": 4743
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6909982969190252e-05,
+      "loss": 2.0667,
+      "step": 4744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6908547892699784e-05,
+      "loss": 1.9668,
+      "step": 4745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6907112523589345e-05,
+      "loss": 2.0264,
+      "step": 4746
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6905676861896573e-05,
+      "loss": 2.0481,
+      "step": 4747
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6904240907659112e-05,
+      "loss": 2.0068,
+      "step": 4748
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.690280466091463e-05,
+      "loss": 2.051,
+      "step": 4749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6901368121700785e-05,
+      "loss": 2.0588,
+      "step": 4750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6899931290055243e-05,
+      "loss": 2.0823,
+      "step": 4751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.689849416601569e-05,
+      "loss": 2.0598,
+      "step": 4752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.689705674961981e-05,
+      "loss": 2.0588,
+      "step": 4753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6895619040905304e-05,
+      "loss": 2.0732,
+      "step": 4754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6894181039909864e-05,
+      "loss": 2.0359,
+      "step": 4755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6892742746671204e-05,
+      "loss": 2.0056,
+      "step": 4756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6891304161227042e-05,
+      "loss": 2.0767,
+      "step": 4757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.688986528361511e-05,
+      "loss": 2.0767,
+      "step": 4758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.688842611387313e-05,
+      "loss": 2.0369,
+      "step": 4759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6886986652038848e-05,
+      "loss": 2.0632,
+      "step": 4760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6885546898150008e-05,
+      "loss": 2.0459,
+      "step": 4761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.688410685224437e-05,
+      "loss": 2.083,
+      "step": 4762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6882666514359703e-05,
+      "loss": 2.0698,
+      "step": 4763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6881225884533763e-05,
+      "loss": 2.0369,
+      "step": 4764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6879784962804342e-05,
+      "loss": 1.9968,
+      "step": 4765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.687834374920922e-05,
+      "loss": 2.0256,
+      "step": 4766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6876902243786195e-05,
+      "loss": 2.0928,
+      "step": 4767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6875460446573067e-05,
+      "loss": 2.1067,
+      "step": 4768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6874018357607647e-05,
+      "loss": 2.0264,
+      "step": 4769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6872575976927753e-05,
+      "loss": 2.0398,
+      "step": 4770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.68711333045712e-05,
+      "loss": 2.0398,
+      "step": 4771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.686969034057583e-05,
+      "loss": 2.0605,
+      "step": 4772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.686824708497948e-05,
+      "loss": 2.0413,
+      "step": 4773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6866803537820005e-05,
+      "loss": 2.0303,
+      "step": 4774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.686535969913525e-05,
+      "loss": 2.1123,
+      "step": 4775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6863915568963086e-05,
+      "loss": 2.1006,
+      "step": 4776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6862471147341374e-05,
+      "loss": 2.061,
+      "step": 4777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6861026434308e-05,
+      "loss": 2.0718,
+      "step": 4778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6859581429900846e-05,
+      "loss": 2.0417,
+      "step": 4779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6858136134157806e-05,
+      "loss": 2.021,
+      "step": 4780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6856690547116788e-05,
+      "loss": 2.0317,
+      "step": 4781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6855244668815694e-05,
+      "loss": 2.0457,
+      "step": 4782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6853798499292442e-05,
+      "loss": 2.0823,
+      "step": 4783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6852352038584952e-05,
+      "loss": 2.0923,
+      "step": 4784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6850905286731164e-05,
+      "loss": 2.0789,
+      "step": 4785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.684945824376901e-05,
+      "loss": 2.1104,
+      "step": 4786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6848010909736443e-05,
+      "loss": 2.1272,
+      "step": 4787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6846563284671416e-05,
+      "loss": 2.0981,
+      "step": 4788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6845115368611885e-05,
+      "loss": 2.0679,
+      "step": 4789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.684366716159583e-05,
+      "loss": 2.0732,
+      "step": 4790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.684221866366122e-05,
+      "loss": 2.0349,
+      "step": 4791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6840769874846047e-05,
+      "loss": 2.0349,
+      "step": 4792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.68393207951883e-05,
+      "loss": 2.0923,
+      "step": 4793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.683787142472598e-05,
+      "loss": 2.0913,
+      "step": 4794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.683642176349709e-05,
+      "loss": 2.1045,
+      "step": 4795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.683497181153966e-05,
+      "loss": 2.0801,
+      "step": 4796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6833521568891695e-05,
+      "loss": 2.0872,
+      "step": 4797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6832071035591238e-05,
+      "loss": 2.0964,
+      "step": 4798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6830620211676327e-05,
+      "loss": 2.1008,
+      "step": 4799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6829169097185005e-05,
+      "loss": 2.116,
+      "step": 4800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.682771769215532e-05,
+      "loss": 2.0591,
+      "step": 4801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.682626599662535e-05,
+      "loss": 2.0544,
+      "step": 4802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.682481401063315e-05,
+      "loss": 2.0808,
+      "step": 4803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6823361734216796e-05,
+      "loss": 2.0637,
+      "step": 4804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6821909167414384e-05,
+      "loss": 2.0181,
+      "step": 4805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6820456310264e-05,
+      "loss": 2.0278,
+      "step": 4806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6819003162803737e-05,
+      "loss": 2.0913,
+      "step": 4807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6817549725071708e-05,
+      "loss": 2.0908,
+      "step": 4808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6816095997106025e-05,
+      "loss": 2.0806,
+      "step": 4809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6814641978944815e-05,
+      "loss": 2.0515,
+      "step": 4810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6813187670626207e-05,
+      "loss": 2.0798,
+      "step": 4811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6811733072188335e-05,
+      "loss": 2.0732,
+      "step": 4812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6810278183669347e-05,
+      "loss": 2.0083,
+      "step": 4813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6808823005107395e-05,
+      "loss": 2.0474,
+      "step": 4814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6807367536540633e-05,
+      "loss": 2.1035,
+      "step": 4815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6805911778007243e-05,
+      "loss": 2.0828,
+      "step": 4816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6804455729545396e-05,
+      "loss": 2.0703,
+      "step": 4817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.680299939119326e-05,
+      "loss": 2.0891,
+      "step": 4818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6801542762989046e-05,
+      "loss": 2.0884,
+      "step": 4819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6800085844970937e-05,
+      "loss": 2.0835,
+      "step": 4820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6798628637177154e-05,
+      "loss": 2.0476,
+      "step": 4821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6797171139645903e-05,
+      "loss": 2.062,
+      "step": 4822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6795713352415397e-05,
+      "loss": 2.1174,
+      "step": 4823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.679425527552388e-05,
+      "loss": 2.092,
+      "step": 4824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.679279690900958e-05,
+      "loss": 2.085,
+      "step": 4825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.679133825291074e-05,
+      "loss": 2.0522,
+      "step": 4826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.678987930726561e-05,
+      "loss": 2.0725,
+      "step": 4827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6788420072112455e-05,
+      "loss": 2.0544,
+      "step": 4828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6786960547489544e-05,
+      "loss": 2.0889,
+      "step": 4829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6785500733435146e-05,
+      "loss": 2.0706,
+      "step": 4830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.678404062998754e-05,
+      "loss": 2.1003,
+      "step": 4831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6782580237185023e-05,
+      "loss": 2.0967,
+      "step": 4832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6781119555065883e-05,
+      "loss": 2.0469,
+      "step": 4833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6779658583668434e-05,
+      "loss": 2.0481,
+      "step": 4834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6778197323030986e-05,
+      "loss": 2.0256,
+      "step": 4835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.677673577319185e-05,
+      "loss": 2.083,
+      "step": 4836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6775273934189365e-05,
+      "loss": 2.0615,
+      "step": 4837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.677381180606186e-05,
+      "loss": 2.062,
+      "step": 4838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6772349388847685e-05,
+      "loss": 2.0881,
+      "step": 4839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.677088668258518e-05,
+      "loss": 2.0796,
+      "step": 4840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6769423687312702e-05,
+      "loss": 2.0833,
+      "step": 4841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.676796040306863e-05,
+      "loss": 2.0544,
+      "step": 4842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6766496829891326e-05,
+      "loss": 2.0701,
+      "step": 4843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.676503296781917e-05,
+      "loss": 2.0728,
+      "step": 4844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6763568816890556e-05,
+      "loss": 2.0967,
+      "step": 4845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.676210437714388e-05,
+      "loss": 2.1123,
+      "step": 4846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.676063964861754e-05,
+      "loss": 2.052,
+      "step": 4847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6759174631349944e-05,
+      "loss": 2.0466,
+      "step": 4848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.675770932537952e-05,
+      "loss": 2.1123,
+      "step": 4849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6756243730744686e-05,
+      "loss": 2.0964,
+      "step": 4850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6754777847483878e-05,
+      "loss": 2.0149,
+      "step": 4851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6753311675635543e-05,
+      "loss": 2.0325,
+      "step": 4852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.675184521523812e-05,
+      "loss": 2.0852,
+      "step": 4853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6750378466330073e-05,
+      "loss": 2.0754,
+      "step": 4854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.674891142894986e-05,
+      "loss": 2.0312,
+      "step": 4855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6747444103135953e-05,
+      "loss": 2.0564,
+      "step": 4856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6745976488926837e-05,
+      "loss": 2.0632,
+      "step": 4857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.674450858636099e-05,
+      "loss": 2.0684,
+      "step": 4858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6743040395476912e-05,
+      "loss": 2.0139,
+      "step": 4859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.674157191631311e-05,
+      "loss": 2.0359,
+      "step": 4860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6740103148908074e-05,
+      "loss": 2.0654,
+      "step": 4861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.673863409330034e-05,
+      "loss": 2.0356,
+      "step": 4862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.673716474952842e-05,
+      "loss": 2.0645,
+      "step": 4863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6735695117630852e-05,
+      "loss": 2.0588,
+      "step": 4864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6734225197646177e-05,
+      "loss": 2.0889,
+      "step": 4865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.673275498961294e-05,
+      "loss": 2.0901,
+      "step": 4866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.673128449356969e-05,
+      "loss": 2.0769,
+      "step": 4867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6729813709555e-05,
+      "loss": 2.0693,
+      "step": 4868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6728342637607427e-05,
+      "loss": 2.0432,
+      "step": 4869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6726871277765557e-05,
+      "loss": 2.075,
+      "step": 4870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6725399630067972e-05,
+      "loss": 2.0837,
+      "step": 4871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6723927694553265e-05,
+      "loss": 2.0776,
+      "step": 4872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6722455471260032e-05,
+      "loss": 2.0422,
+      "step": 4873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.672098296022689e-05,
+      "loss": 2.0632,
+      "step": 4874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6719510161492443e-05,
+      "loss": 2.0149,
+      "step": 4875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6718037075095317e-05,
+      "loss": 1.9844,
+      "step": 4876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6716563701074147e-05,
+      "loss": 2.0647,
+      "step": 4877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6715090039467565e-05,
+      "loss": 2.0916,
+      "step": 4878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6713616090314215e-05,
+      "loss": 2.041,
+      "step": 4879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6712141853652754e-05,
+      "loss": 2.0276,
+      "step": 4880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.671066732952184e-05,
+      "loss": 2.0667,
+      "step": 4881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.670919251796014e-05,
+      "loss": 2.064,
+      "step": 4882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6707717419006333e-05,
+      "loss": 2.032,
+      "step": 4883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6706242032699094e-05,
+      "loss": 2.0168,
+      "step": 4884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6704766359077123e-05,
+      "loss": 2.0276,
+      "step": 4885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6703290398179108e-05,
+      "loss": 2.0022,
+      "step": 4886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6701814150043766e-05,
+      "loss": 2.0674,
+      "step": 4887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6700337614709798e-05,
+      "loss": 2.0913,
+      "step": 4888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.669886079221593e-05,
+      "loss": 2.0649,
+      "step": 4889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6697383682600887e-05,
+      "loss": 2.0342,
+      "step": 4890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6695906285903414e-05,
+      "loss": 2.0718,
+      "step": 4891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6694428602162247e-05,
+      "loss": 2.0793,
+      "step": 4892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.669295063141613e-05,
+      "loss": 2.0908,
+      "step": 4893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.669147237370383e-05,
+      "loss": 2.0969,
+      "step": 4894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6689993829064117e-05,
+      "loss": 2.1042,
+      "step": 4895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.668851499753575e-05,
+      "loss": 2.0815,
+      "step": 4896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.668703587915752e-05,
+      "loss": 2.0793,
+      "step": 4897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6685556473968213e-05,
+      "loss": 2.0764,
+      "step": 4898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.668407678200662e-05,
+      "loss": 2.0789,
+      "step": 4899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.6682596803311553e-05,
+      "loss": 2.0872,
+      "step": 4900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6681116537921817e-05,
+      "loss": 2.0825,
+      "step": 4901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.667963598587623e-05,
+      "loss": 2.0962,
+      "step": 4902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6678155147213616e-05,
+      "loss": 2.0764,
+      "step": 4903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6676674021972814e-05,
+      "loss": 2.0513,
+      "step": 4904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6675192610192662e-05,
+      "loss": 2.0322,
+      "step": 4905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.667371091191201e-05,
+      "loss": 2.0371,
+      "step": 4906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.667222892716971e-05,
+      "loss": 2.0657,
+      "step": 4907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6670746656004627e-05,
+      "loss": 2.0732,
+      "step": 4908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6669264098455634e-05,
+      "loss": 2.095,
+      "step": 4909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6667781254561604e-05,
+      "loss": 2.1233,
+      "step": 4910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6666298124361432e-05,
+      "loss": 2.0396,
+      "step": 4911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6664814707894004e-05,
+      "loss": 2.01,
+      "step": 4912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6663331005198218e-05,
+      "loss": 2.0725,
+      "step": 4913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6661847016312992e-05,
+      "loss": 2.0706,
+      "step": 4914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6660362741277237e-05,
+      "loss": 2.0267,
+      "step": 4915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6658878180129874e-05,
+      "loss": 2.0474,
+      "step": 4916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6657393332909838e-05,
+      "loss": 2.0378,
+      "step": 4917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6655908199656066e-05,
+      "loss": 2.082,
+      "step": 4918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6654422780407498e-05,
+      "loss": 2.0161,
+      "step": 4919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6652937075203103e-05,
+      "loss": 2.0259,
+      "step": 4920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6651451084081823e-05,
+      "loss": 2.0116,
+      "step": 4921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.664996480708264e-05,
+      "loss": 2.0293,
+      "step": 4922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6648478244244522e-05,
+      "loss": 2.0891,
+      "step": 4923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.664699139560646e-05,
+      "loss": 2.0813,
+      "step": 4924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6645504261207433e-05,
+      "loss": 2.0657,
+      "step": 4925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6644016841086456e-05,
+      "loss": 2.0737,
+      "step": 4926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.664252913528252e-05,
+      "loss": 2.0747,
+      "step": 4927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6641041143834648e-05,
+      "loss": 2.0142,
+      "step": 4928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6639552866781853e-05,
+      "loss": 2.0618,
+      "step": 4929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6638064304163165e-05,
+      "loss": 2.0818,
+      "step": 4930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.663657545601763e-05,
+      "loss": 2.074,
+      "step": 4931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6635086322384277e-05,
+      "loss": 2.0537,
+      "step": 4932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6633596903302163e-05,
+      "loss": 2.0859,
+      "step": 4933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6632107198810347e-05,
+      "loss": 2.0266,
+      "step": 4934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6630617208947895e-05,
+      "loss": 2.093,
+      "step": 4935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.662912693375388e-05,
+      "loss": 2.0955,
+      "step": 4936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.662763637326738e-05,
+      "loss": 2.0747,
+      "step": 4937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.662614552752749e-05,
+      "loss": 2.0942,
+      "step": 4938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6624654396573298e-05,
+      "loss": 2.0287,
+      "step": 4939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6623162980443915e-05,
+      "loss": 2.0591,
+      "step": 4940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.662167127917844e-05,
+      "loss": 2.0574,
+      "step": 4941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6620179292816e-05,
+      "loss": 2.0249,
+      "step": 4942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.661868702139572e-05,
+      "loss": 2.1326,
+      "step": 4943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.661719446495673e-05,
+      "loss": 2.1118,
+      "step": 4944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6615701623538177e-05,
+      "loss": 2.115,
+      "step": 4945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6614208497179197e-05,
+      "loss": 2.113,
+      "step": 4946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.661271508591896e-05,
+      "loss": 2.0215,
+      "step": 4947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6611221389796615e-05,
+      "loss": 1.9932,
+      "step": 4948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.660972740885135e-05,
+      "loss": 2.0149,
+      "step": 4949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.660823314312232e-05,
+      "loss": 2.0293,
+      "step": 4950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.660673859264873e-05,
+      "loss": 1.9568,
+      "step": 4951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6605243757469762e-05,
+      "loss": 1.9666,
+      "step": 4952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6603748637624622e-05,
+      "loss": 2.0828,
+      "step": 4953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6602253233152513e-05,
+      "loss": 2.0737,
+      "step": 4954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.660075754409265e-05,
+      "loss": 2.0186,
+      "step": 4955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6599261570484267e-05,
+      "loss": 2.0496,
+      "step": 4956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6597765312366583e-05,
+      "loss": 2.0698,
+      "step": 4957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6596268769778837e-05,
+      "loss": 2.0974,
+      "step": 4958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6594771942760273e-05,
+      "loss": 2.0483,
+      "step": 4959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6593274831350146e-05,
+      "loss": 2.0466,
+      "step": 4960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6591777435587722e-05,
+      "loss": 2.0728,
+      "step": 4961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6590279755512258e-05,
+      "loss": 2.0793,
+      "step": 4962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6588781791163033e-05,
+      "loss": 2.0125,
+      "step": 4963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6587283542579333e-05,
+      "loss": 2.0352,
+      "step": 4964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.658578500980044e-05,
+      "loss": 2.1091,
+      "step": 4965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.658428619286566e-05,
+      "loss": 2.1169,
+      "step": 4966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6582787091814292e-05,
+      "loss": 2.103,
+      "step": 4967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.658128770668565e-05,
+      "loss": 2.1104,
+      "step": 4968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.657978803751905e-05,
+      "loss": 2.0557,
+      "step": 4969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.657828808435383e-05,
+      "loss": 2.05,
+      "step": 4970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6576787847229314e-05,
+      "loss": 2.0542,
+      "step": 4971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6575287326184843e-05,
+      "loss": 2.03,
+      "step": 4972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6573786521259774e-05,
+      "loss": 2.072,
+      "step": 4973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.657228543249346e-05,
+      "loss": 2.0374,
+      "step": 4974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6570784059925263e-05,
+      "loss": 2.0441,
+      "step": 4975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.656928240359456e-05,
+      "loss": 2.0574,
+      "step": 4976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6567780463540727e-05,
+      "loss": 2.0603,
+      "step": 4977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.656627823980315e-05,
+      "loss": 2.052,
+      "step": 4978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6564775732421226e-05,
+      "loss": 2.0718,
+      "step": 4979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6563272941434353e-05,
+      "loss": 2.0723,
+      "step": 4980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6561769866881944e-05,
+      "loss": 2.085,
+      "step": 4981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6560266508803412e-05,
+      "loss": 2.1104,
+      "step": 4982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6558762867238183e-05,
+      "loss": 2.0776,
+      "step": 4983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6557258942225687e-05,
+      "loss": 2.0566,
+      "step": 4984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6555754733805362e-05,
+      "loss": 2.0579,
+      "step": 4985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6554250242016657e-05,
+      "loss": 2.0327,
+      "step": 4986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6552745466899018e-05,
+      "loss": 2.0952,
+      "step": 4987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.655124040849192e-05,
+      "loss": 2.0859,
+      "step": 4988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6549735066834816e-05,
+      "loss": 2.063,
+      "step": 4989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.654822944196719e-05,
+      "loss": 2.0461,
+      "step": 4990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6546723533928527e-05,
+      "loss": 1.9746,
+      "step": 4991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6545217342758314e-05,
+      "loss": 1.9792,
+      "step": 4992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6543710868496046e-05,
+      "loss": 2.0884,
+      "step": 4993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6542204111181235e-05,
+      "loss": 2.1072,
+      "step": 4994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.654069707085339e-05,
+      "loss": 2.0552,
+      "step": 4995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6539189747552033e-05,
+      "loss": 2.0752,
+      "step": 4996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6537682141316696e-05,
+      "loss": 2.0002,
+      "step": 4997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6536174252186907e-05,
+      "loss": 2.0317,
+      "step": 4998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.653466608020221e-05,
+      "loss": 2.0312,
+      "step": 4999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.653315762540216e-05,
+      "loss": 2.0283,
+      "step": 5000
+    },
+    {
+      "Bleu_1": 0.7724589629575086,
+      "Bleu_2": 0.6109441434155072,
+      "Bleu_3": 0.46913198338087564,
+      "Bleu_4": 0.3561108669888887,
+      "CIDEr": 1.203735216668923,
+      "METEOR": 0.27839078757841934,
+      "ROUGE_L": 0.558853893886062,
+      "epoch": 0.25,
+      "step": 5000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6531648887826308e-05,
+      "loss": 2.0535,
+      "step": 5001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.653013986751422e-05,
+      "loss": 2.0388,
+      "step": 5002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6528630564505474e-05,
+      "loss": 2.0618,
+      "step": 5003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6527120978839643e-05,
+      "loss": 2.0745,
+      "step": 5004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6525611110556318e-05,
+      "loss": 2.0466,
+      "step": 5005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6524100959695094e-05,
+      "loss": 2.0432,
+      "step": 5006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6522590526295572e-05,
+      "loss": 2.0598,
+      "step": 5007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6521079810397354e-05,
+      "loss": 2.0652,
+      "step": 5008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6519568812040073e-05,
+      "loss": 2.1018,
+      "step": 5009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6518057531263342e-05,
+      "loss": 2.0867,
+      "step": 5010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.651654596810679e-05,
+      "loss": 2.0386,
+      "step": 5011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.651503412261006e-05,
+      "loss": 2.0251,
+      "step": 5012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6513521994812804e-05,
+      "loss": 2.0574,
+      "step": 5013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6512009584754668e-05,
+      "loss": 2.0322,
+      "step": 5014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.651049689247532e-05,
+      "loss": 2.0266,
+      "step": 5015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6508983918014417e-05,
+      "loss": 2.0032,
+      "step": 5016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6507470661411648e-05,
+      "loss": 2.0588,
+      "step": 5017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6505957122706687e-05,
+      "loss": 2.0725,
+      "step": 5018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.650444330193923e-05,
+      "loss": 2.0642,
+      "step": 5019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6502929199148975e-05,
+      "loss": 2.0452,
+      "step": 5020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.650141481437562e-05,
+      "loss": 2.0549,
+      "step": 5021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.649990014765889e-05,
+      "loss": 2.0649,
+      "step": 5022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6498385199038498e-05,
+      "loss": 2.0691,
+      "step": 5023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6496869968554178e-05,
+      "loss": 2.053,
+      "step": 5024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.649535445624566e-05,
+      "loss": 2.0813,
+      "step": 5025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6493838662152686e-05,
+      "loss": 2.0703,
+      "step": 5026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6492322586315e-05,
+      "loss": 2.0986,
+      "step": 5027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6490806228772378e-05,
+      "loss": 2.0889,
+      "step": 5028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.648928958956456e-05,
+      "loss": 2.0718,
+      "step": 5029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6487772668731344e-05,
+      "loss": 2.071,
+      "step": 5030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6486255466312492e-05,
+      "loss": 2.0918,
+      "step": 5031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6484737982347794e-05,
+      "loss": 2.1194,
+      "step": 5032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.648322021687705e-05,
+      "loss": 2.0825,
+      "step": 5033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6481702169940054e-05,
+      "loss": 2.0969,
+      "step": 5034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6480183841576616e-05,
+      "loss": 2.0967,
+      "step": 5035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.647866523182656e-05,
+      "loss": 2.0898,
+      "step": 5036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6477146340729703e-05,
+      "loss": 2.0366,
+      "step": 5037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6475627168325882e-05,
+      "loss": 2.0339,
+      "step": 5038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6474107714654925e-05,
+      "loss": 2.0898,
+      "step": 5039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6472587979756687e-05,
+      "loss": 2.0593,
+      "step": 5040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6471067963671018e-05,
+      "loss": 2.1113,
+      "step": 5041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.646954766643778e-05,
+      "loss": 2.0837,
+      "step": 5042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6468027088096843e-05,
+      "loss": 2.0134,
+      "step": 5043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6466506228688075e-05,
+      "loss": 2.0244,
+      "step": 5044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6464985088251367e-05,
+      "loss": 2.1143,
+      "step": 5045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6463463666826603e-05,
+      "loss": 2.0813,
+      "step": 5046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6461941964453683e-05,
+      "loss": 2.0928,
+      "step": 5047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6460419981172517e-05,
+      "loss": 2.0786,
+      "step": 5048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6458897717023006e-05,
+      "loss": 2.0779,
+      "step": 5049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.645737517204508e-05,
+      "loss": 2.0972,
+      "step": 5050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6455852346278657e-05,
+      "loss": 2.0649,
+      "step": 5051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6454329239763685e-05,
+      "loss": 2.0449,
+      "step": 5052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.645280585254009e-05,
+      "loss": 2.1055,
+      "step": 5053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.645128218464783e-05,
+      "loss": 2.1157,
+      "step": 5054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6449758236126857e-05,
+      "loss": 2.1143,
+      "step": 5055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.644823400701714e-05,
+      "loss": 2.0952,
+      "step": 5056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6446709497358648e-05,
+      "loss": 2.0869,
+      "step": 5057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6445184707191353e-05,
+      "loss": 2.0867,
+      "step": 5058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6443659636555255e-05,
+      "loss": 2.0464,
+      "step": 5059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6442134285490332e-05,
+      "loss": 2.051,
+      "step": 5060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6440608654036596e-05,
+      "loss": 2.0454,
+      "step": 5061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.643908274223405e-05,
+      "loss": 2.0161,
+      "step": 5062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6437556550122707e-05,
+      "loss": 2.0522,
+      "step": 5063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6436030077742596e-05,
+      "loss": 2.0649,
+      "step": 5064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.643450332513374e-05,
+      "loss": 2.0837,
+      "step": 5065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6432976292336183e-05,
+      "loss": 2.0771,
+      "step": 5066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.643144897938996e-05,
+      "loss": 2.0942,
+      "step": 5067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6429921386335136e-05,
+      "loss": 2.0764,
+      "step": 5068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6428393513211757e-05,
+      "loss": 2.1011,
+      "step": 5069
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6426865360059903e-05,
+      "loss": 2.0764,
+      "step": 5070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6425336926919638e-05,
+      "loss": 2.0427,
+      "step": 5071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6423808213831046e-05,
+      "loss": 2.0625,
+      "step": 5072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6422279220834217e-05,
+      "loss": 2.0688,
+      "step": 5073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6420749947969246e-05,
+      "loss": 2.0593,
+      "step": 5074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6419220395276237e-05,
+      "loss": 2.0273,
+      "step": 5075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6417690562795303e-05,
+      "loss": 2.0354,
+      "step": 5076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6416160450566556e-05,
+      "loss": 2.0955,
+      "step": 5077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.641463005863013e-05,
+      "loss": 2.0938,
+      "step": 5078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.641309938702615e-05,
+      "loss": 2.0527,
+      "step": 5079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6411568435794757e-05,
+      "loss": 2.0603,
+      "step": 5080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6410037204976102e-05,
+      "loss": 2.0471,
+      "step": 5081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.640850569461034e-05,
+      "loss": 2.0791,
+      "step": 5082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.640697390473763e-05,
+      "loss": 2.0947,
+      "step": 5083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6405441835398142e-05,
+      "loss": 2.0989,
+      "step": 5084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6403909486632054e-05,
+      "loss": 2.0737,
+      "step": 5085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6402376858479548e-05,
+      "loss": 2.0676,
+      "step": 5086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6400843950980818e-05,
+      "loss": 2.0645,
+      "step": 5087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6399310764176063e-05,
+      "loss": 2.0566,
+      "step": 5088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6397777298105484e-05,
+      "loss": 2.074,
+      "step": 5089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6396243552809305e-05,
+      "loss": 2.0781,
+      "step": 5090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6394709528327735e-05,
+      "loss": 2.0576,
+      "step": 5091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6393175224701004e-05,
+      "loss": 2.0349,
+      "step": 5092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6391640641969353e-05,
+      "loss": 2.0178,
+      "step": 5093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6390105780173022e-05,
+      "loss": 2.0281,
+      "step": 5094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.638857063935226e-05,
+      "loss": 2.0266,
+      "step": 5095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6387035219547324e-05,
+      "loss": 2.0488,
+      "step": 5096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.638549952079848e-05,
+      "loss": 2.1306,
+      "step": 5097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6383963543146002e-05,
+      "loss": 2.1323,
+      "step": 5098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.6382427286630163e-05,
+      "loss": 2.0505,
+      "step": 5099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6380890751291257e-05,
+      "loss": 2.0569,
+      "step": 5100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6379353937169574e-05,
+      "loss": 2.0854,
+      "step": 5101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.637781684430541e-05,
+      "loss": 2.083,
+      "step": 5102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6376279472739082e-05,
+      "loss": 2.0415,
+      "step": 5103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6374741822510905e-05,
+      "loss": 2.0415,
+      "step": 5104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6373203893661195e-05,
+      "loss": 1.9932,
+      "step": 5105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6371665686230293e-05,
+      "loss": 2.019,
+      "step": 5106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6370127200258525e-05,
+      "loss": 2.0819,
+      "step": 5107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6368588435786245e-05,
+      "loss": 2.0967,
+      "step": 5108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6367049392853797e-05,
+      "loss": 2.0798,
+      "step": 5109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6365510071501552e-05,
+      "loss": 2.0767,
+      "step": 5110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.636397047176987e-05,
+      "loss": 2.0325,
+      "step": 5111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.636243059369912e-05,
+      "loss": 2.0645,
+      "step": 5112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6360890437329693e-05,
+      "loss": 2.1462,
+      "step": 5113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6359350002701978e-05,
+      "loss": 2.1184,
+      "step": 5114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6357809289856357e-05,
+      "loss": 2.0339,
+      "step": 5115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6356268298833254e-05,
+      "loss": 2.0649,
+      "step": 5116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6354727029673066e-05,
+      "loss": 2.0488,
+      "step": 5117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6353185482416217e-05,
+      "loss": 2.062,
+      "step": 5118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6351643657103123e-05,
+      "loss": 2.0493,
+      "step": 5119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6350101553774227e-05,
+      "loss": 2.0859,
+      "step": 5120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6348559172469964e-05,
+      "loss": 2.016,
+      "step": 5121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6347016513230782e-05,
+      "loss": 2.0552,
+      "step": 5122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6345473576097136e-05,
+      "loss": 2.0698,
+      "step": 5123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.634393036110949e-05,
+      "loss": 2.0991,
+      "step": 5124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6342386868308306e-05,
+      "loss": 2.0398,
+      "step": 5125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.634084309773407e-05,
+      "loss": 2.0183,
+      "step": 5126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6339299049427257e-05,
+      "loss": 2.0361,
+      "step": 5127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6337754723428358e-05,
+      "loss": 2.0483,
+      "step": 5128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6336210119777883e-05,
+      "loss": 2.1106,
+      "step": 5129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.633466523851632e-05,
+      "loss": 2.0942,
+      "step": 5130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6333120079684196e-05,
+      "loss": 2.0366,
+      "step": 5131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6331574643322024e-05,
+      "loss": 2.0417,
+      "step": 5132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.633002892947033e-05,
+      "loss": 2.0042,
+      "step": 5133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6328482938169662e-05,
+      "loss": 1.9946,
+      "step": 5134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.632693666946054e-05,
+      "loss": 2.0918,
+      "step": 5135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6325390123383528e-05,
+      "loss": 2.0652,
+      "step": 5136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6323843299979176e-05,
+      "loss": 2.0352,
+      "step": 5137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6322296199288055e-05,
+      "loss": 2.0728,
+      "step": 5138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6320748821350727e-05,
+      "loss": 2.0898,
+      "step": 5139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.631920116620778e-05,
+      "loss": 2.0979,
+      "step": 5140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6317653233899785e-05,
+      "loss": 2.0292,
+      "step": 5141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.631610502446735e-05,
+      "loss": 2.0769,
+      "step": 5142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6314556537951072e-05,
+      "loss": 2.0603,
+      "step": 5143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6313007774391547e-05,
+      "loss": 2.0676,
+      "step": 5144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6311458733829398e-05,
+      "loss": 2.0916,
+      "step": 5145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6309909416305253e-05,
+      "loss": 2.1104,
+      "step": 5146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.630835982185973e-05,
+      "loss": 2.1145,
+      "step": 5147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.630680995053347e-05,
+      "loss": 2.1272,
+      "step": 5148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6305259802367113e-05,
+      "loss": 2.1084,
+      "step": 5149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6303709377401318e-05,
+      "loss": 2.1399,
+      "step": 5150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6302158675676734e-05,
+      "loss": 2.0281,
+      "step": 5151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6300607697234035e-05,
+      "loss": 2.031,
+      "step": 5152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6299056442113885e-05,
+      "loss": 2.0674,
+      "step": 5153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.629750491035697e-05,
+      "loss": 2.0342,
+      "step": 5154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6295953102003975e-05,
+      "loss": 2.0583,
+      "step": 5155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6294401017095594e-05,
+      "loss": 2.0955,
+      "step": 5156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.629284865567253e-05,
+      "loss": 2.0513,
+      "step": 5157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6291296017775487e-05,
+      "loss": 2.0669,
+      "step": 5158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.628974310344519e-05,
+      "loss": 2.1067,
+      "step": 5159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6288189912722355e-05,
+      "loss": 2.0745,
+      "step": 5160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6286636445647715e-05,
+      "loss": 2.0471,
+      "step": 5161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.628508270226201e-05,
+      "loss": 2.0295,
+      "step": 5162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6283528682605976e-05,
+      "loss": 2.0935,
+      "step": 5163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.628197438672038e-05,
+      "loss": 2.0762,
+      "step": 5164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.628041981464597e-05,
+      "loss": 2.0425,
+      "step": 5165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.627886496642352e-05,
+      "loss": 2.0593,
+      "step": 5166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.62773098420938e-05,
+      "loss": 2.0706,
+      "step": 5167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6275754441697593e-05,
+      "loss": 2.0415,
+      "step": 5168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6274198765275682e-05,
+      "loss": 2.0679,
+      "step": 5169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6272642812868874e-05,
+      "loss": 2.0752,
+      "step": 5170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.627108658451796e-05,
+      "loss": 2.0688,
+      "step": 5171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6269530080263763e-05,
+      "loss": 2.0735,
+      "step": 5172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.626797330014709e-05,
+      "loss": 2.0532,
+      "step": 5173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6266416244208767e-05,
+      "loss": 2.0193,
+      "step": 5174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6264858912489632e-05,
+      "loss": 2.0667,
+      "step": 5175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.626330130503052e-05,
+      "loss": 2.0562,
+      "step": 5176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6261743421872275e-05,
+      "loss": 2.0864,
+      "step": 5177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6260185263055754e-05,
+      "loss": 2.0828,
+      "step": 5178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6258626828621816e-05,
+      "loss": 2.0798,
+      "step": 5179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.625706811861133e-05,
+      "loss": 2.0854,
+      "step": 5180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6255509133065177e-05,
+      "loss": 2.0657,
+      "step": 5181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6253949872024238e-05,
+      "loss": 2.0942,
+      "step": 5182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.625239033552939e-05,
+      "loss": 2.0708,
+      "step": 5183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6250830523621543e-05,
+      "loss": 2.083,
+      "step": 5184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6249270436341597e-05,
+      "loss": 2.1055,
+      "step": 5185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6247710073730464e-05,
+      "loss": 2.0972,
+      "step": 5186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6246149435829067e-05,
+      "loss": 2.103,
+      "step": 5187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6244588522678326e-05,
+      "loss": 2.1106,
+      "step": 5188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6243027334319174e-05,
+      "loss": 2.1294,
+      "step": 5189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.624146587079255e-05,
+      "loss": 2.1145,
+      "step": 5190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6239904132139413e-05,
+      "loss": 2.0879,
+      "step": 5191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6238342118400706e-05,
+      "loss": 2.0769,
+      "step": 5192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.623677982961739e-05,
+      "loss": 2.1069,
+      "step": 5193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.623521726583044e-05,
+      "loss": 2.092,
+      "step": 5194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.623365442708084e-05,
+      "loss": 2.0625,
+      "step": 5195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6232091313409557e-05,
+      "loss": 2.0698,
+      "step": 5196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6230527924857592e-05,
+      "loss": 2.0613,
+      "step": 5197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.622896426146594e-05,
+      "loss": 2.0735,
+      "step": 5198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6227400323275605e-05,
+      "loss": 2.0447,
+      "step": 5199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6225836110327607e-05,
+      "loss": 2.0608,
+      "step": 5200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.622427162266296e-05,
+      "loss": 2.0148,
+      "step": 5201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6222706860322686e-05,
+      "loss": 2.0391,
+      "step": 5202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6221141823347823e-05,
+      "loss": 2.1055,
+      "step": 5203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.621957651177942e-05,
+      "loss": 2.1172,
+      "step": 5204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6218010925658516e-05,
+      "loss": 2.075,
+      "step": 5205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.621644506502617e-05,
+      "loss": 2.0583,
+      "step": 5206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.621487892992344e-05,
+      "loss": 2.0854,
+      "step": 5207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6213312520391403e-05,
+      "loss": 2.0588,
+      "step": 5208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.621174583647114e-05,
+      "loss": 2.0408,
+      "step": 5209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.621017887820372e-05,
+      "loss": 2.0664,
+      "step": 5210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.620861164563025e-05,
+      "loss": 2.0286,
+      "step": 5211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.620704413879182e-05,
+      "loss": 2.0354,
+      "step": 5212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.620547635772954e-05,
+      "loss": 2.0361,
+      "step": 5213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6203908302484522e-05,
+      "loss": 2.0569,
+      "step": 5214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6202339973097886e-05,
+      "loss": 2.0923,
+      "step": 5215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6200771369610757e-05,
+      "loss": 2.1084,
+      "step": 5216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.619920249206428e-05,
+      "loss": 2.1123,
+      "step": 5217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6197633340499586e-05,
+      "loss": 2.1052,
+      "step": 5218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6196063914957827e-05,
+      "loss": 2.074,
+      "step": 5219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6194494215480162e-05,
+      "loss": 2.0505,
+      "step": 5220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6192924242107753e-05,
+      "loss": 2.0981,
+      "step": 5221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6191353994881772e-05,
+      "loss": 2.0759,
+      "step": 5222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6189783473843395e-05,
+      "loss": 2.0576,
+      "step": 5223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6188212679033805e-05,
+      "loss": 2.0554,
+      "step": 5224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.61866416104942e-05,
+      "loss": 2.0625,
+      "step": 5225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6185070268265772e-05,
+      "loss": 2.0408,
+      "step": 5226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6183498652389735e-05,
+      "loss": 2.0183,
+      "step": 5227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6181926762907304e-05,
+      "loss": 2.0173,
+      "step": 5228
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6180354599859692e-05,
+      "loss": 2.0608,
+      "step": 5229
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.617878216328813e-05,
+      "loss": 2.0767,
+      "step": 5230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6177209453233854e-05,
+      "loss": 2.0481,
+      "step": 5231
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.617563646973811e-05,
+      "loss": 2.0637,
+      "step": 5232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6174063212842144e-05,
+      "loss": 2.0535,
+      "step": 5233
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.617248968258721e-05,
+      "loss": 2.0801,
+      "step": 5234
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.617091587901458e-05,
+      "loss": 2.1233,
+      "step": 5235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6169341802165513e-05,
+      "loss": 2.1321,
+      "step": 5236
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.61677674520813e-05,
+      "loss": 2.0864,
+      "step": 5237
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6166192828803228e-05,
+      "loss": 2.0752,
+      "step": 5238
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6164617932372574e-05,
+      "loss": 2.0332,
+      "step": 5239
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6163042762830647e-05,
+      "loss": 2.0532,
+      "step": 5240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.616146732021876e-05,
+      "loss": 2.0002,
+      "step": 5241
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6159891604578215e-05,
+      "loss": 2.0002,
+      "step": 5242
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.615831561595034e-05,
+      "loss": 2.0078,
+      "step": 5243
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6156739354376467e-05,
+      "loss": 2.0383,
+      "step": 5244
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.615516281989792e-05,
+      "loss": 2.0461,
+      "step": 5245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.615358601255606e-05,
+      "loss": 2.0234,
+      "step": 5246
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6152008932392216e-05,
+      "loss": 2.0903,
+      "step": 5247
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6150431579447763e-05,
+      "loss": 2.0718,
+      "step": 5248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.614885395376405e-05,
+      "loss": 2.0574,
+      "step": 5249
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6147276055382457e-05,
+      "loss": 2.0828,
+      "step": 5250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6145697884344367e-05,
+      "loss": 2.0857,
+      "step": 5251
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6144119440691156e-05,
+      "loss": 2.0879,
+      "step": 5252
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6142540724464222e-05,
+      "loss": 2.0637,
+      "step": 5253
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6140961735704967e-05,
+      "loss": 2.0715,
+      "step": 5254
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6139382474454795e-05,
+      "loss": 2.0496,
+      "step": 5255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6137802940755118e-05,
+      "loss": 2.0645,
+      "step": 5256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6136223134647358e-05,
+      "loss": 2.0889,
+      "step": 5257
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6134643056172945e-05,
+      "loss": 2.0874,
+      "step": 5258
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.613306270537332e-05,
+      "loss": 2.0476,
+      "step": 5259
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6131482082289918e-05,
+      "loss": 2.0066,
+      "step": 5260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6129901186964195e-05,
+      "loss": 2.0889,
+      "step": 5261
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6128320019437604e-05,
+      "loss": 2.082,
+      "step": 5262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6126738579751605e-05,
+      "loss": 2.0913,
+      "step": 5263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.612515686794768e-05,
+      "loss": 2.1355,
+      "step": 5264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6123574884067305e-05,
+      "loss": 2.0088,
+      "step": 5265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6121992628151963e-05,
+      "loss": 2.0449,
+      "step": 5266
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.612041010024314e-05,
+      "loss": 2.0691,
+      "step": 5267
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6118827300382347e-05,
+      "loss": 2.064,
+      "step": 5268
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.611724422861109e-05,
+      "loss": 2.0425,
+      "step": 5269
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6115660884970874e-05,
+      "loss": 2.0356,
+      "step": 5270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6114077269503232e-05,
+      "loss": 2.0588,
+      "step": 5271
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6112493382249684e-05,
+      "loss": 2.071,
+      "step": 5272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.611090922325177e-05,
+      "loss": 2.0322,
+      "step": 5273
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6109324792551032e-05,
+      "loss": 2.0271,
+      "step": 5274
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.610774009018902e-05,
+      "loss": 2.124,
+      "step": 5275
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6106155116207287e-05,
+      "loss": 2.0906,
+      "step": 5276
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.610456987064741e-05,
+      "loss": 2.0789,
+      "step": 5277
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.610298435355094e-05,
+      "loss": 2.0613,
+      "step": 5278
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6101398564959468e-05,
+      "loss": 2.0515,
+      "step": 5279
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.609981250491458e-05,
+      "loss": 2.051,
+      "step": 5280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6098226173457865e-05,
+      "loss": 2.0039,
+      "step": 5281
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.609663957063092e-05,
+      "loss": 2.0156,
+      "step": 5282
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.609505269647536e-05,
+      "loss": 2.0913,
+      "step": 5283
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6093465551032792e-05,
+      "loss": 2.0986,
+      "step": 5284
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.609187813434484e-05,
+      "loss": 2.0012,
+      "step": 5285
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6090290446453132e-05,
+      "loss": 2.0261,
+      "step": 5286
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6088702487399307e-05,
+      "loss": 1.9907,
+      "step": 5287
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6087114257224998e-05,
+      "loss": 1.9854,
+      "step": 5288
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6085525755971865e-05,
+      "loss": 2.0659,
+      "step": 5289
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6083936983681555e-05,
+      "loss": 2.0696,
+      "step": 5290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6082347940395735e-05,
+      "loss": 2.0391,
+      "step": 5291
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.608075862615608e-05,
+      "loss": 2.073,
+      "step": 5292
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6079169041004266e-05,
+      "loss": 2.0574,
+      "step": 5293
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6077579184981975e-05,
+      "loss": 2.0291,
+      "step": 5294
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6075989058130898e-05,
+      "loss": 2.0613,
+      "step": 5295
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6074398660492745e-05,
+      "loss": 2.0806,
+      "step": 5296
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6072807992109208e-05,
+      "loss": 2.1199,
+      "step": 5297
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6071217053022008e-05,
+      "loss": 2.0776,
+      "step": 5298
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.6069625843272867e-05,
+      "loss": 2.0327,
+      "step": 5299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.606803436290351e-05,
+      "loss": 2.0208,
+      "step": 5300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6066442611955666e-05,
+      "loss": 2.0903,
+      "step": 5301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.606485059047109e-05,
+      "loss": 2.1172,
+      "step": 5302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6063258298491524e-05,
+      "loss": 2.0259,
+      "step": 5303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6061665736058715e-05,
+      "loss": 2.0547,
+      "step": 5304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6060072903214445e-05,
+      "loss": 2.085,
+      "step": 5305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6058479800000468e-05,
+      "loss": 2.0549,
+      "step": 5306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.605688642645857e-05,
+      "loss": 2.0125,
+      "step": 5307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6055292782630528e-05,
+      "loss": 2.022,
+      "step": 5308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6053698868558147e-05,
+      "loss": 2.0564,
+      "step": 5309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6052104684283212e-05,
+      "loss": 2.0581,
+      "step": 5310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6050510229847528e-05,
+      "loss": 2.04,
+      "step": 5311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.604891550529292e-05,
+      "loss": 2.053,
+      "step": 5312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6047320510661194e-05,
+      "loss": 2.054,
+      "step": 5313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.604572524599419e-05,
+      "loss": 2.0823,
+      "step": 5314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6044129711333735e-05,
+      "loss": 2.0859,
+      "step": 5315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.604253390672167e-05,
+      "loss": 2.0784,
+      "step": 5316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6040937832199843e-05,
+      "loss": 2.0836,
+      "step": 5317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6039341487810105e-05,
+      "loss": 2.0945,
+      "step": 5318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.603774487359433e-05,
+      "loss": 2.1201,
+      "step": 5319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6036147989594384e-05,
+      "loss": 2.1201,
+      "step": 5320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.603455083585213e-05,
+      "loss": 2.1223,
+      "step": 5321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.603295341240947e-05,
+      "loss": 2.0898,
+      "step": 5322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.603135571930828e-05,
+      "loss": 2.0747,
+      "step": 5323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6029757756590467e-05,
+      "loss": 2.0806,
+      "step": 5324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6028159524297935e-05,
+      "loss": 2.0527,
+      "step": 5325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6026561022472586e-05,
+      "loss": 2.052,
+      "step": 5326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6024962251156353e-05,
+      "loss": 2.0698,
+      "step": 5327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6023363210391157e-05,
+      "loss": 2.0554,
+      "step": 5328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.602176390021892e-05,
+      "loss": 2.0518,
+      "step": 5329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.60201643206816e-05,
+      "loss": 2.0718,
+      "step": 5330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6018564471821126e-05,
+      "loss": 2.0205,
+      "step": 5331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6016964353679468e-05,
+      "loss": 2.043,
+      "step": 5332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6015363966298578e-05,
+      "loss": 2.0674,
+      "step": 5333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.601376330972043e-05,
+      "loss": 2.1157,
+      "step": 5334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6012162383986995e-05,
+      "loss": 2.0627,
+      "step": 5335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6010561189140256e-05,
+      "loss": 2.0811,
+      "step": 5336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6008959725222202e-05,
+      "loss": 2.0593,
+      "step": 5337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6007357992274835e-05,
+      "loss": 2.0203,
+      "step": 5338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.600575599034015e-05,
+      "loss": 2.0833,
+      "step": 5339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.6004153719460162e-05,
+      "loss": 2.0781,
+      "step": 5340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.600255117967689e-05,
+      "loss": 2.0444,
+      "step": 5341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.600094837103236e-05,
+      "loss": 2.0288,
+      "step": 5342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.59993452935686e-05,
+      "loss": 2.1099,
+      "step": 5343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5997741947327648e-05,
+      "loss": 2.1006,
+      "step": 5344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5996138332351555e-05,
+      "loss": 2.0867,
+      "step": 5345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5994534448682364e-05,
+      "loss": 2.0613,
+      "step": 5346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5992930296362146e-05,
+      "loss": 2.0535,
+      "step": 5347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5991325875432967e-05,
+      "loss": 2.0449,
+      "step": 5348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.59897211859369e-05,
+      "loss": 2.0137,
+      "step": 5349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5988116227916017e-05,
+      "loss": 2.0095,
+      "step": 5350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5986511001412417e-05,
+      "loss": 2.0947,
+      "step": 5351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5984905506468192e-05,
+      "loss": 2.1035,
+      "step": 5352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5983299743125443e-05,
+      "loss": 2.0737,
+      "step": 5353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.598169371142628e-05,
+      "loss": 2.0835,
+      "step": 5354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.598008741141282e-05,
+      "loss": 2.0601,
+      "step": 5355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5978480843127184e-05,
+      "loss": 2.062,
+      "step": 5356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5976874006611503e-05,
+      "loss": 2.0701,
+      "step": 5357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5975266901907918e-05,
+      "loss": 2.0798,
+      "step": 5358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.597365952905857e-05,
+      "loss": 2.0334,
+      "step": 5359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.597205188810561e-05,
+      "loss": 2.0464,
+      "step": 5360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5970443979091203e-05,
+      "loss": 2.094,
+      "step": 5361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5968835802057507e-05,
+      "loss": 2.0859,
+      "step": 5362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5967227357046693e-05,
+      "loss": 2.0315,
+      "step": 5363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.596561864410095e-05,
+      "loss": 2.0247,
+      "step": 5364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5964009663262456e-05,
+      "loss": 2.1069,
+      "step": 5365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.596240041457341e-05,
+      "loss": 2.1155,
+      "step": 5366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.596079089807601e-05,
+      "loss": 2.113,
+      "step": 5367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5959181113812465e-05,
+      "loss": 2.1252,
+      "step": 5368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5957571061824987e-05,
+      "loss": 2.0627,
+      "step": 5369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5955960742155805e-05,
+      "loss": 2.0784,
+      "step": 5370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5954350154847137e-05,
+      "loss": 2.0435,
+      "step": 5371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5952739299941227e-05,
+      "loss": 2.0588,
+      "step": 5372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5951128177480317e-05,
+      "loss": 2.0571,
+      "step": 5373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.594951678750665e-05,
+      "loss": 2.0671,
+      "step": 5374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5947905130062488e-05,
+      "loss": 2.0547,
+      "step": 5375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.59462932051901e-05,
+      "loss": 2.054,
+      "step": 5376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5944681012931747e-05,
+      "loss": 2.083,
+      "step": 5377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5943068553329716e-05,
+      "loss": 2.0776,
+      "step": 5378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.594145582642628e-05,
+      "loss": 2.04,
+      "step": 5379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5939842832263742e-05,
+      "loss": 2.0637,
+      "step": 5380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5938229570884402e-05,
+      "loss": 2.0776,
+      "step": 5381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5936616042330555e-05,
+      "loss": 2.0779,
+      "step": 5382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.593500224664452e-05,
+      "loss": 2.0083,
+      "step": 5383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.593338818386862e-05,
+      "loss": 2.0132,
+      "step": 5384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.593177385404518e-05,
+      "loss": 2.0339,
+      "step": 5385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5930159257216526e-05,
+      "loss": 2.0339,
+      "step": 5386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5928544393425013e-05,
+      "loss": 2.0723,
+      "step": 5387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5926929262712974e-05,
+      "loss": 2.072,
+      "step": 5388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.592531386512278e-05,
+      "loss": 2.0659,
+      "step": 5389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.592369820069678e-05,
+      "loss": 2.0845,
+      "step": 5390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5922082269477347e-05,
+      "loss": 1.9946,
+      "step": 5391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5920466071506857e-05,
+      "loss": 2.0264,
+      "step": 5392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5918849606827696e-05,
+      "loss": 2.0647,
+      "step": 5393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5917232875482244e-05,
+      "loss": 2.0908,
+      "step": 5394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5915615877512914e-05,
+      "loss": 2.0657,
+      "step": 5395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5913998612962095e-05,
+      "loss": 2.0652,
+      "step": 5396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5912381081872207e-05,
+      "loss": 2.052,
+      "step": 5397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.591076328428566e-05,
+      "loss": 2.0681,
+      "step": 5398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.590914522024489e-05,
+      "loss": 2.0137,
+      "step": 5399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5907526889792322e-05,
+      "loss": 2.0002,
+      "step": 5400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5905908292970388e-05,
+      "loss": 2.0664,
+      "step": 5401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5904289429821548e-05,
+      "loss": 2.0598,
+      "step": 5402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5902670300388244e-05,
+      "loss": 2.062,
+      "step": 5403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5901050904712938e-05,
+      "loss": 2.0713,
+      "step": 5404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5899431242838105e-05,
+      "loss": 2.0842,
+      "step": 5405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5897811314806207e-05,
+      "loss": 2.0994,
+      "step": 5406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.589619112065973e-05,
+      "loss": 2.0378,
+      "step": 5407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5894570660441163e-05,
+      "loss": 2.0635,
+      "step": 5408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5892949934193e-05,
+      "loss": 2.0811,
+      "step": 5409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5891328941957738e-05,
+      "loss": 2.0854,
+      "step": 5410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5889707683777897e-05,
+      "loss": 2.0059,
+      "step": 5411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.588808615969598e-05,
+      "loss": 2.0066,
+      "step": 5412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.588646436975452e-05,
+      "loss": 2.0476,
+      "step": 5413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5884842313996033e-05,
+      "loss": 2.0999,
+      "step": 5414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5883219992463068e-05,
+      "loss": 2.0627,
+      "step": 5415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5881597405198165e-05,
+      "loss": 2.0835,
+      "step": 5416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5879974552243874e-05,
+      "loss": 2.0774,
+      "step": 5417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5878351433642748e-05,
+      "loss": 2.0632,
+      "step": 5418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5876728049437358e-05,
+      "loss": 2.0854,
+      "step": 5419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5875104399670277e-05,
+      "loss": 2.0786,
+      "step": 5420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5873480484384076e-05,
+      "loss": 2.085,
+      "step": 5421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5871856303621347e-05,
+      "loss": 2.1021,
+      "step": 5422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5870231857424676e-05,
+      "loss": 1.9927,
+      "step": 5423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5868607145836662e-05,
+      "loss": 2.0464,
+      "step": 5424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.586698216889992e-05,
+      "loss": 2.0691,
+      "step": 5425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5865356926657054e-05,
+      "loss": 2.0779,
+      "step": 5426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.586373141915069e-05,
+      "loss": 2.0227,
+      "step": 5427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.586210564642345e-05,
+      "loss": 2.0708,
+      "step": 5428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.586047960851797e-05,
+      "loss": 1.9905,
+      "step": 5429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5858853305476895e-05,
+      "loss": 1.999,
+      "step": 5430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5857226737342873e-05,
+      "loss": 2.051,
+      "step": 5431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5855599904158548e-05,
+      "loss": 2.042,
+      "step": 5432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5853972805966588e-05,
+      "loss": 2.0708,
+      "step": 5433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5852345442809668e-05,
+      "loss": 2.0688,
+      "step": 5434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5850717814730458e-05,
+      "loss": 2.0869,
+      "step": 5435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.584908992177164e-05,
+      "loss": 2.0979,
+      "step": 5436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5847461763975905e-05,
+      "loss": 2.0295,
+      "step": 5437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.584583334138595e-05,
+      "loss": 2.0261,
+      "step": 5438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5844204654044476e-05,
+      "loss": 2.1194,
+      "step": 5439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.58425757019942e-05,
+      "loss": 2.1216,
+      "step": 5440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.584094648527783e-05,
+      "loss": 2.0645,
+      "step": 5441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.58393170039381e-05,
+      "loss": 2.0579,
+      "step": 5442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5837687258017728e-05,
+      "loss": 2.0681,
+      "step": 5443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.583605724755947e-05,
+      "loss": 2.082,
+      "step": 5444
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.583442697260606e-05,
+      "loss": 2.0571,
+      "step": 5445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5832796433200242e-05,
+      "loss": 2.0691,
+      "step": 5446
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5831165629384794e-05,
+      "loss": 2.0798,
+      "step": 5447
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.582953456120247e-05,
+      "loss": 2.064,
+      "step": 5448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5827903228696052e-05,
+      "loss": 2.0479,
+      "step": 5449
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5826271631908304e-05,
+      "loss": 2.019,
+      "step": 5450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5824639770882026e-05,
+      "loss": 2.1001,
+      "step": 5451
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5823007645660008e-05,
+      "loss": 2.0698,
+      "step": 5452
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5821375256285046e-05,
+      "loss": 2.1033,
+      "step": 5453
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5819742602799956e-05,
+      "loss": 2.0925,
+      "step": 5454
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5818109685247548e-05,
+      "loss": 2.053,
+      "step": 5455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5816476503670643e-05,
+      "loss": 2.0801,
+      "step": 5456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.581484305811207e-05,
+      "loss": 2.0874,
+      "step": 5457
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5813209348614665e-05,
+      "loss": 2.0798,
+      "step": 5458
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.581157537522127e-05,
+      "loss": 2.1089,
+      "step": 5459
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5809941137974732e-05,
+      "loss": 2.0918,
+      "step": 5460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5808306636917908e-05,
+      "loss": 2.0547,
+      "step": 5461
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5806671872093663e-05,
+      "loss": 2.0176,
+      "step": 5462
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5805036843544865e-05,
+      "loss": 2.0249,
+      "step": 5463
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.580340155131439e-05,
+      "loss": 2.0498,
+      "step": 5464
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5801765995445115e-05,
+      "loss": 2.0586,
+      "step": 5465
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.580013017597995e-05,
+      "loss": 2.0938,
+      "step": 5466
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5798494092961775e-05,
+      "loss": 2.1042,
+      "step": 5467
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.57968577464335e-05,
+      "loss": 2.0818,
+      "step": 5468
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.579522113643804e-05,
+      "loss": 2.1042,
+      "step": 5469
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5793584263018305e-05,
+      "loss": 2.0754,
+      "step": 5470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.579194712621722e-05,
+      "loss": 2.0918,
+      "step": 5471
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.579030972607773e-05,
+      "loss": 2.0854,
+      "step": 5472
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.578867206264276e-05,
+      "loss": 2.0574,
+      "step": 5473
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5787034135955265e-05,
+      "loss": 2.0515,
+      "step": 5474
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.578539594605819e-05,
+      "loss": 2.0674,
+      "step": 5475
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.57837574929945e-05,
+      "loss": 2.0427,
+      "step": 5476
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5782118776807158e-05,
+      "loss": 2.0266,
+      "step": 5477
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5780479797539146e-05,
+      "loss": 2.0527,
+      "step": 5478
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.577884055523343e-05,
+      "loss": 2.0808,
+      "step": 5479
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.577720104993301e-05,
+      "loss": 2.0615,
+      "step": 5480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5775561281680872e-05,
+      "loss": 2.0286,
+      "step": 5481
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.577392125052002e-05,
+      "loss": 2.0222,
+      "step": 5482
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5772280956493466e-05,
+      "loss": 2.0793,
+      "step": 5483
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5770640399644222e-05,
+      "loss": 2.0652,
+      "step": 5484
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5768999580015305e-05,
+      "loss": 2.0088,
+      "step": 5485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5767358497649748e-05,
+      "loss": 2.0337,
+      "step": 5486
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5765717152590583e-05,
+      "loss": 2.1018,
+      "step": 5487
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.576407554488086e-05,
+      "loss": 2.1064,
+      "step": 5488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.576243367456362e-05,
+      "loss": 2.0708,
+      "step": 5489
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.576079154168192e-05,
+      "loss": 2.0737,
+      "step": 5490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5759149146278832e-05,
+      "loss": 2.0669,
+      "step": 5491
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5757506488397417e-05,
+      "loss": 2.085,
+      "step": 5492
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5755863568080753e-05,
+      "loss": 2.0432,
+      "step": 5493
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5754220385371927e-05,
+      "loss": 2.0483,
+      "step": 5494
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5752576940314025e-05,
+      "loss": 2.0574,
+      "step": 5495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5750933232950145e-05,
+      "loss": 2.0491,
+      "step": 5496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5749289263323398e-05,
+      "loss": 2.0852,
+      "step": 5497
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5747645031476886e-05,
+      "loss": 2.0747,
+      "step": 5498
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.5746000537453735e-05,
+      "loss": 2.0535,
+      "step": 5499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5744355781297065e-05,
+      "loss": 2.0496,
+      "step": 5500
+    },
+    {
+      "Bleu_1": 0.7721925443582018,
+      "Bleu_2": 0.6111149865649975,
+      "Bleu_3": 0.4688068885045631,
+      "Bleu_4": 0.35581603515860927,
+      "CIDEr": 1.200469119281331,
+      "METEOR": 0.2778902099598547,
+      "ROUGE_L": 0.5579503931021045,
+      "epoch": 0.28,
+      "step": 5500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.574271076305001e-05,
+      "loss": 2.0569,
+      "step": 5501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5741065482755706e-05,
+      "loss": 2.0579,
+      "step": 5502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5739419940457298e-05,
+      "loss": 2.0991,
+      "step": 5503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.573777413619795e-05,
+      "loss": 2.0752,
+      "step": 5504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5736128070020804e-05,
+      "loss": 2.064,
+      "step": 5505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.573448174196904e-05,
+      "loss": 2.061,
+      "step": 5506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5732835152085817e-05,
+      "loss": 2.0745,
+      "step": 5507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5731188300414332e-05,
+      "loss": 2.0454,
+      "step": 5508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.572954118699776e-05,
+      "loss": 2.095,
+      "step": 5509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5727893811879293e-05,
+      "loss": 2.0999,
+      "step": 5510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5726246175102144e-05,
+      "loss": 2.0874,
+      "step": 5511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.572459827670951e-05,
+      "loss": 2.0857,
+      "step": 5512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5722950116744607e-05,
+      "loss": 2.0461,
+      "step": 5513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.572130169525066e-05,
+      "loss": 2.0498,
+      "step": 5514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.571965301227089e-05,
+      "loss": 2.0884,
+      "step": 5515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5718004067848536e-05,
+      "loss": 2.0608,
+      "step": 5516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5716354862026844e-05,
+      "loss": 2.1116,
+      "step": 5517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5714705394849053e-05,
+      "loss": 2.0884,
+      "step": 5518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5713055666358424e-05,
+      "loss": 1.9695,
+      "step": 5519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5711405676598222e-05,
+      "loss": 2.02,
+      "step": 5520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5709755425611704e-05,
+      "loss": 2.084,
+      "step": 5521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5708104913442163e-05,
+      "loss": 2.1011,
+      "step": 5522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.570645414013287e-05,
+      "loss": 2.0737,
+      "step": 5523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.570480310572712e-05,
+      "loss": 2.061,
+      "step": 5524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5703151810268205e-05,
+      "loss": 2.0828,
+      "step": 5525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5701500253799427e-05,
+      "loss": 2.104,
+      "step": 5526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5699848436364106e-05,
+      "loss": 2.0532,
+      "step": 5527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.569819635800555e-05,
+      "loss": 2.0747,
+      "step": 5528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5696544018767086e-05,
+      "loss": 2.0986,
+      "step": 5529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5694891418692044e-05,
+      "loss": 2.095,
+      "step": 5530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.569323855782376e-05,
+      "loss": 1.9919,
+      "step": 5531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.569158543620558e-05,
+      "loss": 2.0488,
+      "step": 5532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5689932053880857e-05,
+      "loss": 2.0796,
+      "step": 5533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5688278410892948e-05,
+      "loss": 2.0479,
+      "step": 5534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5686624507285218e-05,
+      "loss": 2.0781,
+      "step": 5535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.568497034310104e-05,
+      "loss": 2.0828,
+      "step": 5536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5683315918383784e-05,
+      "loss": 2.1077,
+      "step": 5537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5681661233176842e-05,
+      "loss": 2.1082,
+      "step": 5538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.568000628752361e-05,
+      "loss": 2.0332,
+      "step": 5539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.567835108146748e-05,
+      "loss": 2.0784,
+      "step": 5540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5676695615051867e-05,
+      "loss": 2.0354,
+      "step": 5541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5675039888320173e-05,
+      "loss": 2.0513,
+      "step": 5542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.567338390131582e-05,
+      "loss": 2.1216,
+      "step": 5543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.567172765408224e-05,
+      "loss": 2.1118,
+      "step": 5544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.567007114666286e-05,
+      "loss": 2.0637,
+      "step": 5545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5668414379101127e-05,
+      "loss": 2.0432,
+      "step": 5546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.566675735144048e-05,
+      "loss": 2.0427,
+      "step": 5547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.566510006372438e-05,
+      "loss": 2.0559,
+      "step": 5548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5663442515996277e-05,
+      "loss": 2.0278,
+      "step": 5549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.566178470829965e-05,
+      "loss": 2.0613,
+      "step": 5550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5660126640677964e-05,
+      "loss": 2.1218,
+      "step": 5551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5658468313174705e-05,
+      "loss": 2.1399,
+      "step": 5552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5656809725833357e-05,
+      "loss": 2.063,
+      "step": 5553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5655150878697422e-05,
+      "loss": 2.0242,
+      "step": 5554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5653491771810394e-05,
+      "loss": 2.0542,
+      "step": 5555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5651832405215783e-05,
+      "loss": 2.051,
+      "step": 5556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5650172778957106e-05,
+      "loss": 2.0498,
+      "step": 5557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.564851289307788e-05,
+      "loss": 2.0408,
+      "step": 5558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.564685274762164e-05,
+      "loss": 2.0356,
+      "step": 5559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5645192342631916e-05,
+      "loss": 2.0364,
+      "step": 5560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5643531678152254e-05,
+      "loss": 2.0728,
+      "step": 5561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5641870754226204e-05,
+      "loss": 2.0403,
+      "step": 5562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5640209570897313e-05,
+      "loss": 2.104,
+      "step": 5563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5638548128209153e-05,
+      "loss": 2.0881,
+      "step": 5564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5636886426205292e-05,
+      "loss": 1.9148,
+      "step": 5565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.56352244649293e-05,
+      "loss": 1.9612,
+      "step": 5566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5633562244424768e-05,
+      "loss": 2.0215,
+      "step": 5567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5631899764735282e-05,
+      "loss": 2.0405,
+      "step": 5568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5630237025904434e-05,
+      "loss": 2.0613,
+      "step": 5569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5628574027975837e-05,
+      "loss": 2.0671,
+      "step": 5570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5626910770993096e-05,
+      "loss": 2.0195,
+      "step": 5571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.562524725499983e-05,
+      "loss": 2.0647,
+      "step": 5572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5623583480039656e-05,
+      "loss": 2.0635,
+      "step": 5573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5621919446156212e-05,
+      "loss": 2.0874,
+      "step": 5574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5620255153393136e-05,
+      "loss": 2.0266,
+      "step": 5575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5618590601794068e-05,
+      "loss": 2.0259,
+      "step": 5576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5616925791402656e-05,
+      "loss": 2.0725,
+      "step": 5577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5615260722262567e-05,
+      "loss": 2.0715,
+      "step": 5578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5613595394417458e-05,
+      "loss": 2.0911,
+      "step": 5579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5611929807911003e-05,
+      "loss": 2.094,
+      "step": 5580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.561026396278688e-05,
+      "loss": 2.032,
+      "step": 5581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5608597859088773e-05,
+      "loss": 2.02,
+      "step": 5582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.560693149686037e-05,
+      "loss": 2.1265,
+      "step": 5583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5605264876145383e-05,
+      "loss": 2.1021,
+      "step": 5584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5603597996987505e-05,
+      "loss": 2.0381,
+      "step": 5585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5601930859430445e-05,
+      "loss": 2.0637,
+      "step": 5586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.560026346351793e-05,
+      "loss": 2.0696,
+      "step": 5587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5598595809293688e-05,
+      "loss": 2.0664,
+      "step": 5588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.559692789680144e-05,
+      "loss": 2.0105,
+      "step": 5589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.559525972608494e-05,
+      "loss": 2.0193,
+      "step": 5590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5593591297187917e-05,
+      "loss": 2.0894,
+      "step": 5591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5591922610154134e-05,
+      "loss": 2.0979,
+      "step": 5592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.559025366502735e-05,
+      "loss": 2.0828,
+      "step": 5593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5588584461851322e-05,
+      "loss": 2.1045,
+      "step": 5594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.558691500066984e-05,
+      "loss": 2.0422,
+      "step": 5595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.558524528152667e-05,
+      "loss": 2.019,
+      "step": 5596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.558357530446561e-05,
+      "loss": 2.0789,
+      "step": 5597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5581905069530434e-05,
+      "loss": 2.0813,
+      "step": 5598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.558023457676496e-05,
+      "loss": 2.061,
+      "step": 5599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5578563826212987e-05,
+      "loss": 2.0693,
+      "step": 5600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.557689281791833e-05,
+      "loss": 2.0347,
+      "step": 5601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5575221551924808e-05,
+      "loss": 2.0535,
+      "step": 5602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.557355002827625e-05,
+      "loss": 2.0667,
+      "step": 5603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.557187824701649e-05,
+      "loss": 2.0493,
+      "step": 5604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5570206208189366e-05,
+      "loss": 2.0299,
+      "step": 5605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.556853391183873e-05,
+      "loss": 2.0403,
+      "step": 5606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.556686135800843e-05,
+      "loss": 2.0112,
+      "step": 5607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.556518854674233e-05,
+      "loss": 2.0081,
+      "step": 5608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5563515478084292e-05,
+      "loss": 2.1343,
+      "step": 5609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5561842152078204e-05,
+      "loss": 2.1147,
+      "step": 5610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5560168568767935e-05,
+      "loss": 2.049,
+      "step": 5611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5558494728197375e-05,
+      "loss": 2.0583,
+      "step": 5612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.555682063041042e-05,
+      "loss": 2.0725,
+      "step": 5613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5555146275450972e-05,
+      "loss": 2.0657,
+      "step": 5614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.555347166336294e-05,
+      "loss": 2.0549,
+      "step": 5615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.555179679419023e-05,
+      "loss": 2.0295,
+      "step": 5616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5550121667976778e-05,
+      "loss": 2.0625,
+      "step": 5617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.55484462847665e-05,
+      "loss": 2.0652,
+      "step": 5618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5546770644603337e-05,
+      "loss": 2.0923,
+      "step": 5619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.554509474753123e-05,
+      "loss": 2.0664,
+      "step": 5620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5543418593594126e-05,
+      "loss": 2.0435,
+      "step": 5621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.554174218283598e-05,
+      "loss": 2.0657,
+      "step": 5622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5540065515300754e-05,
+      "loss": 2.0322,
+      "step": 5623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5538388591032418e-05,
+      "loss": 2.0305,
+      "step": 5624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5536711410074947e-05,
+      "loss": 2.0759,
+      "step": 5625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.553503397247233e-05,
+      "loss": 2.0969,
+      "step": 5626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.553335627826854e-05,
+      "loss": 2.106,
+      "step": 5627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5531678327507584e-05,
+      "loss": 2.0986,
+      "step": 5628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5530000120233455e-05,
+      "loss": 2.0676,
+      "step": 5629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.552832165649018e-05,
+      "loss": 2.0706,
+      "step": 5630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5526642936321755e-05,
+      "loss": 2.0972,
+      "step": 5631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.552496395977222e-05,
+      "loss": 2.103,
+      "step": 5632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.552328472688559e-05,
+      "loss": 2.0304,
+      "step": 5633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5521605237705905e-05,
+      "loss": 2.0442,
+      "step": 5634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.551992549227721e-05,
+      "loss": 2.1145,
+      "step": 5635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5518245490643556e-05,
+      "loss": 2.0996,
+      "step": 5636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.551656523284899e-05,
+      "loss": 2.0273,
+      "step": 5637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5514884718937586e-05,
+      "loss": 2.0652,
+      "step": 5638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5513203948953406e-05,
+      "loss": 2.051,
+      "step": 5639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.551152292294053e-05,
+      "loss": 2.0247,
+      "step": 5640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.550984164094304e-05,
+      "loss": 2.113,
+      "step": 5641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5508160103005025e-05,
+      "loss": 2.0942,
+      "step": 5642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.550647830917058e-05,
+      "loss": 2.0381,
+      "step": 5643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5504796259483813e-05,
+      "loss": 2.0813,
+      "step": 5644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5503113953988833e-05,
+      "loss": 2.1094,
+      "step": 5645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.550143139272975e-05,
+      "loss": 2.1145,
+      "step": 5646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5499748575750697e-05,
+      "loss": 2.0989,
+      "step": 5647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.549806550309579e-05,
+      "loss": 2.0728,
+      "step": 5648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5496382174809175e-05,
+      "loss": 2.0645,
+      "step": 5649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5494698590935002e-05,
+      "loss": 2.0479,
+      "step": 5650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.549301475151741e-05,
+      "loss": 2.0637,
+      "step": 5651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5491330656600557e-05,
+      "loss": 2.073,
+      "step": 5652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5489646306228607e-05,
+      "loss": 2.0671,
+      "step": 5653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5487961700445738e-05,
+      "loss": 2.0793,
+      "step": 5654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5486276839296112e-05,
+      "loss": 2.0789,
+      "step": 5655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.548459172282393e-05,
+      "loss": 2.0867,
+      "step": 5656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5482906351073367e-05,
+      "loss": 2.042,
+      "step": 5657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5481220724088633e-05,
+      "loss": 2.0388,
+      "step": 5658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5479534841913916e-05,
+      "loss": 2.0955,
+      "step": 5659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5477848704593444e-05,
+      "loss": 2.085,
+      "step": 5660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5476162312171422e-05,
+      "loss": 2.0466,
+      "step": 5661
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5474475664692073e-05,
+      "loss": 2.075,
+      "step": 5662
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5472788762199635e-05,
+      "loss": 2.0564,
+      "step": 5663
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5471101604738336e-05,
+      "loss": 2.0569,
+      "step": 5664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.546941419235243e-05,
+      "loss": 2.0247,
+      "step": 5665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.546772652508616e-05,
+      "loss": 2.0781,
+      "step": 5666
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.546603860298379e-05,
+      "loss": 2.0391,
+      "step": 5667
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5464350426089576e-05,
+      "loss": 2.0532,
+      "step": 5668
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5462661994447793e-05,
+      "loss": 2.052,
+      "step": 5669
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5460973308102714e-05,
+      "loss": 2.0027,
+      "step": 5670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.545928436709863e-05,
+      "loss": 2.0808,
+      "step": 5671
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5457595171479826e-05,
+      "loss": 2.1052,
+      "step": 5672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5455905721290603e-05,
+      "loss": 2.0513,
+      "step": 5673
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5454216016575257e-05,
+      "loss": 2.0623,
+      "step": 5674
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.545252605737811e-05,
+      "loss": 2.0566,
+      "step": 5675
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.545083584374346e-05,
+      "loss": 2.0571,
+      "step": 5676
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5449145375715656e-05,
+      "loss": 2.0696,
+      "step": 5677
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5447454653339017e-05,
+      "loss": 2.1089,
+      "step": 5678
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5445763676657872e-05,
+      "loss": 2.0422,
+      "step": 5679
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5444072445716583e-05,
+      "loss": 2.0415,
+      "step": 5680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5442380960559483e-05,
+      "loss": 2.1035,
+      "step": 5681
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5440689221230937e-05,
+      "loss": 2.0962,
+      "step": 5682
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5438997227775315e-05,
+      "loss": 1.9685,
+      "step": 5683
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5437304980236975e-05,
+      "loss": 2.0225,
+      "step": 5684
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.54356124786603e-05,
+      "loss": 2.0938,
+      "step": 5685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5433919723089678e-05,
+      "loss": 2.0815,
+      "step": 5686
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.543222671356949e-05,
+      "loss": 2.0735,
+      "step": 5687
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5430533450144144e-05,
+      "loss": 2.0864,
+      "step": 5688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.542883993285804e-05,
+      "loss": 2.063,
+      "step": 5689
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5427146161755583e-05,
+      "loss": 2.0457,
+      "step": 5690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5425452136881193e-05,
+      "loss": 2.0447,
+      "step": 5691
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5423757858279296e-05,
+      "loss": 2.0789,
+      "step": 5692
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.542206332599433e-05,
+      "loss": 2.0215,
+      "step": 5693
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.542036854007071e-05,
+      "loss": 2.0088,
+      "step": 5694
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5418673500552908e-05,
+      "loss": 2.0647,
+      "step": 5695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5416978207485353e-05,
+      "loss": 2.0483,
+      "step": 5696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5415282660912512e-05,
+      "loss": 1.9573,
+      "step": 5697
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5413586860878844e-05,
+      "loss": 2.0024,
+      "step": 5698
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.5411890807428823e-05,
+      "loss": 2.0591,
+      "step": 5699
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.541019450060692e-05,
+      "loss": 2.0417,
+      "step": 5700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5408497940457626e-05,
+      "loss": 2.0425,
+      "step": 5701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5406801127025433e-05,
+      "loss": 2.0574,
+      "step": 5702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5405104060354827e-05,
+      "loss": 2.0718,
+      "step": 5703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5403406740490322e-05,
+      "loss": 2.0811,
+      "step": 5704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5401709167476424e-05,
+      "loss": 2.0671,
+      "step": 5705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5400011341357646e-05,
+      "loss": 2.0647,
+      "step": 5706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.539831326217852e-05,
+      "loss": 2.1123,
+      "step": 5707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5396614929983567e-05,
+      "loss": 2.1047,
+      "step": 5708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5394916344817335e-05,
+      "loss": 2.071,
+      "step": 5709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5393217506724358e-05,
+      "loss": 2.062,
+      "step": 5710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5391518415749187e-05,
+      "loss": 2.0842,
+      "step": 5711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5389819071936385e-05,
+      "loss": 2.0637,
+      "step": 5712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5388119475330513e-05,
+      "loss": 2.0408,
+      "step": 5713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.538641962597613e-05,
+      "loss": 2.0422,
+      "step": 5714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.538471952391783e-05,
+      "loss": 2.0483,
+      "step": 5715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.538301916920018e-05,
+      "loss": 2.03,
+      "step": 5716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.538131856186779e-05,
+      "loss": 2.094,
+      "step": 5717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.537961770196523e-05,
+      "loss": 2.054,
+      "step": 5718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5377916589537127e-05,
+      "loss": 2.0256,
+      "step": 5719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5376215224628076e-05,
+      "loss": 2.0723,
+      "step": 5720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.53745136072827e-05,
+      "loss": 2.0891,
+      "step": 5721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.537281173754562e-05,
+      "loss": 2.1062,
+      "step": 5722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5371109615461464e-05,
+      "loss": 2.1201,
+      "step": 5723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5369407241074874e-05,
+      "loss": 2.1226,
+      "step": 5724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5367704614430485e-05,
+      "loss": 2.1235,
+      "step": 5725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.536600173557295e-05,
+      "loss": 2.1106,
+      "step": 5726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.536429860454693e-05,
+      "loss": 2.0667,
+      "step": 5727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5362595221397075e-05,
+      "loss": 2.0684,
+      "step": 5728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5360891586168065e-05,
+      "loss": 2.0591,
+      "step": 5729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5359187698904578e-05,
+      "loss": 2.0386,
+      "step": 5730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5357483559651286e-05,
+      "loss": 2.0637,
+      "step": 5731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5355779168452886e-05,
+      "loss": 2.0627,
+      "step": 5732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5354074525354073e-05,
+      "loss": 2.0718,
+      "step": 5733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5352369630399544e-05,
+      "loss": 2.084,
+      "step": 5734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.535066448363401e-05,
+      "loss": 2.0615,
+      "step": 5735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.534895908510219e-05,
+      "loss": 2.0793,
+      "step": 5736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5347253434848804e-05,
+      "loss": 1.9956,
+      "step": 5737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.534554753291858e-05,
+      "loss": 2.0234,
+      "step": 5738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5343841379356253e-05,
+      "loss": 2.0269,
+      "step": 5739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5342134974206568e-05,
+      "loss": 2.0137,
+      "step": 5740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.534042831751427e-05,
+      "loss": 2.0942,
+      "step": 5741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5338721409324114e-05,
+      "loss": 2.1128,
+      "step": 5742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.533701424968087e-05,
+      "loss": 2.073,
+      "step": 5743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5335306838629294e-05,
+      "loss": 2.0615,
+      "step": 5744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5333599176214168e-05,
+      "loss": 2.0771,
+      "step": 5745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5331891262480266e-05,
+      "loss": 2.0542,
+      "step": 5746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5330183097472385e-05,
+      "loss": 2.0437,
+      "step": 5747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5328474681235317e-05,
+      "loss": 2.0137,
+      "step": 5748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5326766013813864e-05,
+      "loss": 2.0681,
+      "step": 5749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5325057095252827e-05,
+      "loss": 2.0374,
+      "step": 5750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.532334792559703e-05,
+      "loss": 2.0818,
+      "step": 5751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.532163850489129e-05,
+      "loss": 2.0989,
+      "step": 5752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.531992883318044e-05,
+      "loss": 2.0691,
+      "step": 5753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5318218910509295e-05,
+      "loss": 2.0771,
+      "step": 5754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.531650873692272e-05,
+      "loss": 2.0844,
+      "step": 5755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5314798312465546e-05,
+      "loss": 2.0881,
+      "step": 5756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5313087637182636e-05,
+      "loss": 2.0359,
+      "step": 5757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5311376711118848e-05,
+      "loss": 2.0586,
+      "step": 5758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5309665534319046e-05,
+      "loss": 2.0662,
+      "step": 5759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5307954106828103e-05,
+      "loss": 2.075,
+      "step": 5760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5306242428690906e-05,
+      "loss": 2.0205,
+      "step": 5761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5304530499952334e-05,
+      "loss": 2.0042,
+      "step": 5762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.530281832065729e-05,
+      "loss": 2.1333,
+      "step": 5763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.530110589085066e-05,
+      "loss": 2.1147,
+      "step": 5764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5299393210577362e-05,
+      "loss": 2.0884,
+      "step": 5765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5297680279882304e-05,
+      "loss": 2.0618,
+      "step": 5766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5295967098810414e-05,
+      "loss": 2.0398,
+      "step": 5767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5294253667406605e-05,
+      "loss": 2.0649,
+      "step": 5768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5292539985715814e-05,
+      "loss": 2.0735,
+      "step": 5769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.529082605378299e-05,
+      "loss": 2.0793,
+      "step": 5770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5289111871653066e-05,
+      "loss": 2.0515,
+      "step": 5771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5287397439371002e-05,
+      "loss": 2.0796,
+      "step": 5772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5285682756981752e-05,
+      "loss": 2.0284,
+      "step": 5773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5283967824530284e-05,
+      "loss": 2.0649,
+      "step": 5774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.528225264206157e-05,
+      "loss": 2.0464,
+      "step": 5775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5280537209620587e-05,
+      "loss": 2.042,
+      "step": 5776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5278821527252328e-05,
+      "loss": 2.0377,
+      "step": 5777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5277105595001776e-05,
+      "loss": 2.0508,
+      "step": 5778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5275389412913937e-05,
+      "loss": 2.0718,
+      "step": 5779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5273672981033802e-05,
+      "loss": 2.0801,
+      "step": 5780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.52719562994064e-05,
+      "loss": 2.0776,
+      "step": 5781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5270239368076735e-05,
+      "loss": 2.1025,
+      "step": 5782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.526852218708984e-05,
+      "loss": 2.0239,
+      "step": 5783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5266804756490744e-05,
+      "loss": 2.0332,
+      "step": 5784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.526508707632448e-05,
+      "loss": 2.0803,
+      "step": 5785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.52633691466361e-05,
+      "loss": 2.0564,
+      "step": 5786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5261650967470647e-05,
+      "loss": 2.0376,
+      "step": 5787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5259932538873185e-05,
+      "loss": 2.0405,
+      "step": 5788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5258213860888774e-05,
+      "loss": 2.0535,
+      "step": 5789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5256494933562484e-05,
+      "loss": 2.0557,
+      "step": 5790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5254775756939398e-05,
+      "loss": 2.1013,
+      "step": 5791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5253056331064586e-05,
+      "loss": 2.0737,
+      "step": 5792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5251336655983147e-05,
+      "loss": 2.0574,
+      "step": 5793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5249616731740186e-05,
+      "loss": 2.0557,
+      "step": 5794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.524789655838079e-05,
+      "loss": 2.1189,
+      "step": 5795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5246176135950075e-05,
+      "loss": 2.0996,
+      "step": 5796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.524445546449316e-05,
+      "loss": 2.04,
+      "step": 5797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5242734544055165e-05,
+      "loss": 2.0479,
+      "step": 5798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.524101337468122e-05,
+      "loss": 2.0129,
+      "step": 5799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5239291956416454e-05,
+      "loss": 2.0549,
+      "step": 5800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5237570289306023e-05,
+      "loss": 2.1289,
+      "step": 5801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5235848373395062e-05,
+      "loss": 2.1216,
+      "step": 5802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5234126208728733e-05,
+      "loss": 2.1023,
+      "step": 5803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5232403795352202e-05,
+      "loss": 2.106,
+      "step": 5804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5230681133310626e-05,
+      "loss": 2.052,
+      "step": 5805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5228958222649187e-05,
+      "loss": 2.0159,
+      "step": 5806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5227235063413068e-05,
+      "loss": 2.0894,
+      "step": 5807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5225511655647453e-05,
+      "loss": 2.064,
+      "step": 5808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.522378799939754e-05,
+      "loss": 2.0637,
+      "step": 5809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.522206409470852e-05,
+      "loss": 2.0718,
+      "step": 5810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5220339941625616e-05,
+      "loss": 2.0073,
+      "step": 5811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.521861554019403e-05,
+      "loss": 2.032,
+      "step": 5812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5216890890458987e-05,
+      "loss": 2.0703,
+      "step": 5813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5215165992465716e-05,
+      "loss": 2.0808,
+      "step": 5814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5213440846259438e-05,
+      "loss": 2.0684,
+      "step": 5815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5211715451885412e-05,
+      "loss": 2.0432,
+      "step": 5816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5209989809388866e-05,
+      "loss": 2.0537,
+      "step": 5817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5208263918815072e-05,
+      "loss": 2.0698,
+      "step": 5818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5206537780209272e-05,
+      "loss": 2.0354,
+      "step": 5819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5204811393616746e-05,
+      "loss": 2.0205,
+      "step": 5820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.520308475908275e-05,
+      "loss": 2.0033,
+      "step": 5821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5201357876652583e-05,
+      "loss": 2.0522,
+      "step": 5822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5199630746371515e-05,
+      "loss": 2.0652,
+      "step": 5823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5197903368284846e-05,
+      "loss": 2.0437,
+      "step": 5824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.519617574243787e-05,
+      "loss": 2.0754,
+      "step": 5825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.519444786887589e-05,
+      "loss": 2.0586,
+      "step": 5826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5192719747644224e-05,
+      "loss": 2.0898,
+      "step": 5827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.519099137878819e-05,
+      "loss": 2.063,
+      "step": 5828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.518926276235311e-05,
+      "loss": 2.0269,
+      "step": 5829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5187533898384316e-05,
+      "loss": 2.074,
+      "step": 5830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5185804786927137e-05,
+      "loss": 1.9952,
+      "step": 5831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.518407542802693e-05,
+      "loss": 2.0322,
+      "step": 5832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5182345821729035e-05,
+      "loss": 2.0352,
+      "step": 5833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5180615968078816e-05,
+      "loss": 2.0337,
+      "step": 5834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5178885867121634e-05,
+      "loss": 1.9797,
+      "step": 5835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.517715551890286e-05,
+      "loss": 2.0322,
+      "step": 5836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5175424923467864e-05,
+      "loss": 2.0781,
+      "step": 5837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5173694080862037e-05,
+      "loss": 2.0415,
+      "step": 5838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.517196299113077e-05,
+      "loss": 2.1133,
+      "step": 5839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.517023165431945e-05,
+      "loss": 2.125,
+      "step": 5840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5168500070473485e-05,
+      "loss": 2.0759,
+      "step": 5841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.516676823963828e-05,
+      "loss": 2.0708,
+      "step": 5842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.516503616185925e-05,
+      "loss": 2.114,
+      "step": 5843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.516330383718182e-05,
+      "loss": 2.1091,
+      "step": 5844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5161571265651423e-05,
+      "loss": 2.1042,
+      "step": 5845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5159838447313485e-05,
+      "loss": 2.0713,
+      "step": 5846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.515810538221345e-05,
+      "loss": 2.05,
+      "step": 5847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5156372070396768e-05,
+      "loss": 2.0969,
+      "step": 5848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5154638511908883e-05,
+      "loss": 2.0732,
+      "step": 5849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5152904706795274e-05,
+      "loss": 2.0671,
+      "step": 5850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5151170655101392e-05,
+      "loss": 2.0579,
+      "step": 5851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5149436356872713e-05,
+      "loss": 2.0723,
+      "step": 5852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5147701812154725e-05,
+      "loss": 2.0758,
+      "step": 5853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5145967020992908e-05,
+      "loss": 2.0591,
+      "step": 5854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5144231983432757e-05,
+      "loss": 2.0999,
+      "step": 5855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5142496699519772e-05,
+      "loss": 2.0908,
+      "step": 5856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.514076116929945e-05,
+      "loss": 2.041,
+      "step": 5857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.513902539281732e-05,
+      "loss": 2.0173,
+      "step": 5858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5137289370118884e-05,
+      "loss": 2.1035,
+      "step": 5859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5135553101249682e-05,
+      "loss": 2.0691,
+      "step": 5860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.513381658625523e-05,
+      "loss": 2.0913,
+      "step": 5861
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5132079825181078e-05,
+      "loss": 2.0869,
+      "step": 5862
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5130342818072766e-05,
+      "loss": 2.0573,
+      "step": 5863
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5128605564975847e-05,
+      "loss": 2.0334,
+      "step": 5864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.512686806593588e-05,
+      "loss": 2.0193,
+      "step": 5865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5125130320998422e-05,
+      "loss": 2.0002,
+      "step": 5866
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.512339233020905e-05,
+      "loss": 2.0996,
+      "step": 5867
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5121654093613337e-05,
+      "loss": 2.1179,
+      "step": 5868
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.511991561125687e-05,
+      "loss": 2.0867,
+      "step": 5869
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5118176883185236e-05,
+      "loss": 2.0857,
+      "step": 5870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5116437909444026e-05,
+      "loss": 2.0143,
+      "step": 5871
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5114698690078853e-05,
+      "loss": 2.0693,
+      "step": 5872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5112959225135323e-05,
+      "loss": 2.0491,
+      "step": 5873
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.511121951465904e-05,
+      "loss": 2.0383,
+      "step": 5874
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5109479558695647e-05,
+      "loss": 2.0476,
+      "step": 5875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5107739357290752e-05,
+      "loss": 2.0886,
+      "step": 5876
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5105998910490004e-05,
+      "loss": 2.0593,
+      "step": 5877
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5104258218339034e-05,
+      "loss": 2.0417,
+      "step": 5878
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5102517280883493e-05,
+      "loss": 2.0955,
+      "step": 5879
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5100776098169045e-05,
+      "loss": 2.085,
+      "step": 5880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5099034670241333e-05,
+      "loss": 2.1169,
+      "step": 5881
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.509729299714604e-05,
+      "loss": 2.0867,
+      "step": 5882
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5095551078928824e-05,
+      "loss": 2.0701,
+      "step": 5883
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5093808915635378e-05,
+      "loss": 2.0632,
+      "step": 5884
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5092066507311387e-05,
+      "loss": 2.0433,
+      "step": 5885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.509032385400253e-05,
+      "loss": 2.103,
+      "step": 5886
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5088580955754523e-05,
+      "loss": 2.0515,
+      "step": 5887
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5086837812613062e-05,
+      "loss": 2.0269,
+      "step": 5888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5085094424623862e-05,
+      "loss": 2.0286,
+      "step": 5889
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5083350791832638e-05,
+      "loss": 2.0461,
+      "step": 5890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5081606914285117e-05,
+      "loss": 1.9877,
+      "step": 5891
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5079862792027034e-05,
+      "loss": 2.0227,
+      "step": 5892
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.507811842510412e-05,
+      "loss": 2.0879,
+      "step": 5893
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.507637381356212e-05,
+      "loss": 2.094,
+      "step": 5894
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5074628957446794e-05,
+      "loss": 2.0662,
+      "step": 5895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5072883856803884e-05,
+      "loss": 2.0625,
+      "step": 5896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5071138511679165e-05,
+      "loss": 2.0684,
+      "step": 5897
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.50693929221184e-05,
+      "loss": 2.0662,
+      "step": 5898
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.5067647088167367e-05,
+      "loss": 2.0294,
+      "step": 5899
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.506590100987185e-05,
+      "loss": 2.0569,
+      "step": 5900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5064154687277635e-05,
+      "loss": 2.082,
+      "step": 5901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5062408120430522e-05,
+      "loss": 2.0886,
+      "step": 5902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5060661309376306e-05,
+      "loss": 2.0898,
+      "step": 5903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5058914254160798e-05,
+      "loss": 2.0688,
+      "step": 5904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5057166954829814e-05,
+      "loss": 2.0168,
+      "step": 5905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5055419411429174e-05,
+      "loss": 1.9648,
+      "step": 5906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5053671624004708e-05,
+      "loss": 2.052,
+      "step": 5907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5051923592602244e-05,
+      "loss": 2.0502,
+      "step": 5908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5050175317267623e-05,
+      "loss": 2.0442,
+      "step": 5909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5048426798046698e-05,
+      "loss": 2.075,
+      "step": 5910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.504667803498532e-05,
+      "loss": 1.9956,
+      "step": 5911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5044929028129338e-05,
+      "loss": 1.9946,
+      "step": 5912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5043179777524628e-05,
+      "loss": 2.1106,
+      "step": 5913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5041430283217063e-05,
+      "loss": 2.0801,
+      "step": 5914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5039680545252512e-05,
+      "loss": 2.0608,
+      "step": 5915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.503793056367687e-05,
+      "loss": 2.0691,
+      "step": 5916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5036180338536024e-05,
+      "loss": 2.1213,
+      "step": 5917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.503442986987587e-05,
+      "loss": 2.0728,
+      "step": 5918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5032679157742317e-05,
+      "loss": 1.9965,
+      "step": 5919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5030928202181267e-05,
+      "loss": 2.0288,
+      "step": 5920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.502917700323865e-05,
+      "loss": 2.114,
+      "step": 5921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5027425560960372e-05,
+      "loss": 2.0854,
+      "step": 5922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5025673875392375e-05,
+      "loss": 2.0737,
+      "step": 5923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5023921946580596e-05,
+      "loss": 2.071,
+      "step": 5924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5022169774570963e-05,
+      "loss": 2.004,
+      "step": 5925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.502041735940944e-05,
+      "loss": 2.0491,
+      "step": 5926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5018664701141978e-05,
+      "loss": 2.0845,
+      "step": 5927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5016911799814534e-05,
+      "loss": 2.1106,
+      "step": 5928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5015158655473084e-05,
+      "loss": 2.1021,
+      "step": 5929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5013405268163588e-05,
+      "loss": 2.084,
+      "step": 5930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.501165163793204e-05,
+      "loss": 2.0691,
+      "step": 5931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5009897764824426e-05,
+      "loss": 2.0518,
+      "step": 5932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.500814364888673e-05,
+      "loss": 2.083,
+      "step": 5933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5006389290164958e-05,
+      "loss": 2.0928,
+      "step": 5934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5004634688705116e-05,
+      "loss": 2.0803,
+      "step": 5935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.5002879844553218e-05,
+      "loss": 2.093,
+      "step": 5936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.500112475775528e-05,
+      "loss": 2.073,
+      "step": 5937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4999369428357326e-05,
+      "loss": 2.0444,
+      "step": 5938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.499761385640539e-05,
+      "loss": 2.0481,
+      "step": 5939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4995858041945506e-05,
+      "loss": 2.0583,
+      "step": 5940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4994101985023726e-05,
+      "loss": 2.0024,
+      "step": 5941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4992345685686094e-05,
+      "loss": 2.0676,
+      "step": 5942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4990589143978668e-05,
+      "loss": 2.0879,
+      "step": 5943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4988832359947513e-05,
+      "loss": 2.0754,
+      "step": 5944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4987075333638694e-05,
+      "loss": 2.0913,
+      "step": 5945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4985318065098292e-05,
+      "loss": 2.0955,
+      "step": 5946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.498356055437239e-05,
+      "loss": 2.0437,
+      "step": 5947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4981802801507072e-05,
+      "loss": 2.0569,
+      "step": 5948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.498004480654844e-05,
+      "loss": 2.0552,
+      "step": 5949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4978286569542585e-05,
+      "loss": 2.0581,
+      "step": 5950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4976528090535622e-05,
+      "loss": 2.0195,
+      "step": 5951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4974769369573664e-05,
+      "loss": 2.0454,
+      "step": 5952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4973010406702833e-05,
+      "loss": 2.0403,
+      "step": 5953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4971251201969252e-05,
+      "loss": 2.0129,
+      "step": 5954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4969491755419054e-05,
+      "loss": 2.0066,
+      "step": 5955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4967732067098385e-05,
+      "loss": 2.031,
+      "step": 5956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4965972137053385e-05,
+      "loss": 2.0671,
+      "step": 5957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4964211965330203e-05,
+      "loss": 2.0681,
+      "step": 5958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4962451551975006e-05,
+      "loss": 2.0486,
+      "step": 5959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.496069089703395e-05,
+      "loss": 2.0669,
+      "step": 5960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.495893000055322e-05,
+      "loss": 2.0615,
+      "step": 5961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4957168862578975e-05,
+      "loss": 2.0784,
+      "step": 5962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4955407483157413e-05,
+      "loss": 2.0486,
+      "step": 5963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4953645862334717e-05,
+      "loss": 2.0449,
+      "step": 5964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4951884000157084e-05,
+      "loss": 2.0779,
+      "step": 5965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.495012189667072e-05,
+      "loss": 2.0442,
+      "step": 5966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4948359551921835e-05,
+      "loss": 2.0891,
+      "step": 5967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4946596965956636e-05,
+      "loss": 2.0742,
+      "step": 5968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4944834138821353e-05,
+      "loss": 2.0391,
+      "step": 5969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4943071070562214e-05,
+      "loss": 2.1074,
+      "step": 5970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.494130776122545e-05,
+      "loss": 2.0073,
+      "step": 5971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4939544210857303e-05,
+      "loss": 2.0195,
+      "step": 5972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.493778041950402e-05,
+      "loss": 2.0046,
+      "step": 5973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4936016387211852e-05,
+      "loss": 2.0505,
+      "step": 5974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4934252114027065e-05,
+      "loss": 2.1028,
+      "step": 5975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.493248759999592e-05,
+      "loss": 2.0903,
+      "step": 5976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.493072284516469e-05,
+      "loss": 2.0718,
+      "step": 5977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4928957849579654e-05,
+      "loss": 2.0776,
+      "step": 5978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4927192613287096e-05,
+      "loss": 2.0767,
+      "step": 5979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.492542713633331e-05,
+      "loss": 2.0786,
+      "step": 5980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.492366141876459e-05,
+      "loss": 2.0735,
+      "step": 5981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4921895460627246e-05,
+      "loss": 2.051,
+      "step": 5982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.492012926196758e-05,
+      "loss": 2.0295,
+      "step": 5983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4918362822831913e-05,
+      "loss": 2.033,
+      "step": 5984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4916596143266567e-05,
+      "loss": 2.0591,
+      "step": 5985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.491482922331787e-05,
+      "loss": 2.04,
+      "step": 5986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4913062063032163e-05,
+      "loss": 2.0786,
+      "step": 5987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4911294662455784e-05,
+      "loss": 2.0896,
+      "step": 5988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4909527021635075e-05,
+      "loss": 2.006,
+      "step": 5989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.49077591406164e-05,
+      "loss": 2.0308,
+      "step": 5990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4905991019446115e-05,
+      "loss": 2.0813,
+      "step": 5991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4904222658170592e-05,
+      "loss": 2.0637,
+      "step": 5992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4902454056836194e-05,
+      "loss": 2.0889,
+      "step": 5993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4900685215489304e-05,
+      "loss": 2.0691,
+      "step": 5994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4898916134176314e-05,
+      "loss": 2.0911,
+      "step": 5995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4897146812943613e-05,
+      "loss": 2.0686,
+      "step": 5996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4895377251837596e-05,
+      "loss": 2.0913,
+      "step": 5997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.489360745090467e-05,
+      "loss": 2.0938,
+      "step": 5998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.489183741019125e-05,
+      "loss": 2.0715,
+      "step": 5999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4890067129743744e-05,
+      "loss": 2.0552,
+      "step": 6000
+    },
+    {
+      "Bleu_1": 0.7711423729623096,
+      "Bleu_2": 0.6096883575087889,
+      "Bleu_3": 0.4672946775280815,
+      "Bleu_4": 0.35414019781135964,
+      "CIDEr": 1.1958407780946136,
+      "METEOR": 0.276749604978806,
+      "ROUGE_L": 0.5566008110285593,
+      "epoch": 0.3,
+      "step": 6000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4888296609608585e-05,
+      "loss": 2.0928,
+      "step": 6001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.48865258498322e-05,
+      "loss": 2.0649,
+      "step": 6002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4884754850461022e-05,
+      "loss": 2.0139,
+      "step": 6003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4882983611541493e-05,
+      "loss": 2.042,
+      "step": 6004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4881212133120065e-05,
+      "loss": 1.9976,
+      "step": 6005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.487944041524319e-05,
+      "loss": 2.019,
+      "step": 6006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.487766845795734e-05,
+      "loss": 2.1157,
+      "step": 6007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.487589626130897e-05,
+      "loss": 2.1057,
+      "step": 6008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4874123825344556e-05,
+      "loss": 2.0481,
+      "step": 6009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4872351150110578e-05,
+      "loss": 2.0859,
+      "step": 6010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4870578235653526e-05,
+      "loss": 1.9872,
+      "step": 6011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.486880508201989e-05,
+      "loss": 2.0071,
+      "step": 6012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4867031689256173e-05,
+      "loss": 2.0759,
+      "step": 6013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4865258057408873e-05,
+      "loss": 2.074,
+      "step": 6014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4863484186524508e-05,
+      "loss": 2.0735,
+      "step": 6015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.486171007664959e-05,
+      "loss": 2.1035,
+      "step": 6016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4859935727830646e-05,
+      "loss": 2.0271,
+      "step": 6017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4858161140114207e-05,
+      "loss": 2.0496,
+      "step": 6018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.485638631354681e-05,
+      "loss": 2.0471,
+      "step": 6019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4854611248174996e-05,
+      "loss": 2.0391,
+      "step": 6020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.485283594404531e-05,
+      "loss": 2.0999,
+      "step": 6021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.485106040120432e-05,
+      "loss": 2.0981,
+      "step": 6022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.484928461969857e-05,
+      "loss": 2.0933,
+      "step": 6023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4847508599574644e-05,
+      "loss": 2.0867,
+      "step": 6024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4845732340879105e-05,
+      "loss": 2.0696,
+      "step": 6025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.484395584365854e-05,
+      "loss": 2.0566,
+      "step": 6026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4842179107959533e-05,
+      "loss": 2.1189,
+      "step": 6027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4840402133828677e-05,
+      "loss": 2.0862,
+      "step": 6028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.483862492131257e-05,
+      "loss": 2.0918,
+      "step": 6029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4836847470457814e-05,
+      "loss": 2.0679,
+      "step": 6030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4835069781311028e-05,
+      "loss": 2.054,
+      "step": 6031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.483329185391883e-05,
+      "loss": 2.092,
+      "step": 6032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.483151368832784e-05,
+      "loss": 2.0449,
+      "step": 6033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4829735284584686e-05,
+      "loss": 2.054,
+      "step": 6034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4827956642736007e-05,
+      "loss": 2.0754,
+      "step": 6035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4826177762828447e-05,
+      "loss": 2.0781,
+      "step": 6036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.482439864490866e-05,
+      "loss": 2.0969,
+      "step": 6037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4822619289023295e-05,
+      "loss": 2.0742,
+      "step": 6038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.482083969521901e-05,
+      "loss": 2.0906,
+      "step": 6039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4819059863542475e-05,
+      "loss": 2.0828,
+      "step": 6040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.481727979404037e-05,
+      "loss": 2.0757,
+      "step": 6041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4815499486759373e-05,
+      "loss": 2.0195,
+      "step": 6042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4813718941746168e-05,
+      "loss": 2.0806,
+      "step": 6043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4811938159047447e-05,
+      "loss": 2.1074,
+      "step": 6044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4810157138709913e-05,
+      "loss": 2.0615,
+      "step": 6045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4808375880780267e-05,
+      "loss": 2.105,
+      "step": 6046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4806594385305226e-05,
+      "loss": 2.0603,
+      "step": 6047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.48048126523315e-05,
+      "loss": 2.0798,
+      "step": 6048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4803030681905817e-05,
+      "loss": 2.1055,
+      "step": 6049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.480124847407491e-05,
+      "loss": 2.0852,
+      "step": 6050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.479946602888551e-05,
+      "loss": 2.103,
+      "step": 6051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4797683346384357e-05,
+      "loss": 2.104,
+      "step": 6052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4795900426618214e-05,
+      "loss": 2.094,
+      "step": 6053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.479411726963382e-05,
+      "loss": 2.0903,
+      "step": 6054
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4792333875477947e-05,
+      "loss": 2.0481,
+      "step": 6055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4790550244197355e-05,
+      "loss": 2.0249,
+      "step": 6056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4788766375838818e-05,
+      "loss": 2.0305,
+      "step": 6057
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4786982270449125e-05,
+      "loss": 2.0366,
+      "step": 6058
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4785197928075053e-05,
+      "loss": 2.105,
+      "step": 6059
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.47834133487634e-05,
+      "loss": 2.082,
+      "step": 6060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4781628532560958e-05,
+      "loss": 2.0217,
+      "step": 6061
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4779843479514537e-05,
+      "loss": 2.02,
+      "step": 6062
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4778058189670946e-05,
+      "loss": 2.0808,
+      "step": 6063
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4776272663077004e-05,
+      "loss": 2.0967,
+      "step": 6064
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.477448689977953e-05,
+      "loss": 2.03,
+      "step": 6065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4772700899825357e-05,
+      "loss": 2.0444,
+      "step": 6066
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.477091466326132e-05,
+      "loss": 2.0886,
+      "step": 6067
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4769128190134265e-05,
+      "loss": 2.093,
+      "step": 6068
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4767341480491035e-05,
+      "loss": 2.094,
+      "step": 6069
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4765554534378483e-05,
+      "loss": 2.0828,
+      "step": 6070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4763767351843474e-05,
+      "loss": 2.0701,
+      "step": 6071
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4761979932932876e-05,
+      "loss": 2.0945,
+      "step": 6072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4760192277693557e-05,
+      "loss": 2.1401,
+      "step": 6073
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.47584043861724e-05,
+      "loss": 2.125,
+      "step": 6074
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.475661625841628e-05,
+      "loss": 2.0918,
+      "step": 6075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.475482789447211e-05,
+      "loss": 2.0669,
+      "step": 6076
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4753039294386767e-05,
+      "loss": 2.103,
+      "step": 6077
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4751250458207167e-05,
+      "loss": 2.0957,
+      "step": 6078
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4749461385980213e-05,
+      "loss": 2.1121,
+      "step": 6079
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.474767207775283e-05,
+      "loss": 2.054,
+      "step": 6080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.474588253357193e-05,
+      "loss": 2.0239,
+      "step": 6081
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4744092753484447e-05,
+      "loss": 2.0576,
+      "step": 6082
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4742302737537317e-05,
+      "loss": 2.1064,
+      "step": 6083
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4740512485777475e-05,
+      "loss": 2.0862,
+      "step": 6084
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4738721998251877e-05,
+      "loss": 2.0613,
+      "step": 6085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4736931275007476e-05,
+      "loss": 2.0503,
+      "step": 6086
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4735140316091228e-05,
+      "loss": 2.0898,
+      "step": 6087
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4733349121550092e-05,
+      "loss": 2.1084,
+      "step": 6088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4731557691431055e-05,
+      "loss": 2.1455,
+      "step": 6089
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4729766025781083e-05,
+      "loss": 2.1182,
+      "step": 6090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4727974124647166e-05,
+      "loss": 2.0896,
+      "step": 6091
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.472618198807629e-05,
+      "loss": 2.0957,
+      "step": 6092
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.472438961611546e-05,
+      "loss": 2.0327,
+      "step": 6093
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4722597008811668e-05,
+      "loss": 2.032,
+      "step": 6094
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.472080416621193e-05,
+      "loss": 2.0964,
+      "step": 6095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4719011088363265e-05,
+      "loss": 2.0974,
+      "step": 6096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.471721777531268e-05,
+      "loss": 2.0542,
+      "step": 6097
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.471542422710722e-05,
+      "loss": 2.0608,
+      "step": 6098
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.4713630443793906e-05,
+      "loss": 2.0686,
+      "step": 6099
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.471183642541978e-05,
+      "loss": 2.0601,
+      "step": 6100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4710042172031897e-05,
+      "loss": 2.0676,
+      "step": 6101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4708247683677304e-05,
+      "loss": 2.0818,
+      "step": 6102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.470645296040305e-05,
+      "loss": 2.0916,
+      "step": 6103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4704658002256217e-05,
+      "loss": 2.063,
+      "step": 6104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.470286280928386e-05,
+      "loss": 2.0696,
+      "step": 6105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4701067381533065e-05,
+      "loss": 2.074,
+      "step": 6106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.469927171905091e-05,
+      "loss": 2.0471,
+      "step": 6107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4697475821884485e-05,
+      "loss": 2.0715,
+      "step": 6108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.469567969008089e-05,
+      "loss": 2.0581,
+      "step": 6109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.469388332368722e-05,
+      "loss": 2.0928,
+      "step": 6110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.469208672275059e-05,
+      "loss": 2.0193,
+      "step": 6111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4690289887318105e-05,
+      "loss": 2.0129,
+      "step": 6112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.468849281743689e-05,
+      "loss": 2.0574,
+      "step": 6113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4686695513154076e-05,
+      "loss": 2.0627,
+      "step": 6114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4684897974516786e-05,
+      "loss": 2.0286,
+      "step": 6115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4683100201572158e-05,
+      "loss": 2.03,
+      "step": 6116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.468130219436735e-05,
+      "loss": 2.0417,
+      "step": 6117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.46795039529495e-05,
+      "loss": 2.0801,
+      "step": 6118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.467770547736576e-05,
+      "loss": 2.0447,
+      "step": 6119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.467590676766331e-05,
+      "loss": 2.0613,
+      "step": 6120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4674107823889304e-05,
+      "loss": 2.0508,
+      "step": 6121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4672308646090928e-05,
+      "loss": 2.0542,
+      "step": 6122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4670509234315356e-05,
+      "loss": 2.062,
+      "step": 6123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4668709588609783e-05,
+      "loss": 2.0327,
+      "step": 6124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4666909709021394e-05,
+      "loss": 2.0874,
+      "step": 6125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4665109595597393e-05,
+      "loss": 2.0869,
+      "step": 6126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.466330924838498e-05,
+      "loss": 2.0417,
+      "step": 6127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4661508667431377e-05,
+      "loss": 2.0217,
+      "step": 6128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4659707852783802e-05,
+      "loss": 2.0845,
+      "step": 6129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4657906804489467e-05,
+      "loss": 2.0674,
+      "step": 6130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4656105522595615e-05,
+      "loss": 2.0571,
+      "step": 6131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4654304007149476e-05,
+      "loss": 2.075,
+      "step": 6132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4652502258198296e-05,
+      "loss": 2.0476,
+      "step": 6133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4650700275789324e-05,
+      "loss": 2.0947,
+      "step": 6134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4648898059969806e-05,
+      "loss": 2.043,
+      "step": 6135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4647095610787018e-05,
+      "loss": 2.0342,
+      "step": 6136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4645292928288212e-05,
+      "loss": 2.0898,
+      "step": 6137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4643490012520673e-05,
+      "loss": 2.0979,
+      "step": 6138
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4641686863531675e-05,
+      "loss": 2.1123,
+      "step": 6139
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4639883481368507e-05,
+      "loss": 2.1196,
+      "step": 6140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4638079866078456e-05,
+      "loss": 2.0527,
+      "step": 6141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.463627601770882e-05,
+      "loss": 2.04,
+      "step": 6142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4634471936306906e-05,
+      "loss": 2.0356,
+      "step": 6143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4632667621920018e-05,
+      "loss": 2.0498,
+      "step": 6144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4630863074595485e-05,
+      "loss": 2.0859,
+      "step": 6145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4629058294380615e-05,
+      "loss": 2.0496,
+      "step": 6146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.462725328132274e-05,
+      "loss": 2.1304,
+      "step": 6147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4625448035469204e-05,
+      "loss": 2.1252,
+      "step": 6148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4623642556867334e-05,
+      "loss": 2.0693,
+      "step": 6149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4621836845564484e-05,
+      "loss": 2.0391,
+      "step": 6150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4620030901608005e-05,
+      "loss": 2.0203,
+      "step": 6151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.461822472504526e-05,
+      "loss": 2.0579,
+      "step": 6152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.46164183159236e-05,
+      "loss": 2.0884,
+      "step": 6153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4614611674290413e-05,
+      "loss": 2.0857,
+      "step": 6154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.461280480019307e-05,
+      "loss": 2.0574,
+      "step": 6155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4610997693678947e-05,
+      "loss": 2.04,
+      "step": 6156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.460919035479544e-05,
+      "loss": 2.0825,
+      "step": 6157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4607382783589947e-05,
+      "loss": 2.1047,
+      "step": 6158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4605574980109866e-05,
+      "loss": 2.0645,
+      "step": 6159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4603766944402603e-05,
+      "loss": 2.0696,
+      "step": 6160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4601958676515573e-05,
+      "loss": 2.0818,
+      "step": 6161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.46001501764962e-05,
+      "loss": 2.0911,
+      "step": 6162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4598341444391902e-05,
+      "loss": 2.0645,
+      "step": 6163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4596532480250115e-05,
+      "loss": 2.064,
+      "step": 6164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.459472328411828e-05,
+      "loss": 2.0366,
+      "step": 6165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.459291385604383e-05,
+      "loss": 2.0549,
+      "step": 6166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.459110419607423e-05,
+      "loss": 2.0154,
+      "step": 6167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4589294304256924e-05,
+      "loss": 1.9915,
+      "step": 6168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4587484180639386e-05,
+      "loss": 2.0828,
+      "step": 6169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4585673825269073e-05,
+      "loss": 2.083,
+      "step": 6170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4583863238193463e-05,
+      "loss": 2.0496,
+      "step": 6171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.458205241946004e-05,
+      "loss": 2.0471,
+      "step": 6172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4580241369116292e-05,
+      "loss": 2.0898,
+      "step": 6173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4578430087209703e-05,
+      "loss": 2.0691,
+      "step": 6174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4576618573787783e-05,
+      "loss": 2.0457,
+      "step": 6175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4574806828898023e-05,
+      "loss": 2.0537,
+      "step": 6176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4572994852587945e-05,
+      "loss": 2.104,
+      "step": 6177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4571182644905065e-05,
+      "loss": 2.0654,
+      "step": 6178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.45693702058969e-05,
+      "loss": 2.1165,
+      "step": 6179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4567557535610985e-05,
+      "loss": 2.0933,
+      "step": 6180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4565744634094852e-05,
+      "loss": 2.0784,
+      "step": 6181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4563931501396046e-05,
+      "loss": 2.0708,
+      "step": 6182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4562118137562107e-05,
+      "loss": 2.0424,
+      "step": 6183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4560304542640596e-05,
+      "loss": 2.0649,
+      "step": 6184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.455849071667907e-05,
+      "loss": 2.0139,
+      "step": 6185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.455667665972509e-05,
+      "loss": 2.0391,
+      "step": 6186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4554862371826238e-05,
+      "loss": 2.0908,
+      "step": 6187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4553047853030078e-05,
+      "loss": 2.0603,
+      "step": 6188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4551233103384203e-05,
+      "loss": 2.1042,
+      "step": 6189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4549418122936202e-05,
+      "loss": 2.0615,
+      "step": 6190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4547602911733668e-05,
+      "loss": 2.052,
+      "step": 6191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4545787469824204e-05,
+      "loss": 2.0793,
+      "step": 6192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4543971797255414e-05,
+      "loss": 2.1001,
+      "step": 6193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4542155894074918e-05,
+      "loss": 2.1226,
+      "step": 6194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.454033976033034e-05,
+      "loss": 2.0857,
+      "step": 6195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4538523396069294e-05,
+      "loss": 2.0718,
+      "step": 6196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4536706801339414e-05,
+      "loss": 2.1028,
+      "step": 6197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.453488997618835e-05,
+      "loss": 2.1106,
+      "step": 6198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.453307292066373e-05,
+      "loss": 2.0466,
+      "step": 6199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4531255634813213e-05,
+      "loss": 2.033,
+      "step": 6200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4529438118684457e-05,
+      "loss": 2.0488,
+      "step": 6201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4527620372325123e-05,
+      "loss": 2.0825,
+      "step": 6202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4525802395782874e-05,
+      "loss": 2.0525,
+      "step": 6203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.452398418910539e-05,
+      "loss": 2.0442,
+      "step": 6204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.452216575234035e-05,
+      "loss": 1.9849,
+      "step": 6205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.452034708553544e-05,
+      "loss": 2.0103,
+      "step": 6206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.451852818873835e-05,
+      "loss": 2.0732,
+      "step": 6207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4516709061996784e-05,
+      "loss": 2.0781,
+      "step": 6208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4514889705358438e-05,
+      "loss": 2.0447,
+      "step": 6209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.451307011887103e-05,
+      "loss": 2.0449,
+      "step": 6210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.451125030258228e-05,
+      "loss": 1.9651,
+      "step": 6211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.45094302565399e-05,
+      "loss": 2.0173,
+      "step": 6212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4507609980791623e-05,
+      "loss": 2.0212,
+      "step": 6213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4505789475385185e-05,
+      "loss": 2.0229,
+      "step": 6214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4503968740368325e-05,
+      "loss": 2.0808,
+      "step": 6215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.450214777578879e-05,
+      "loss": 2.0645,
+      "step": 6216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4500326581694335e-05,
+      "loss": 2.0933,
+      "step": 6217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.449850515813272e-05,
+      "loss": 2.074,
+      "step": 6218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.44966835051517e-05,
+      "loss": 2.0759,
+      "step": 6219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.449486162279905e-05,
+      "loss": 2.0403,
+      "step": 6220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4493039511122558e-05,
+      "loss": 2.1104,
+      "step": 6221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4491217170169993e-05,
+      "loss": 2.1399,
+      "step": 6222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4489394599989147e-05,
+      "loss": 2.0857,
+      "step": 6223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4487571800627817e-05,
+      "loss": 2.0864,
+      "step": 6224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4485748772133804e-05,
+      "loss": 2.0823,
+      "step": 6225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4483925514554917e-05,
+      "loss": 2.1143,
+      "step": 6226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4482102027938964e-05,
+      "loss": 2.1174,
+      "step": 6227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.448027831233376e-05,
+      "loss": 2.1169,
+      "step": 6228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4478454367787144e-05,
+      "loss": 2.0325,
+      "step": 6229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4476630194346933e-05,
+      "loss": 2.0032,
+      "step": 6230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.447480579206097e-05,
+      "loss": 2.0867,
+      "step": 6231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4472981160977096e-05,
+      "loss": 2.0574,
+      "step": 6232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.447115630114316e-05,
+      "loss": 1.9871,
+      "step": 6233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4469331212607023e-05,
+      "loss": 2.0046,
+      "step": 6234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4467505895416533e-05,
+      "loss": 2.0452,
+      "step": 6235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4465680349619568e-05,
+      "loss": 2.0476,
+      "step": 6236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4463854575263997e-05,
+      "loss": 2.0347,
+      "step": 6237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.44620285723977e-05,
+      "loss": 2.0476,
+      "step": 6238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.446020234106856e-05,
+      "loss": 2.0503,
+      "step": 6239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4458375881324464e-05,
+      "loss": 2.1167,
+      "step": 6240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4456549193213317e-05,
+      "loss": 2.0728,
+      "step": 6241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.445472227678302e-05,
+      "loss": 2.0632,
+      "step": 6242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4452895132081477e-05,
+      "loss": 2.0801,
+      "step": 6243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4451067759156603e-05,
+      "loss": 2.0881,
+      "step": 6244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4449240158056327e-05,
+      "loss": 2.0457,
+      "step": 6245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4447412328828567e-05,
+      "loss": 2.0688,
+      "step": 6246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.444558427152126e-05,
+      "loss": 2.0288,
+      "step": 6247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4443755986182342e-05,
+      "loss": 2.033,
+      "step": 6248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.444192747285976e-05,
+      "loss": 2.1265,
+      "step": 6249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.444009873160146e-05,
+      "loss": 2.125,
+      "step": 6250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4438269762455408e-05,
+      "loss": 2.0264,
+      "step": 6251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.443644056546956e-05,
+      "loss": 2.0259,
+      "step": 6252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.443461114069188e-05,
+      "loss": 2.0427,
+      "step": 6253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4432781488170352e-05,
+      "loss": 2.0786,
+      "step": 6254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.443095160795295e-05,
+      "loss": 2.1064,
+      "step": 6255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.442912150008767e-05,
+      "loss": 2.0801,
+      "step": 6256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.442729116462249e-05,
+      "loss": 2.1101,
+      "step": 6257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4425460601605418e-05,
+      "loss": 2.0874,
+      "step": 6258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4423629811084456e-05,
+      "loss": 2.0029,
+      "step": 6259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4421798793107616e-05,
+      "loss": 2.0576,
+      "step": 6260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4419967547722915e-05,
+      "loss": 2.04,
+      "step": 6261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4418136074978368e-05,
+      "loss": 2.0664,
+      "step": 6262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.441630437492201e-05,
+      "loss": 2.0911,
+      "step": 6263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4414472447601877e-05,
+      "loss": 2.1021,
+      "step": 6264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4412640293065997e-05,
+      "loss": 2.0137,
+      "step": 6265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4410807911362436e-05,
+      "loss": 2.0381,
+      "step": 6266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.440897530253923e-05,
+      "loss": 1.98,
+      "step": 6267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.440714246664444e-05,
+      "loss": 2.0254,
+      "step": 6268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4405309403726138e-05,
+      "loss": 2.0706,
+      "step": 6269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4403476113832384e-05,
+      "loss": 2.0803,
+      "step": 6270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4401642597011253e-05,
+      "loss": 2.0793,
+      "step": 6271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4399808853310842e-05,
+      "loss": 2.0964,
+      "step": 6272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4397974882779223e-05,
+      "loss": 2.095,
+      "step": 6273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4396140685464497e-05,
+      "loss": 2.0789,
+      "step": 6274
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.439430626141476e-05,
+      "loss": 2.0515,
+      "step": 6275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.439247161067812e-05,
+      "loss": 2.0608,
+      "step": 6276
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.439063673330269e-05,
+      "loss": 2.0625,
+      "step": 6277
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4388801629336585e-05,
+      "loss": 2.0801,
+      "step": 6278
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4386966298827927e-05,
+      "loss": 1.9541,
+      "step": 6279
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.438513074182485e-05,
+      "loss": 1.97,
+      "step": 6280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4383294958375483e-05,
+      "loss": 2.0552,
+      "step": 6281
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.438145894852798e-05,
+      "loss": 2.0444,
+      "step": 6282
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4379622712330473e-05,
+      "loss": 2.0823,
+      "step": 6283
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.437778624983112e-05,
+      "loss": 2.0876,
+      "step": 6284
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.437594956107809e-05,
+      "loss": 2.054,
+      "step": 6285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4374112646119535e-05,
+      "loss": 2.093,
+      "step": 6286
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.437227550500363e-05,
+      "loss": 2.073,
+      "step": 6287
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4370438137778552e-05,
+      "loss": 2.0579,
+      "step": 6288
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.436860054449249e-05,
+      "loss": 2.0537,
+      "step": 6289
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.436676272519362e-05,
+      "loss": 2.0237,
+      "step": 6290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.436492467993015e-05,
+      "loss": 2.0454,
+      "step": 6291
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.436308640875027e-05,
+      "loss": 2.01,
+      "step": 6292
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4361247911702197e-05,
+      "loss": 2.0642,
+      "step": 6293
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.435940918883413e-05,
+      "loss": 2.0793,
+      "step": 6294
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4357570240194302e-05,
+      "loss": 2.1099,
+      "step": 6295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4355731065830922e-05,
+      "loss": 2.1287,
+      "step": 6296
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4353891665792236e-05,
+      "loss": 2.0891,
+      "step": 6297
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4352052040126467e-05,
+      "loss": 2.083,
+      "step": 6298
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.4350212188881865e-05,
+      "loss": 2.052,
+      "step": 6299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.434837211210668e-05,
+      "loss": 2.0654,
+      "step": 6300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4346531809849156e-05,
+      "loss": 2.1125,
+      "step": 6301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.434469128215756e-05,
+      "loss": 2.0913,
+      "step": 6302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4342850529080157e-05,
+      "loss": 2.1204,
+      "step": 6303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.434100955066521e-05,
+      "loss": 2.1433,
+      "step": 6304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.433916834696101e-05,
+      "loss": 2.0942,
+      "step": 6305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4337326918015836e-05,
+      "loss": 2.0654,
+      "step": 6306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.433548526387798e-05,
+      "loss": 1.9993,
+      "step": 6307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4333643384595724e-05,
+      "loss": 2.0183,
+      "step": 6308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4331801280217383e-05,
+      "loss": 2.0696,
+      "step": 6309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4329958950791255e-05,
+      "loss": 2.0833,
+      "step": 6310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4328116396365658e-05,
+      "loss": 2.0635,
+      "step": 6311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.432627361698892e-05,
+      "loss": 2.0312,
+      "step": 6312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4324430612709342e-05,
+      "loss": 2.1399,
+      "step": 6313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4322587383575276e-05,
+      "loss": 2.0857,
+      "step": 6314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4320743929635048e-05,
+      "loss": 2.0464,
+      "step": 6315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4318900250937006e-05,
+      "loss": 2.0298,
+      "step": 6316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4317056347529495e-05,
+      "loss": 2.0466,
+      "step": 6317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4315212219460873e-05,
+      "loss": 2.0315,
+      "step": 6318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.431336786677949e-05,
+      "loss": 2.0488,
+      "step": 6319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4311523289533726e-05,
+      "loss": 2.0757,
+      "step": 6320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4309678487771948e-05,
+      "loss": 2.1182,
+      "step": 6321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4307833461542525e-05,
+      "loss": 2.0845,
+      "step": 6322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.430598821089385e-05,
+      "loss": 1.9988,
+      "step": 6323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4304142735874308e-05,
+      "loss": 2.0605,
+      "step": 6324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.43022970365323e-05,
+      "loss": 2.084,
+      "step": 6325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4300451112916222e-05,
+      "loss": 2.0764,
+      "step": 6326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4298604965074485e-05,
+      "loss": 2.0786,
+      "step": 6327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4296758593055498e-05,
+      "loss": 2.0764,
+      "step": 6328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4294911996907682e-05,
+      "loss": 2.073,
+      "step": 6329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4293065176679462e-05,
+      "loss": 2.0779,
+      "step": 6330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.429121813241927e-05,
+      "loss": 2.0271,
+      "step": 6331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.428937086417554e-05,
+      "loss": 2.073,
+      "step": 6332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4287523371996718e-05,
+      "loss": 2.0879,
+      "step": 6333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4285675655931247e-05,
+      "loss": 2.094,
+      "step": 6334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4283827716027586e-05,
+      "loss": 2.083,
+      "step": 6335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.428197955233419e-05,
+      "loss": 2.0813,
+      "step": 6336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4280131164899534e-05,
+      "loss": 2.1055,
+      "step": 6337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4278282553772084e-05,
+      "loss": 2.0598,
+      "step": 6338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4276433719000313e-05,
+      "loss": 2.041,
+      "step": 6339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.427458466063271e-05,
+      "loss": 2.0295,
+      "step": 6340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.427273537871777e-05,
+      "loss": 2.0762,
+      "step": 6341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4270885873303973e-05,
+      "loss": 2.0735,
+      "step": 6342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4269036144439835e-05,
+      "loss": 2.0916,
+      "step": 6343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4267186192173858e-05,
+      "loss": 2.0798,
+      "step": 6344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.426533601655455e-05,
+      "loss": 2.0657,
+      "step": 6345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4263485617630433e-05,
+      "loss": 2.0574,
+      "step": 6346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4261634995450038e-05,
+      "loss": 2.0549,
+      "step": 6347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.425978415006189e-05,
+      "loss": 2.0649,
+      "step": 6348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.425793308151452e-05,
+      "loss": 2.1033,
+      "step": 6349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4256081789856476e-05,
+      "loss": 2.1121,
+      "step": 6350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4254230275136306e-05,
+      "loss": 2.0535,
+      "step": 6351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4252378537402562e-05,
+      "loss": 2.0291,
+      "step": 6352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4250526576703804e-05,
+      "loss": 2.0652,
+      "step": 6353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.42486743930886e-05,
+      "loss": 2.0869,
+      "step": 6354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.424682198660552e-05,
+      "loss": 2.0007,
+      "step": 6355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.424496935730314e-05,
+      "loss": 2.0249,
+      "step": 6356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4243116505230042e-05,
+      "loss": 2.0889,
+      "step": 6357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.424126343043482e-05,
+      "loss": 2.0557,
+      "step": 6358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.423941013296606e-05,
+      "loss": 2.1182,
+      "step": 6359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4237556612872372e-05,
+      "loss": 2.0964,
+      "step": 6360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4235702870202356e-05,
+      "loss": 2.0903,
+      "step": 6361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4233848905004627e-05,
+      "loss": 2.0759,
+      "step": 6362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.423199471732781e-05,
+      "loss": 2.1121,
+      "step": 6363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4230140307220513e-05,
+      "loss": 2.1072,
+      "step": 6364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4228285674731374e-05,
+      "loss": 2.1045,
+      "step": 6365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4226430819909036e-05,
+      "loss": 2.0798,
+      "step": 6366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4224575742802125e-05,
+      "loss": 2.0144,
+      "step": 6367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4222720443459304e-05,
+      "loss": 1.989,
+      "step": 6368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4220864921929215e-05,
+      "loss": 2.0442,
+      "step": 6369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.421900917826052e-05,
+      "loss": 2.0493,
+      "step": 6370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4217153212501888e-05,
+      "loss": 2.0491,
+      "step": 6371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4215297024701978e-05,
+      "loss": 2.0452,
+      "step": 6372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4213440614909484e-05,
+      "loss": 2.061,
+      "step": 6373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.421158398317307e-05,
+      "loss": 2.0771,
+      "step": 6374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4209727129541438e-05,
+      "loss": 2.0598,
+      "step": 6375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4207870054063274e-05,
+      "loss": 2.053,
+      "step": 6376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.420601275678728e-05,
+      "loss": 2.0686,
+      "step": 6377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4204155237762163e-05,
+      "loss": 2.0818,
+      "step": 6378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.420229749703663e-05,
+      "loss": 2.0911,
+      "step": 6379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4200439534659407e-05,
+      "loss": 2.0847,
+      "step": 6380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4198581350679203e-05,
+      "loss": 2.0901,
+      "step": 6381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4196722945144755e-05,
+      "loss": 2.0811,
+      "step": 6382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4194864318104805e-05,
+      "loss": 2.0847,
+      "step": 6383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.419300546960808e-05,
+      "loss": 2.0454,
+      "step": 6384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4191146399703327e-05,
+      "loss": 2.0486,
+      "step": 6385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4189287108439306e-05,
+      "loss": 2.0508,
+      "step": 6386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4187427595864773e-05,
+      "loss": 2.0571,
+      "step": 6387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4185567862028487e-05,
+      "loss": 2.0525,
+      "step": 6388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4183707906979223e-05,
+      "loss": 2.0437,
+      "step": 6389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4181847730765754e-05,
+      "loss": 2.0415,
+      "step": 6390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4179987333436858e-05,
+      "loss": 2.0217,
+      "step": 6391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4178126715041324e-05,
+      "loss": 2.0708,
+      "step": 6392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.417626587562794e-05,
+      "loss": 2.0613,
+      "step": 6393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4174404815245513e-05,
+      "loss": 2.0327,
+      "step": 6394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4172543533942845e-05,
+      "loss": 2.075,
+      "step": 6395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.417068203176874e-05,
+      "loss": 2.0928,
+      "step": 6396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.416882030877202e-05,
+      "loss": 2.1292,
+      "step": 6397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4166958365001502e-05,
+      "loss": 2.1348,
+      "step": 6398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.416509620050602e-05,
+      "loss": 2.0884,
+      "step": 6399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.41632338153344e-05,
+      "loss": 2.0776,
+      "step": 6400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.416137120953548e-05,
+      "loss": 2.0615,
+      "step": 6401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4159508383158115e-05,
+      "loss": 2.0803,
+      "step": 6402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4157645336251142e-05,
+      "loss": 2.0747,
+      "step": 6403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4155782068863427e-05,
+      "loss": 2.0703,
+      "step": 6404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.415391858104383e-05,
+      "loss": 2.0776,
+      "step": 6405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4152054872841217e-05,
+      "loss": 2.0962,
+      "step": 6406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.415019094430446e-05,
+      "loss": 2.0889,
+      "step": 6407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4148326795482442e-05,
+      "loss": 2.1013,
+      "step": 6408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4146462426424052e-05,
+      "loss": 2.0603,
+      "step": 6409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4144597837178167e-05,
+      "loss": 2.0627,
+      "step": 6410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4142733027793698e-05,
+      "loss": 2.0732,
+      "step": 6411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.414086799831954e-05,
+      "loss": 2.0593,
+      "step": 6412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4139002748804602e-05,
+      "loss": 2.1187,
+      "step": 6413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4137137279297802e-05,
+      "loss": 2.1072,
+      "step": 6414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.413527158984806e-05,
+      "loss": 2.0376,
+      "step": 6415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.413340568050429e-05,
+      "loss": 2.0149,
+      "step": 6416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4131539551315437e-05,
+      "loss": 2.095,
+      "step": 6417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4129673202330428e-05,
+      "loss": 2.0645,
+      "step": 6418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4127806633598218e-05,
+      "loss": 2.0857,
+      "step": 6419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4125939845167742e-05,
+      "loss": 2.0571,
+      "step": 6420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4124072837087962e-05,
+      "loss": 2.0708,
+      "step": 6421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4122205609407837e-05,
+      "loss": 2.0422,
+      "step": 6422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4120338162176337e-05,
+      "loss": 2.0637,
+      "step": 6423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.411847049544243e-05,
+      "loss": 2.0513,
+      "step": 6424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4116602609255086e-05,
+      "loss": 2.0627,
+      "step": 6425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4114734503663295e-05,
+      "loss": 2.0583,
+      "step": 6426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.411286617871605e-05,
+      "loss": 2.0898,
+      "step": 6427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4110997634462337e-05,
+      "loss": 2.0911,
+      "step": 6428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.410912887095116e-05,
+      "loss": 2.0488,
+      "step": 6429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4107259888231532e-05,
+      "loss": 2.0371,
+      "step": 6430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4105390686352455e-05,
+      "loss": 2.0981,
+      "step": 6431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4103521265362952e-05,
+      "loss": 2.082,
+      "step": 6432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.410165162531204e-05,
+      "loss": 2.0586,
+      "step": 6433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.409978176624876e-05,
+      "loss": 2.0403,
+      "step": 6434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4097911688222135e-05,
+      "loss": 2.0889,
+      "step": 6435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4096041391281215e-05,
+      "loss": 2.0852,
+      "step": 6436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4094170875475035e-05,
+      "loss": 2.1279,
+      "step": 6437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4092300140852658e-05,
+      "loss": 2.1233,
+      "step": 6438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.409042918746314e-05,
+      "loss": 2.1077,
+      "step": 6439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4088558015355542e-05,
+      "loss": 2.0901,
+      "step": 6440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4086686624578927e-05,
+      "loss": 2.0864,
+      "step": 6441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4084815015182382e-05,
+      "loss": 2.0732,
+      "step": 6442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4082943187214983e-05,
+      "loss": 2.0146,
+      "step": 6443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4081071140725816e-05,
+      "loss": 2.0176,
+      "step": 6444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.407919887576397e-05,
+      "loss": 2.0002,
+      "step": 6445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.407732639237855e-05,
+      "loss": 2.0244,
+      "step": 6446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4075453690618656e-05,
+      "loss": 2.0754,
+      "step": 6447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4073580770533395e-05,
+      "loss": 2.0696,
+      "step": 6448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4071707632171888e-05,
+      "loss": 2.0884,
+      "step": 6449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.406983427558325e-05,
+      "loss": 2.072,
+      "step": 6450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4067960700816613e-05,
+      "loss": 2.1233,
+      "step": 6451
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4066086907921102e-05,
+      "loss": 2.0928,
+      "step": 6452
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.406421289694586e-05,
+      "loss": 2.0776,
+      "step": 6453
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4062338667940034e-05,
+      "loss": 2.0933,
+      "step": 6454
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.406046422095277e-05,
+      "loss": 2.0327,
+      "step": 6455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4058589556033213e-05,
+      "loss": 2.0461,
+      "step": 6456
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4056714673230543e-05,
+      "loss": 2.0562,
+      "step": 6457
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4054839572593915e-05,
+      "loss": 2.0505,
+      "step": 6458
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4052964254172503e-05,
+      "loss": 2.092,
+      "step": 6459
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.405108871801549e-05,
+      "loss": 2.0776,
+      "step": 6460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4049212964172054e-05,
+      "loss": 2.095,
+      "step": 6461
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4047336992691384e-05,
+      "loss": 2.0964,
+      "step": 6462
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4045460803622677e-05,
+      "loss": 2.0996,
+      "step": 6463
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4043584397015138e-05,
+      "loss": 2.0706,
+      "step": 6464
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4041707772917964e-05,
+      "loss": 2.0569,
+      "step": 6465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4039830931380378e-05,
+      "loss": 2.0669,
+      "step": 6466
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.403795387245159e-05,
+      "loss": 2.0701,
+      "step": 6467
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4036076596180822e-05,
+      "loss": 2.0894,
+      "step": 6468
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4034199102617318e-05,
+      "loss": 1.9985,
+      "step": 6469
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4032321391810297e-05,
+      "loss": 2.0244,
+      "step": 6470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4030443463809008e-05,
+      "loss": 2.0825,
+      "step": 6471
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4028565318662695e-05,
+      "loss": 2.0525,
+      "step": 6472
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.402668695642061e-05,
+      "loss": 2.0842,
+      "step": 6473
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4024808377132014e-05,
+      "loss": 2.0857,
+      "step": 6474
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4022929580846165e-05,
+      "loss": 2.0415,
+      "step": 6475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4021050567612336e-05,
+      "loss": 2.0562,
+      "step": 6476
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.40191713374798e-05,
+      "loss": 2.0291,
+      "step": 6477
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.401729189049784e-05,
+      "loss": 1.9949,
+      "step": 6478
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4015412226715744e-05,
+      "loss": 2.0469,
+      "step": 6479
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4013532346182796e-05,
+      "loss": 2.0728,
+      "step": 6480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.40116522489483e-05,
+      "loss": 2.0581,
+      "step": 6481
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4009771935061564e-05,
+      "loss": 2.0415,
+      "step": 6482
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4007891404571883e-05,
+      "loss": 2.0593,
+      "step": 6483
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4006010657528584e-05,
+      "loss": 2.0417,
+      "step": 6484
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.4004129693980987e-05,
+      "loss": 2.1016,
+      "step": 6485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.400224851397841e-05,
+      "loss": 2.1167,
+      "step": 6486
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.400036711757019e-05,
+      "loss": 2.0122,
+      "step": 6487
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3998485504805663e-05,
+      "loss": 2.0227,
+      "step": 6488
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3996603675734172e-05,
+      "loss": 2.0918,
+      "step": 6489
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3994721630405074e-05,
+      "loss": 2.0647,
+      "step": 6490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.399283936886771e-05,
+      "loss": 2.1238,
+      "step": 6491
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3990956891171442e-05,
+      "loss": 2.1272,
+      "step": 6492
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3989074197365645e-05,
+      "loss": 2.0496,
+      "step": 6493
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3987191287499688e-05,
+      "loss": 2.0649,
+      "step": 6494
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3985308161622944e-05,
+      "loss": 2.0752,
+      "step": 6495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3983424819784792e-05,
+      "loss": 2.085,
+      "step": 6496
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.398154126203463e-05,
+      "loss": 2.0361,
+      "step": 6497
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.397965748842185e-05,
+      "loss": 2.0305,
+      "step": 6498
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.3977773498995846e-05,
+      "loss": 2.0684,
+      "step": 6499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.397588929380603e-05,
+      "loss": 2.0437,
+      "step": 6500
+    },
+    {
+      "Bleu_1": 0.7737048918915991,
+      "Bleu_2": 0.6133997960711146,
+      "Bleu_3": 0.4715725923057629,
+      "Bleu_4": 0.35862465095996005,
+      "CIDEr": 1.2077309124014373,
+      "METEOR": 0.27913622121520965,
+      "ROUGE_L": 0.5598123387021305,
+      "epoch": 0.33,
+      "step": 6500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3974004872901807e-05,
+      "loss": 2.1042,
+      "step": 6501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.39721202363326e-05,
+      "loss": 2.074,
+      "step": 6502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3970235384147824e-05,
+      "loss": 2.1211,
+      "step": 6503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3968350316396915e-05,
+      "loss": 2.0986,
+      "step": 6504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3966465033129308e-05,
+      "loss": 2.0386,
+      "step": 6505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.396457953439443e-05,
+      "loss": 2.0386,
+      "step": 6506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3962693820241737e-05,
+      "loss": 2.135,
+      "step": 6507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3960807890720676e-05,
+      "loss": 2.1187,
+      "step": 6508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.39589217458807e-05,
+      "loss": 2.0347,
+      "step": 6509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.395703538577128e-05,
+      "loss": 2.0181,
+      "step": 6510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.395514881044187e-05,
+      "loss": 2.1047,
+      "step": 6511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3953262019941957e-05,
+      "loss": 2.0947,
+      "step": 6512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3951375014321017e-05,
+      "loss": 2.0769,
+      "step": 6513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3949487793628526e-05,
+      "loss": 2.0657,
+      "step": 6514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3947600357913983e-05,
+      "loss": 2.0974,
+      "step": 6515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3945712707226883e-05,
+      "loss": 2.0491,
+      "step": 6516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3943824841616722e-05,
+      "loss": 2.0994,
+      "step": 6517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3941936761133012e-05,
+      "loss": 2.0928,
+      "step": 6518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.394004846582526e-05,
+      "loss": 2.0522,
+      "step": 6519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3938159955742994e-05,
+      "loss": 2.0637,
+      "step": 6520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3936271230935727e-05,
+      "loss": 2.0315,
+      "step": 6521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3934382291452995e-05,
+      "loss": 2.0669,
+      "step": 6522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3932493137344335e-05,
+      "loss": 2.0815,
+      "step": 6523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3930603768659285e-05,
+      "loss": 2.1082,
+      "step": 6524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3928714185447393e-05,
+      "loss": 2.0554,
+      "step": 6525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3926824387758204e-05,
+      "loss": 2.0505,
+      "step": 6526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3924934375641282e-05,
+      "loss": 2.0471,
+      "step": 6527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3923044149146196e-05,
+      "loss": 2.0374,
+      "step": 6528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.392115370832251e-05,
+      "loss": 2.0828,
+      "step": 6529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3919263053219787e-05,
+      "loss": 2.054,
+      "step": 6530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3917372183887626e-05,
+      "loss": 2.0806,
+      "step": 6531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3915481100375603e-05,
+      "loss": 2.0737,
+      "step": 6532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.391358980273331e-05,
+      "loss": 2.0286,
+      "step": 6533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3911698291010344e-05,
+      "loss": 2.0615,
+      "step": 6534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.390980656525631e-05,
+      "loss": 2.1147,
+      "step": 6535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3907914625520815e-05,
+      "loss": 2.1196,
+      "step": 6536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3906022471853474e-05,
+      "loss": 2.0864,
+      "step": 6537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3904130104303904e-05,
+      "loss": 2.084,
+      "step": 6538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.390223752292173e-05,
+      "loss": 2.1133,
+      "step": 6539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.390034472775659e-05,
+      "loss": 2.0474,
+      "step": 6540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3898451718858113e-05,
+      "loss": 2.0242,
+      "step": 6541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3896558496275942e-05,
+      "loss": 2.0168,
+      "step": 6542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3894665060059723e-05,
+      "loss": 2.0469,
+      "step": 6543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3892771410259114e-05,
+      "loss": 2.0471,
+      "step": 6544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.389087754692377e-05,
+      "loss": 2.0811,
+      "step": 6545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3888983470103358e-05,
+      "loss": 2.0715,
+      "step": 6546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3887089179847548e-05,
+      "loss": 2.1096,
+      "step": 6547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.388519467620601e-05,
+      "loss": 2.1003,
+      "step": 6548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3883299959228434e-05,
+      "loss": 2.0447,
+      "step": 6549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3881405028964503e-05,
+      "loss": 2.0847,
+      "step": 6550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.38795098854639e-05,
+      "loss": 2.0701,
+      "step": 6551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3877614528776338e-05,
+      "loss": 2.0886,
+      "step": 6552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3875718958951517e-05,
+      "loss": 2.1226,
+      "step": 6553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3873823176039137e-05,
+      "loss": 2.0886,
+      "step": 6554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.387192718008892e-05,
+      "loss": 2.0508,
+      "step": 6555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3870030971150588e-05,
+      "loss": 2.0547,
+      "step": 6556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.386813454927386e-05,
+      "loss": 2.1072,
+      "step": 6557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3866237914508473e-05,
+      "loss": 2.1108,
+      "step": 6558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3864341066904165e-05,
+      "loss": 2.0391,
+      "step": 6559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.386244400651067e-05,
+      "loss": 2.063,
+      "step": 6560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3860546733377747e-05,
+      "loss": 1.9976,
+      "step": 6561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3858649247555143e-05,
+      "loss": 2.0183,
+      "step": 6562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.385675154909262e-05,
+      "loss": 2.0327,
+      "step": 6563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3854853638039943e-05,
+      "loss": 2.0337,
+      "step": 6564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3852955514446884e-05,
+      "loss": 2.0432,
+      "step": 6565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3851057178363216e-05,
+      "loss": 2.0691,
+      "step": 6566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3849158629838718e-05,
+      "loss": 2.0608,
+      "step": 6567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3847259868923185e-05,
+      "loss": 2.0466,
+      "step": 6568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3845360895666405e-05,
+      "loss": 2.0393,
+      "step": 6569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3843461710118172e-05,
+      "loss": 2.0671,
+      "step": 6570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.38415623123283e-05,
+      "loss": 2.114,
+      "step": 6571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3839662702346592e-05,
+      "loss": 2.0913,
+      "step": 6572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3837762880222866e-05,
+      "loss": 2.0957,
+      "step": 6573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3835862846006936e-05,
+      "loss": 2.0952,
+      "step": 6574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3833962599748636e-05,
+      "loss": 2.0645,
+      "step": 6575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.38320621414978e-05,
+      "loss": 2.0583,
+      "step": 6576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.383016147130426e-05,
+      "loss": 2.0798,
+      "step": 6577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3828260589217854e-05,
+      "loss": 2.0854,
+      "step": 6578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.382635949528844e-05,
+      "loss": 2.0757,
+      "step": 6579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3824458189565862e-05,
+      "loss": 2.0842,
+      "step": 6580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.382255667209999e-05,
+      "loss": 2.0859,
+      "step": 6581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3820654942940687e-05,
+      "loss": 2.0911,
+      "step": 6582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3818753002137817e-05,
+      "loss": 2.0266,
+      "step": 6583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.381685084974126e-05,
+      "loss": 2.0376,
+      "step": 6584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.38149484858009e-05,
+      "loss": 2.0164,
+      "step": 6585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3813045910366617e-05,
+      "loss": 2.0322,
+      "step": 6586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3811143123488317e-05,
+      "loss": 2.0569,
+      "step": 6587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3809240125215886e-05,
+      "loss": 2.0571,
+      "step": 6588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.380733691559923e-05,
+      "loss": 2.0388,
+      "step": 6589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3805433494688266e-05,
+      "loss": 2.075,
+      "step": 6590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.38035298625329e-05,
+      "loss": 2.0398,
+      "step": 6591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3801626019183064e-05,
+      "loss": 2.0815,
+      "step": 6592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3799721964688666e-05,
+      "loss": 2.0723,
+      "step": 6593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3797817699099654e-05,
+      "loss": 2.0786,
+      "step": 6594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3795913222465952e-05,
+      "loss": 2.0657,
+      "step": 6595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.379400853483752e-05,
+      "loss": 2.0955,
+      "step": 6596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.379210363626429e-05,
+      "loss": 2.0645,
+      "step": 6597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3790198526796223e-05,
+      "loss": 2.0505,
+      "step": 6598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.378829320648328e-05,
+      "loss": 2.092,
+      "step": 6599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3786387675375416e-05,
+      "loss": 2.0894,
+      "step": 6600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3784481933522615e-05,
+      "loss": 2.105,
+      "step": 6601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3782575980974846e-05,
+      "loss": 2.0867,
+      "step": 6602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.378066981778209e-05,
+      "loss": 1.9775,
+      "step": 6603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.377876344399433e-05,
+      "loss": 2.0007,
+      "step": 6604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3776856859661567e-05,
+      "loss": 2.0566,
+      "step": 6605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3774950064833794e-05,
+      "loss": 2.0393,
+      "step": 6606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3773043059561018e-05,
+      "loss": 2.0481,
+      "step": 6607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3771135843893243e-05,
+      "loss": 2.0586,
+      "step": 6608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3769228417880485e-05,
+      "loss": 2.0278,
+      "step": 6609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3767320781572767e-05,
+      "loss": 2.0249,
+      "step": 6610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.376541293502011e-05,
+      "loss": 2.0662,
+      "step": 6611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.376350487827255e-05,
+      "loss": 2.0605,
+      "step": 6612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.376159661138012e-05,
+      "loss": 2.0693,
+      "step": 6613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3759688134392862e-05,
+      "loss": 2.072,
+      "step": 6614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3757779447360824e-05,
+      "loss": 2.0254,
+      "step": 6615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3755870550334062e-05,
+      "loss": 2.0273,
+      "step": 6616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3753961443362635e-05,
+      "loss": 2.0979,
+      "step": 6617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3752052126496603e-05,
+      "loss": 2.0776,
+      "step": 6618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3750142599786035e-05,
+      "loss": 2.0991,
+      "step": 6619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.374823286328101e-05,
+      "loss": 2.0908,
+      "step": 6620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3746322917031608e-05,
+      "loss": 2.0359,
+      "step": 6621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.374441276108791e-05,
+      "loss": 2.053,
+      "step": 6622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.374250239550002e-05,
+      "loss": 2.0999,
+      "step": 6623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.374059182031802e-05,
+      "loss": 2.0674,
+      "step": 6624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.373868103559202e-05,
+      "loss": 2.0681,
+      "step": 6625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3736770041372132e-05,
+      "loss": 2.0574,
+      "step": 6626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3734858837708463e-05,
+      "loss": 2.0542,
+      "step": 6627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3732947424651135e-05,
+      "loss": 2.0745,
+      "step": 6628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.373103580225027e-05,
+      "loss": 2.0671,
+      "step": 6629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3729123970555998e-05,
+      "loss": 2.106,
+      "step": 6630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.372721192961846e-05,
+      "loss": 2.0513,
+      "step": 6631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.372529967948779e-05,
+      "loss": 2.0652,
+      "step": 6632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3723387220214138e-05,
+      "loss": 2.1113,
+      "step": 6633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3721474551847663e-05,
+      "loss": 2.1187,
+      "step": 6634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.371956167443851e-05,
+      "loss": 2.0984,
+      "step": 6635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.371764858803685e-05,
+      "loss": 2.0559,
+      "step": 6636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3715735292692844e-05,
+      "loss": 2.0671,
+      "step": 6637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3713821788456683e-05,
+      "loss": 2.106,
+      "step": 6638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3711908075378527e-05,
+      "loss": 2.1084,
+      "step": 6639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3709994153508566e-05,
+      "loss": 2.0923,
+      "step": 6640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3708080022896998e-05,
+      "loss": 2.0264,
+      "step": 6641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3706165683594004e-05,
+      "loss": 2.0269,
+      "step": 6642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3704251135649808e-05,
+      "loss": 2.1333,
+      "step": 6643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3702336379114597e-05,
+      "loss": 2.127,
+      "step": 6644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3700421414038593e-05,
+      "loss": 2.0771,
+      "step": 6645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3698506240472003e-05,
+      "loss": 2.0466,
+      "step": 6646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.369659085846507e-05,
+      "loss": 2.043,
+      "step": 6647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3694675268068e-05,
+      "loss": 2.0146,
+      "step": 6648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3692759469331048e-05,
+      "loss": 2.0854,
+      "step": 6649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3690843462304434e-05,
+      "loss": 2.0608,
+      "step": 6650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3688927247038416e-05,
+      "loss": 2.0984,
+      "step": 6651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.368701082358324e-05,
+      "loss": 2.0701,
+      "step": 6652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3685094191989164e-05,
+      "loss": 2.0498,
+      "step": 6653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3683177352306452e-05,
+      "loss": 2.0645,
+      "step": 6654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3681260304585364e-05,
+      "loss": 1.9705,
+      "step": 6655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3679343048876176e-05,
+      "loss": 1.9946,
+      "step": 6656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.367742558522916e-05,
+      "loss": 2.1099,
+      "step": 6657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3675507913694613e-05,
+      "loss": 2.072,
+      "step": 6658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.367359003432281e-05,
+      "loss": 2.0676,
+      "step": 6659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3671671947164056e-05,
+      "loss": 2.0916,
+      "step": 6660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.366975365226864e-05,
+      "loss": 2.0103,
+      "step": 6661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3667835149686873e-05,
+      "loss": 2.0366,
+      "step": 6662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3665916439469072e-05,
+      "loss": 2.0625,
+      "step": 6663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3663997521665536e-05,
+      "loss": 2.0627,
+      "step": 6664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3662078396326602e-05,
+      "loss": 2.0459,
+      "step": 6665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3660159063502587e-05,
+      "loss": 2.061,
+      "step": 6666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3658239523243832e-05,
+      "loss": 2.0989,
+      "step": 6667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3656319775600664e-05,
+      "loss": 2.0879,
+      "step": 6668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3654399820623435e-05,
+      "loss": 2.0264,
+      "step": 6669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.365247965836249e-05,
+      "loss": 2.0366,
+      "step": 6670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3650559288868184e-05,
+      "loss": 2.0945,
+      "step": 6671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.364863871219088e-05,
+      "loss": 2.0974,
+      "step": 6672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.364671792838094e-05,
+      "loss": 2.0308,
+      "step": 6673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3644796937488727e-05,
+      "loss": 2.0422,
+      "step": 6674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3642875739564628e-05,
+      "loss": 2.0288,
+      "step": 6675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3640954334659015e-05,
+      "loss": 2.0598,
+      "step": 6676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.363903272282228e-05,
+      "loss": 2.0356,
+      "step": 6677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.363711090410482e-05,
+      "loss": 2.0535,
+      "step": 6678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3635188878557017e-05,
+      "loss": 2.0996,
+      "step": 6679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.363326664622929e-05,
+      "loss": 2.1296,
+      "step": 6680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3631344207172037e-05,
+      "loss": 2.0776,
+      "step": 6681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3629421561435676e-05,
+      "loss": 2.0737,
+      "step": 6682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3627498709070628e-05,
+      "loss": 2.0842,
+      "step": 6683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3625575650127316e-05,
+      "loss": 2.0554,
+      "step": 6684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3623652384656168e-05,
+      "loss": 2.0737,
+      "step": 6685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3621728912707617e-05,
+      "loss": 2.0725,
+      "step": 6686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.361980523433211e-05,
+      "loss": 2.02,
+      "step": 6687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.361788134958009e-05,
+      "loss": 2.0356,
+      "step": 6688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.361595725850201e-05,
+      "loss": 2.0691,
+      "step": 6689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3614032961148325e-05,
+      "loss": 2.0762,
+      "step": 6690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3612108457569504e-05,
+      "loss": 2.1104,
+      "step": 6691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3610183747816e-05,
+      "loss": 2.1147,
+      "step": 6692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3608258831938302e-05,
+      "loss": 2.0493,
+      "step": 6693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3606333709986884e-05,
+      "loss": 2.0422,
+      "step": 6694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3604408382012224e-05,
+      "loss": 2.0879,
+      "step": 6695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.360248284806482e-05,
+      "loss": 2.0564,
+      "step": 6696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3600557108195162e-05,
+      "loss": 2.0918,
+      "step": 6697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.359863116245375e-05,
+      "loss": 2.0869,
+      "step": 6698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.3596705010891096e-05,
+      "loss": 2.0747,
+      "step": 6699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3594778653557703e-05,
+      "loss": 2.0798,
+      "step": 6700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.359285209050409e-05,
+      "loss": 2.0405,
+      "step": 6701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.359092532178078e-05,
+      "loss": 2.002,
+      "step": 6702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3588998347438297e-05,
+      "loss": 2.1003,
+      "step": 6703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.358707116752718e-05,
+      "loss": 2.1167,
+      "step": 6704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.358514378209796e-05,
+      "loss": 2.084,
+      "step": 6705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.358321619120119e-05,
+      "loss": 2.0959,
+      "step": 6706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3581288394887405e-05,
+      "loss": 2.0806,
+      "step": 6707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3579360393207172e-05,
+      "loss": 2.1006,
+      "step": 6708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.357743218621105e-05,
+      "loss": 2.0208,
+      "step": 6709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3575503773949592e-05,
+      "loss": 2.0564,
+      "step": 6710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3573575156473378e-05,
+      "loss": 2.064,
+      "step": 6711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3571646333832983e-05,
+      "loss": 2.0769,
+      "step": 6712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3569717306078983e-05,
+      "loss": 2.0601,
+      "step": 6713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.356778807326198e-05,
+      "loss": 2.0974,
+      "step": 6714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3565858635432542e-05,
+      "loss": 2.085,
+      "step": 6715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3563928992641285e-05,
+      "loss": 2.0842,
+      "step": 6716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.35619991449388e-05,
+      "loss": 2.0491,
+      "step": 6717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3560069092375702e-05,
+      "loss": 2.0623,
+      "step": 6718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3558138835002604e-05,
+      "loss": 2.1013,
+      "step": 6719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3556208372870122e-05,
+      "loss": 2.1094,
+      "step": 6720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3554277706028883e-05,
+      "loss": 2.0559,
+      "step": 6721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3552346834529515e-05,
+      "loss": 2.0557,
+      "step": 6722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3550415758422654e-05,
+      "loss": 2.0591,
+      "step": 6723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3548484477758935e-05,
+      "loss": 2.0776,
+      "step": 6724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.354655299258901e-05,
+      "loss": 2.0725,
+      "step": 6725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3544621302963528e-05,
+      "loss": 2.0337,
+      "step": 6726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3542689408933145e-05,
+      "loss": 2.0432,
+      "step": 6727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3540757310548525e-05,
+      "loss": 2.0649,
+      "step": 6728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3538825007860328e-05,
+      "loss": 2.0488,
+      "step": 6729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3536892500919235e-05,
+      "loss": 2.0352,
+      "step": 6730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3534959789775917e-05,
+      "loss": 2.0752,
+      "step": 6731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.353302687448106e-05,
+      "loss": 2.0662,
+      "step": 6732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3531093755085356e-05,
+      "loss": 2.0918,
+      "step": 6733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3529160431639495e-05,
+      "loss": 2.0784,
+      "step": 6734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.352722690419417e-05,
+      "loss": 2.0759,
+      "step": 6735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.35252931728001e-05,
+      "loss": 2.0579,
+      "step": 6736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3523359237507978e-05,
+      "loss": 2.1074,
+      "step": 6737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3521425098368535e-05,
+      "loss": 2.0654,
+      "step": 6738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3519490755432483e-05,
+      "loss": 2.0427,
+      "step": 6739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3517556208750544e-05,
+      "loss": 2.0339,
+      "step": 6740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3515621458373465e-05,
+      "loss": 2.105,
+      "step": 6741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3513686504351965e-05,
+      "loss": 2.1121,
+      "step": 6742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3511751346736793e-05,
+      "loss": 2.0686,
+      "step": 6743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3509815985578702e-05,
+      "loss": 2.0801,
+      "step": 6744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3507880420928433e-05,
+      "loss": 2.0679,
+      "step": 6745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3505944652836755e-05,
+      "loss": 2.0554,
+      "step": 6746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3504008681354423e-05,
+      "loss": 2.0747,
+      "step": 6747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3502072506532208e-05,
+      "loss": 2.0779,
+      "step": 6748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3500136128420885e-05,
+      "loss": 2.1233,
+      "step": 6749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3498199547071237e-05,
+      "loss": 2.1177,
+      "step": 6750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.349626276253404e-05,
+      "loss": 2.0872,
+      "step": 6751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3494325774860094e-05,
+      "loss": 2.0637,
+      "step": 6752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.349238858410018e-05,
+      "loss": 2.0981,
+      "step": 6753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3490451190305116e-05,
+      "loss": 2.0857,
+      "step": 6754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3488513593525693e-05,
+      "loss": 2.1187,
+      "step": 6755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3486575793812734e-05,
+      "loss": 2.0933,
+      "step": 6756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3484637791217047e-05,
+      "loss": 2.0544,
+      "step": 6757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.348269958578946e-05,
+      "loss": 2.0786,
+      "step": 6758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3480761177580793e-05,
+      "loss": 2.0042,
+      "step": 6759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3478822566641887e-05,
+      "loss": 2.0208,
+      "step": 6760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3476883753023575e-05,
+      "loss": 2.0903,
+      "step": 6761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3474944736776696e-05,
+      "loss": 2.0894,
+      "step": 6762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.347300551795211e-05,
+      "loss": 2.0371,
+      "step": 6763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3471066096600655e-05,
+      "loss": 2.0535,
+      "step": 6764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3469126472773208e-05,
+      "loss": 2.0159,
+      "step": 6765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3467186646520617e-05,
+      "loss": 2.0176,
+      "step": 6766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.346524661789376e-05,
+      "loss": 2.063,
+      "step": 6767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3463306386943518e-05,
+      "loss": 2.0618,
+      "step": 6768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3461365953720755e-05,
+      "loss": 2.0688,
+      "step": 6769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3459425318276373e-05,
+      "loss": 2.0471,
+      "step": 6770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.345748448066125e-05,
+      "loss": 2.019,
+      "step": 6771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3455543440926294e-05,
+      "loss": 2.009,
+      "step": 6772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3453602199122397e-05,
+      "loss": 2.0859,
+      "step": 6773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.345166075530047e-05,
+      "loss": 2.0613,
+      "step": 6774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3449719109511428e-05,
+      "loss": 2.0168,
+      "step": 6775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.344777726180618e-05,
+      "loss": 2.0403,
+      "step": 6776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3445835212235656e-05,
+      "loss": 2.124,
+      "step": 6777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.344389296085078e-05,
+      "loss": 2.1416,
+      "step": 6778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3441950507702485e-05,
+      "loss": 2.0986,
+      "step": 6779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.344000785284172e-05,
+      "loss": 2.0833,
+      "step": 6780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.343806499631941e-05,
+      "loss": 2.0977,
+      "step": 6781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3436121938186518e-05,
+      "loss": 2.1206,
+      "step": 6782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3434178678493997e-05,
+      "loss": 2.0671,
+      "step": 6783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.343223521729281e-05,
+      "loss": 2.0674,
+      "step": 6784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3430291554633908e-05,
+      "loss": 2.0996,
+      "step": 6785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3428347690568268e-05,
+      "loss": 2.0735,
+      "step": 6786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.342640362514687e-05,
+      "loss": 2.0708,
+      "step": 6787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3424459358420698e-05,
+      "loss": 2.0347,
+      "step": 6788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.342251489044073e-05,
+      "loss": 2.075,
+      "step": 6789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3420570221257957e-05,
+      "loss": 2.0928,
+      "step": 6790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3418625350923383e-05,
+      "loss": 2.0562,
+      "step": 6791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3416680279488002e-05,
+      "loss": 2.0493,
+      "step": 6792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.341473500700283e-05,
+      "loss": 2.0859,
+      "step": 6793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3412789533518868e-05,
+      "loss": 2.1077,
+      "step": 6794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.341084385908714e-05,
+      "loss": 2.0696,
+      "step": 6795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3408897983758675e-05,
+      "loss": 2.0486,
+      "step": 6796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.340695190758449e-05,
+      "loss": 2.1064,
+      "step": 6797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3405005630615627e-05,
+      "loss": 2.1216,
+      "step": 6798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3403059152903122e-05,
+      "loss": 2.0786,
+      "step": 6799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3401112474498023e-05,
+      "loss": 2.094,
+      "step": 6800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3399165595451374e-05,
+      "loss": 2.0344,
+      "step": 6801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3397218515814225e-05,
+      "loss": 2.02,
+      "step": 6802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3395271235637644e-05,
+      "loss": 2.1111,
+      "step": 6803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3393323754972695e-05,
+      "loss": 2.061,
+      "step": 6804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3391376073870447e-05,
+      "loss": 2.0559,
+      "step": 6805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3389428192381982e-05,
+      "loss": 2.0759,
+      "step": 6806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3387480110558367e-05,
+      "loss": 2.0193,
+      "step": 6807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.33855318284507e-05,
+      "loss": 2.0154,
+      "step": 6808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3383583346110074e-05,
+      "loss": 2.1013,
+      "step": 6809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3381634663587576e-05,
+      "loss": 2.0679,
+      "step": 6810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.337968578093431e-05,
+      "loss": 2.0505,
+      "step": 6811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.337773669820139e-05,
+      "loss": 2.0554,
+      "step": 6812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.337578741543992e-05,
+      "loss": 2.0906,
+      "step": 6813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3373837932701027e-05,
+      "loss": 2.0708,
+      "step": 6814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.337188825003582e-05,
+      "loss": 2.0325,
+      "step": 6815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3369938367495448e-05,
+      "loss": 2.0708,
+      "step": 6816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3367988285131024e-05,
+      "loss": 2.0947,
+      "step": 6817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3366038002993696e-05,
+      "loss": 2.0303,
+      "step": 6818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3364087521134612e-05,
+      "loss": 2.0491,
+      "step": 6819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.336213683960491e-05,
+      "loss": 2.0459,
+      "step": 6820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3360185958455753e-05,
+      "loss": 2.0571,
+      "step": 6821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3358234877738297e-05,
+      "loss": 2.053,
+      "step": 6822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3356283597503708e-05,
+      "loss": 2.0,
+      "step": 6823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3354332117803152e-05,
+      "loss": 2.0022,
+      "step": 6824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3352380438687812e-05,
+      "loss": 2.0422,
+      "step": 6825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3350428560208868e-05,
+      "loss": 2.0627,
+      "step": 6826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3348476482417494e-05,
+      "loss": 2.05,
+      "step": 6827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3346524205364896e-05,
+      "loss": 2.0522,
+      "step": 6828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3344571729102262e-05,
+      "loss": 2.0483,
+      "step": 6829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3342619053680795e-05,
+      "loss": 2.0444,
+      "step": 6830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3340666179151706e-05,
+      "loss": 2.0183,
+      "step": 6831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3338713105566194e-05,
+      "loss": 2.0493,
+      "step": 6832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3336759832975488e-05,
+      "loss": 2.0432,
+      "step": 6833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.333480636143081e-05,
+      "loss": 2.0596,
+      "step": 6834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.333285269098338e-05,
+      "loss": 2.0298,
+      "step": 6835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.333089882168444e-05,
+      "loss": 2.0261,
+      "step": 6836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.332894475358522e-05,
+      "loss": 2.0383,
+      "step": 6837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3326990486736964e-05,
+      "loss": 2.0171,
+      "step": 6838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3325036021190928e-05,
+      "loss": 1.9944,
+      "step": 6839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3323081356998355e-05,
+      "loss": 1.991,
+      "step": 6840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3321126494210508e-05,
+      "loss": 2.062,
+      "step": 6841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3319171432878657e-05,
+      "loss": 2.0845,
+      "step": 6842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.331721617305406e-05,
+      "loss": 2.0549,
+      "step": 6843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3315260714788005e-05,
+      "loss": 2.0906,
+      "step": 6844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.331330505813176e-05,
+      "loss": 2.0994,
+      "step": 6845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.331134920313662e-05,
+      "loss": 2.1003,
+      "step": 6846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3309393149853862e-05,
+      "loss": 2.0654,
+      "step": 6847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.330743689833479e-05,
+      "loss": 2.0505,
+      "step": 6848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3305480448630703e-05,
+      "loss": 1.9883,
+      "step": 6849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.330352380079291e-05,
+      "loss": 1.97,
+      "step": 6850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.330156695487272e-05,
+      "loss": 2.0349,
+      "step": 6851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.329960991092144e-05,
+      "loss": 2.043,
+      "step": 6852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3297652668990404e-05,
+      "loss": 2.0623,
+      "step": 6853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.329569522913093e-05,
+      "loss": 2.0527,
+      "step": 6854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3293737591394353e-05,
+      "loss": 2.075,
+      "step": 6855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3291779755832007e-05,
+      "loss": 2.0732,
+      "step": 6856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3289821722495237e-05,
+      "loss": 2.0183,
+      "step": 6857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.328786349143539e-05,
+      "loss": 2.0203,
+      "step": 6858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.328590506270382e-05,
+      "loss": 2.0664,
+      "step": 6859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.328394643635188e-05,
+      "loss": 2.0649,
+      "step": 6860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3281987612430936e-05,
+      "loss": 2.0789,
+      "step": 6861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3280028590992357e-05,
+      "loss": 2.0874,
+      "step": 6862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3278069372087508e-05,
+      "loss": 2.0884,
+      "step": 6863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.327610995576778e-05,
+      "loss": 2.0725,
+      "step": 6864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3274150342084544e-05,
+      "loss": 2.0007,
+      "step": 6865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.32721905310892e-05,
+      "loss": 2.0381,
+      "step": 6866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.327023052283313e-05,
+      "loss": 2.0488,
+      "step": 6867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.326827031736774e-05,
+      "loss": 2.0632,
+      "step": 6868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3266309914744438e-05,
+      "loss": 2.0989,
+      "step": 6869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3264349315014624e-05,
+      "loss": 2.0979,
+      "step": 6870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.326238851822972e-05,
+      "loss": 2.0234,
+      "step": 6871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3260427524441136e-05,
+      "loss": 2.0115,
+      "step": 6872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3258466333700306e-05,
+      "loss": 2.1348,
+      "step": 6873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3256504946058655e-05,
+      "loss": 2.074,
+      "step": 6874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3254543361567625e-05,
+      "loss": 2.0354,
+      "step": 6875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3252581580278646e-05,
+      "loss": 2.0161,
+      "step": 6876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3250619602243173e-05,
+      "loss": 2.1023,
+      "step": 6877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3248657427512655e-05,
+      "loss": 2.0637,
+      "step": 6878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.324669505613854e-05,
+      "loss": 2.0696,
+      "step": 6879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3244732488172296e-05,
+      "loss": 2.1079,
+      "step": 6880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3242769723665386e-05,
+      "loss": 2.0852,
+      "step": 6881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3240806762669285e-05,
+      "loss": 2.1167,
+      "step": 6882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3238843605235466e-05,
+      "loss": 2.1213,
+      "step": 6883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.323688025141541e-05,
+      "loss": 2.0916,
+      "step": 6884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3234916701260607e-05,
+      "loss": 2.0066,
+      "step": 6885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3232952954822546e-05,
+      "loss": 2.0129,
+      "step": 6886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3230989012152726e-05,
+      "loss": 2.0732,
+      "step": 6887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3229024873302644e-05,
+      "loss": 2.0908,
+      "step": 6888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3227060538323814e-05,
+      "loss": 2.1023,
+      "step": 6889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3225096007267747e-05,
+      "loss": 2.0967,
+      "step": 6890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3223131280185956e-05,
+      "loss": 2.0603,
+      "step": 6891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.322116635712997e-05,
+      "loss": 2.0508,
+      "step": 6892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.321920123815131e-05,
+      "loss": 2.1104,
+      "step": 6893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3217235923301516e-05,
+      "loss": 2.103,
+      "step": 6894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3215270412632123e-05,
+      "loss": 2.0811,
+      "step": 6895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3213304706194673e-05,
+      "loss": 2.0774,
+      "step": 6896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3211338804040716e-05,
+      "loss": 2.0818,
+      "step": 6897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3209372706221802e-05,
+      "loss": 2.063,
+      "step": 6898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3207406412789493e-05,
+      "loss": 2.0947,
+      "step": 6899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3205439923795355e-05,
+      "loss": 2.0706,
+      "step": 6900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.320347323929095e-05,
+      "loss": 2.0906,
+      "step": 6901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3201506359327864e-05,
+      "loss": 2.0469,
+      "step": 6902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3199539283957662e-05,
+      "loss": 2.03,
+      "step": 6903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3197572013231935e-05,
+      "loss": 2.0447,
+      "step": 6904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3195604547202274e-05,
+      "loss": 2.0793,
+      "step": 6905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.319363688592027e-05,
+      "loss": 2.0776,
+      "step": 6906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3191669029437528e-05,
+      "loss": 2.0093,
+      "step": 6907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.318970097780564e-05,
+      "loss": 2.0251,
+      "step": 6908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3187732731076233e-05,
+      "loss": 2.0686,
+      "step": 6909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.318576428930091e-05,
+      "loss": 2.0686,
+      "step": 6910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3183795652531295e-05,
+      "loss": 2.0396,
+      "step": 6911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3181826820819012e-05,
+      "loss": 2.0332,
+      "step": 6912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3179857794215697e-05,
+      "loss": 2.0952,
+      "step": 6913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3177888572772974e-05,
+      "loss": 2.1086,
+      "step": 6914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3175919156542498e-05,
+      "loss": 2.0679,
+      "step": 6915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3173949545575903e-05,
+      "loss": 2.0911,
+      "step": 6916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3171979739924842e-05,
+      "loss": 2.0591,
+      "step": 6917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.317000973964098e-05,
+      "loss": 2.0596,
+      "step": 6918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.316803954477596e-05,
+      "loss": 2.0173,
+      "step": 6919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3166069155381466e-05,
+      "loss": 2.0312,
+      "step": 6920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3164098571509164e-05,
+      "loss": 2.041,
+      "step": 6921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3162127793210722e-05,
+      "loss": 2.0342,
+      "step": 6922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.316015682053783e-05,
+      "loss": 2.0701,
+      "step": 6923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3158185653542172e-05,
+      "loss": 2.0698,
+      "step": 6924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.315621429227544e-05,
+      "loss": 2.1204,
+      "step": 6925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3154242736789326e-05,
+      "loss": 2.0959,
+      "step": 6926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.315227098713554e-05,
+      "loss": 2.1074,
+      "step": 6927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3150299043365785e-05,
+      "loss": 2.1294,
+      "step": 6928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3148326905531775e-05,
+      "loss": 2.0095,
+      "step": 6929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.314635457368522e-05,
+      "loss": 1.9976,
+      "step": 6930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3144382047877845e-05,
+      "loss": 2.095,
+      "step": 6931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.314240932816139e-05,
+      "loss": 2.0974,
+      "step": 6932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.314043641458757e-05,
+      "loss": 2.0503,
+      "step": 6933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3138463307208124e-05,
+      "loss": 2.0774,
+      "step": 6934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3136490006074802e-05,
+      "loss": 2.0579,
+      "step": 6935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.313451651123935e-05,
+      "loss": 2.0659,
+      "step": 6936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.313254282275352e-05,
+      "loss": 2.0984,
+      "step": 6937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3130568940669073e-05,
+      "loss": 2.1201,
+      "step": 6938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3128594865037762e-05,
+      "loss": 2.0625,
+      "step": 6939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3126620595911363e-05,
+      "loss": 2.063,
+      "step": 6940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3124646133341642e-05,
+      "loss": 2.0681,
+      "step": 6941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.312267147738039e-05,
+      "loss": 2.0762,
+      "step": 6942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.312069662807938e-05,
+      "loss": 2.1023,
+      "step": 6943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3118721585490397e-05,
+      "loss": 2.0723,
+      "step": 6944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.311674634966524e-05,
+      "loss": 2.0498,
+      "step": 6945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3114770920655705e-05,
+      "loss": 2.0981,
+      "step": 6946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3112795298513597e-05,
+      "loss": 2.094,
+      "step": 6947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3110819483290726e-05,
+      "loss": 2.0837,
+      "step": 6948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.31088434750389e-05,
+      "loss": 2.1262,
+      "step": 6949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.310686727380995e-05,
+      "loss": 2.1213,
+      "step": 6950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.310489087965568e-05,
+      "loss": 2.0452,
+      "step": 6951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3102914292627935e-05,
+      "loss": 2.0393,
+      "step": 6952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3100937512778543e-05,
+      "loss": 2.0603,
+      "step": 6953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.309896054015934e-05,
+      "loss": 2.0647,
+      "step": 6954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3096983374822175e-05,
+      "loss": 2.1023,
+      "step": 6955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.309500601681889e-05,
+      "loss": 2.1099,
+      "step": 6956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3093028466201348e-05,
+      "loss": 2.0598,
+      "step": 6957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3091050723021404e-05,
+      "loss": 2.064,
+      "step": 6958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3089072787330917e-05,
+      "loss": 2.032,
+      "step": 6959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3087094659181763e-05,
+      "loss": 2.0312,
+      "step": 6960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3085116338625813e-05,
+      "loss": 2.114,
+      "step": 6961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3083137825714945e-05,
+      "loss": 2.0984,
+      "step": 6962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3081159120501045e-05,
+      "loss": 2.0818,
+      "step": 6963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3079180223036006e-05,
+      "loss": 2.0759,
+      "step": 6964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.307720113337171e-05,
+      "loss": 2.0251,
+      "step": 6965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.307522185156007e-05,
+      "loss": 2.0566,
+      "step": 6966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3073242377652986e-05,
+      "loss": 2.104,
+      "step": 6967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3071262711702368e-05,
+      "loss": 2.0811,
+      "step": 6968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3069282853760124e-05,
+      "loss": 2.094,
+      "step": 6969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3067302803878185e-05,
+      "loss": 2.0732,
+      "step": 6970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3065322562108463e-05,
+      "loss": 2.0547,
+      "step": 6971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3063342128502892e-05,
+      "loss": 2.0891,
+      "step": 6972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3061361503113416e-05,
+      "loss": 2.0776,
+      "step": 6973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.305938068599196e-05,
+      "loss": 2.1096,
+      "step": 6974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3057399677190476e-05,
+      "loss": 2.0552,
+      "step": 6975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3055418476760913e-05,
+      "loss": 2.0515,
+      "step": 6976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3053437084755224e-05,
+      "loss": 2.0054,
+      "step": 6977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3051455501225378e-05,
+      "loss": 2.0254,
+      "step": 6978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3049473726223324e-05,
+      "loss": 2.0598,
+      "step": 6979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3047491759801044e-05,
+      "loss": 2.04,
+      "step": 6980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3045509602010503e-05,
+      "loss": 2.103,
+      "step": 6981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.304352725290369e-05,
+      "loss": 2.0693,
+      "step": 6982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.304154471253259e-05,
+      "loss": 2.0718,
+      "step": 6983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3039561980949184e-05,
+      "loss": 2.0579,
+      "step": 6984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.303757905820547e-05,
+      "loss": 2.1089,
+      "step": 6985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3035595944353455e-05,
+      "loss": 2.0913,
+      "step": 6986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3033612639445133e-05,
+      "loss": 2.0955,
+      "step": 6987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3031629143532526e-05,
+      "loss": 2.0947,
+      "step": 6988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3029645456667636e-05,
+      "loss": 2.0657,
+      "step": 6989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3027661578902492e-05,
+      "loss": 2.0718,
+      "step": 6990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3025677510289116e-05,
+      "loss": 2.0806,
+      "step": 6991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.302369325087954e-05,
+      "loss": 2.0562,
+      "step": 6992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3021708800725797e-05,
+      "loss": 2.0791,
+      "step": 6993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.301972415987992e-05,
+      "loss": 2.0662,
+      "step": 6994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3017739328393967e-05,
+      "loss": 2.0454,
+      "step": 6995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3015754306319977e-05,
+      "loss": 2.1113,
+      "step": 6996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.301376909371001e-05,
+      "loss": 2.073,
+      "step": 6997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3011783690616125e-05,
+      "loss": 2.0413,
+      "step": 6998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.300979809709039e-05,
+      "loss": 2.0493,
+      "step": 6999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3007812313184875e-05,
+      "loss": 2.0525,
+      "step": 7000
+    },
+    {
+      "Bleu_1": 0.7722608461280476,
+      "Bleu_2": 0.6117017221393226,
+      "Bleu_3": 0.46962782299186484,
+      "Bleu_4": 0.356714600469744,
+      "CIDEr": 1.201525215513983,
+      "METEOR": 0.2781299740896172,
+      "ROUGE_L": 0.559012583070473,
+      "epoch": 0.35,
+      "step": 7000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3005826338951645e-05,
+      "loss": 2.0608,
+      "step": 7001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.300384017444279e-05,
+      "loss": 2.04,
+      "step": 7002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.300185381971039e-05,
+      "loss": 2.0549,
+      "step": 7003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2999867274806544e-05,
+      "loss": 2.0701,
+      "step": 7004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2997880539783334e-05,
+      "loss": 2.0735,
+      "step": 7005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2995893614692862e-05,
+      "loss": 2.1011,
+      "step": 7006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2993906499587242e-05,
+      "loss": 2.0427,
+      "step": 7007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2991919194518573e-05,
+      "loss": 2.0366,
+      "step": 7008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.298993169953898e-05,
+      "loss": 2.0737,
+      "step": 7009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.298794401470057e-05,
+      "loss": 2.063,
+      "step": 7010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.298595614005548e-05,
+      "loss": 2.0513,
+      "step": 7011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2983968075655832e-05,
+      "loss": 2.0232,
+      "step": 7012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2981979821553766e-05,
+      "loss": 2.1011,
+      "step": 7013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.297999137780142e-05,
+      "loss": 2.1042,
+      "step": 7014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2978002744450933e-05,
+      "loss": 2.1191,
+      "step": 7015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2976013921554463e-05,
+      "loss": 2.1387,
+      "step": 7016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.297402490916416e-05,
+      "loss": 2.1086,
+      "step": 7017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.297203570733218e-05,
+      "loss": 2.0872,
+      "step": 7018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.29700463161107e-05,
+      "loss": 2.0317,
+      "step": 7019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.296805673555188e-05,
+      "loss": 2.0156,
+      "step": 7020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.296606696570789e-05,
+      "loss": 2.0693,
+      "step": 7021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2964077006630923e-05,
+      "loss": 2.0737,
+      "step": 7022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.296208685837315e-05,
+      "loss": 2.0735,
+      "step": 7023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.296009652098677e-05,
+      "loss": 2.0801,
+      "step": 7024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.295810599452397e-05,
+      "loss": 2.0254,
+      "step": 7025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2956115279036952e-05,
+      "loss": 1.9841,
+      "step": 7026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2954124374577922e-05,
+      "loss": 2.0896,
+      "step": 7027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.295213328119909e-05,
+      "loss": 2.1169,
+      "step": 7028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2950141998952668e-05,
+      "loss": 2.0596,
+      "step": 7029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2948150527890873e-05,
+      "loss": 2.0471,
+      "step": 7030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.294615886806593e-05,
+      "loss": 2.0845,
+      "step": 7031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2944167019530072e-05,
+      "loss": 2.1028,
+      "step": 7032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2942174982335528e-05,
+      "loss": 2.1194,
+      "step": 7033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2940182756534536e-05,
+      "loss": 2.1206,
+      "step": 7034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2938190342179345e-05,
+      "loss": 2.0925,
+      "step": 7035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2936197739322202e-05,
+      "loss": 2.1306,
+      "step": 7036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2934204948015357e-05,
+      "loss": 2.0752,
+      "step": 7037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2932211968311078e-05,
+      "loss": 2.0779,
+      "step": 7038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2930218800261623e-05,
+      "loss": 2.1064,
+      "step": 7039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2928225443919255e-05,
+      "loss": 2.0745,
+      "step": 7040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.292623189933625e-05,
+      "loss": 2.009,
+      "step": 7041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2924238166564893e-05,
+      "loss": 2.0447,
+      "step": 7042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.292224424565746e-05,
+      "loss": 2.0652,
+      "step": 7043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2920250136666247e-05,
+      "loss": 2.0906,
+      "step": 7044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.291825583964354e-05,
+      "loss": 2.0603,
+      "step": 7045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.291626135464164e-05,
+      "loss": 2.0139,
+      "step": 7046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2914266681712847e-05,
+      "loss": 2.1133,
+      "step": 7047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2912271820909477e-05,
+      "loss": 2.0994,
+      "step": 7048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2910276772283832e-05,
+      "loss": 2.1155,
+      "step": 7049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2908281535888242e-05,
+      "loss": 2.0964,
+      "step": 7050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2906286111775013e-05,
+      "loss": 2.0972,
+      "step": 7051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2904290499996494e-05,
+      "loss": 2.1013,
+      "step": 7052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2902294700605e-05,
+      "loss": 2.1262,
+      "step": 7053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2900298713652874e-05,
+      "loss": 2.104,
+      "step": 7054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2898302539192463e-05,
+      "loss": 2.0369,
+      "step": 7055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2896306177276108e-05,
+      "loss": 2.0562,
+      "step": 7056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2894309627956166e-05,
+      "loss": 2.0908,
+      "step": 7057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2892312891284988e-05,
+      "loss": 2.113,
+      "step": 7058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2890315967314944e-05,
+      "loss": 2.0903,
+      "step": 7059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.28883188560984e-05,
+      "loss": 2.0803,
+      "step": 7060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.288632155768772e-05,
+      "loss": 2.073,
+      "step": 7061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.288432407213529e-05,
+      "loss": 2.0588,
+      "step": 7062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2882326399493485e-05,
+      "loss": 1.9604,
+      "step": 7063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2880328539814695e-05,
+      "loss": 2.0176,
+      "step": 7064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2878330493151316e-05,
+      "loss": 2.0583,
+      "step": 7065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2876332259555734e-05,
+      "loss": 2.061,
+      "step": 7066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2874333839080357e-05,
+      "loss": 2.0493,
+      "step": 7067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.287233523177759e-05,
+      "loss": 2.0122,
+      "step": 7068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2870336437699842e-05,
+      "loss": 2.0361,
+      "step": 7069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2868337456899533e-05,
+      "loss": 2.0569,
+      "step": 7070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2866338289429083e-05,
+      "loss": 2.0645,
+      "step": 7071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2864338935340917e-05,
+      "loss": 2.1052,
+      "step": 7072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.286233939468747e-05,
+      "loss": 1.9678,
+      "step": 7073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.286033966752117e-05,
+      "loss": 1.9941,
+      "step": 7074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.285833975389446e-05,
+      "loss": 2.0625,
+      "step": 7075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2856339653859786e-05,
+      "loss": 2.0867,
+      "step": 7076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.28543393674696e-05,
+      "loss": 2.0835,
+      "step": 7077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2852338894776354e-05,
+      "loss": 2.0815,
+      "step": 7078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.285033823583251e-05,
+      "loss": 2.0557,
+      "step": 7079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.284833739069054e-05,
+      "loss": 2.0469,
+      "step": 7080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.28463363594029e-05,
+      "loss": 2.0991,
+      "step": 7081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2844335142022075e-05,
+      "loss": 2.104,
+      "step": 7082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.284233373860054e-05,
+      "loss": 2.1201,
+      "step": 7083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2840332149190776e-05,
+      "loss": 2.1086,
+      "step": 7084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2838330373845286e-05,
+      "loss": 2.0684,
+      "step": 7085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2836328412616545e-05,
+      "loss": 2.0457,
+      "step": 7086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2834326265557067e-05,
+      "loss": 2.0703,
+      "step": 7087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2832323932719354e-05,
+      "loss": 2.0952,
+      "step": 7088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2830321414155904e-05,
+      "loss": 1.9883,
+      "step": 7089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2828318709919247e-05,
+      "loss": 2.0132,
+      "step": 7090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2826315820061893e-05,
+      "loss": 2.1128,
+      "step": 7091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2824312744636365e-05,
+      "loss": 2.1042,
+      "step": 7092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.282230948369519e-05,
+      "loss": 2.0957,
+      "step": 7093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2820306037290905e-05,
+      "loss": 2.0867,
+      "step": 7094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2818302405476044e-05,
+      "loss": 2.064,
+      "step": 7095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2816298588303157e-05,
+      "loss": 2.0723,
+      "step": 7096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2814294585824788e-05,
+      "loss": 2.0669,
+      "step": 7097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2812290398093486e-05,
+      "loss": 2.0959,
+      "step": 7098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2810286025161807e-05,
+      "loss": 2.0645,
+      "step": 7099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2808281467082324e-05,
+      "loss": 2.0815,
+      "step": 7100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2806276723907598e-05,
+      "loss": 2.0901,
+      "step": 7101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.28042717956902e-05,
+      "loss": 2.093,
+      "step": 7102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2802266682482703e-05,
+      "loss": 2.0608,
+      "step": 7103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.28002613843377e-05,
+      "loss": 2.0583,
+      "step": 7104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2798255901307764e-05,
+      "loss": 2.0269,
+      "step": 7105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2796250233445503e-05,
+      "loss": 2.0466,
+      "step": 7106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.27942443808035e-05,
+      "loss": 2.0051,
+      "step": 7107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2792238343434356e-05,
+      "loss": 1.9578,
+      "step": 7108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.279023212139069e-05,
+      "loss": 2.0557,
+      "step": 7109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2788225714725098e-05,
+      "loss": 2.0955,
+      "step": 7110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.27862191234902e-05,
+      "loss": 2.1396,
+      "step": 7111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2784212347738623e-05,
+      "loss": 2.1033,
+      "step": 7112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2782205387522985e-05,
+      "loss": 2.0762,
+      "step": 7113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.278019824289592e-05,
+      "loss": 2.105,
+      "step": 7114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2778190913910065e-05,
+      "loss": 2.052,
+      "step": 7115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2776183400618057e-05,
+      "loss": 2.0605,
+      "step": 7116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2774175703072537e-05,
+      "loss": 2.0789,
+      "step": 7117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2772167821326153e-05,
+      "loss": 2.0889,
+      "step": 7118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.277015975543157e-05,
+      "loss": 2.0889,
+      "step": 7119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2768151505441442e-05,
+      "loss": 2.0686,
+      "step": 7120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.276614307140843e-05,
+      "loss": 2.1084,
+      "step": 7121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.276413445338521e-05,
+      "loss": 2.123,
+      "step": 7122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.276212565142445e-05,
+      "loss": 2.0413,
+      "step": 7123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.276011666557882e-05,
+      "loss": 2.0383,
+      "step": 7124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2758107495901027e-05,
+      "loss": 2.0977,
+      "step": 7125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2756098142443737e-05,
+      "loss": 2.1169,
+      "step": 7126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2754088605259654e-05,
+      "loss": 2.0398,
+      "step": 7127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.275207888440147e-05,
+      "loss": 2.0679,
+      "step": 7128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2750068979921893e-05,
+      "loss": 2.0339,
+      "step": 7129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2748058891873626e-05,
+      "loss": 2.0212,
+      "step": 7130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.274604862030939e-05,
+      "loss": 2.0818,
+      "step": 7131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.274403816528189e-05,
+      "loss": 2.0532,
+      "step": 7132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2742027526843852e-05,
+      "loss": 2.0264,
+      "step": 7133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2740016705048002e-05,
+      "loss": 1.9963,
+      "step": 7134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2738005699947074e-05,
+      "loss": 2.0808,
+      "step": 7135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.273599451159381e-05,
+      "loss": 2.0403,
+      "step": 7136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.273398314004094e-05,
+      "loss": 2.0515,
+      "step": 7137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2731971585341213e-05,
+      "loss": 2.0789,
+      "step": 7138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.272995984754738e-05,
+      "loss": 2.1099,
+      "step": 7139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.27279479267122e-05,
+      "loss": 2.1177,
+      "step": 7140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.272593582288843e-05,
+      "loss": 2.0281,
+      "step": 7141
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2723923536128834e-05,
+      "loss": 2.0168,
+      "step": 7142
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2721911066486188e-05,
+      "loss": 2.0654,
+      "step": 7143
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.271989841401326e-05,
+      "loss": 2.0715,
+      "step": 7144
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.271788557876283e-05,
+      "loss": 2.1042,
+      "step": 7145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.271587256078769e-05,
+      "loss": 2.0894,
+      "step": 7146
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.271385936014062e-05,
+      "loss": 2.072,
+      "step": 7147
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2711845976874416e-05,
+      "loss": 2.0901,
+      "step": 7148
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.270983241104188e-05,
+      "loss": 2.0435,
+      "step": 7149
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2707818662695812e-05,
+      "loss": 2.0654,
+      "step": 7150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2705804731889023e-05,
+      "loss": 2.0667,
+      "step": 7151
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2703790618674322e-05,
+      "loss": 2.0718,
+      "step": 7152
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.270177632310453e-05,
+      "loss": 2.0481,
+      "step": 7153
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.269976184523247e-05,
+      "loss": 2.0674,
+      "step": 7154
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2697747185110966e-05,
+      "loss": 2.0901,
+      "step": 7155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2695732342792856e-05,
+      "loss": 2.0652,
+      "step": 7156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2693717318330966e-05,
+      "loss": 2.0979,
+      "step": 7157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.269170211177815e-05,
+      "loss": 2.0662,
+      "step": 7158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.268968672318725e-05,
+      "loss": 2.0747,
+      "step": 7159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2687671152611112e-05,
+      "loss": 2.0779,
+      "step": 7160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.26856554001026e-05,
+      "loss": 2.0801,
+      "step": 7161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2683639465714575e-05,
+      "loss": 2.1025,
+      "step": 7162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.268162334949989e-05,
+      "loss": 2.0874,
+      "step": 7163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.267960705151143e-05,
+      "loss": 2.0823,
+      "step": 7164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2677590571802063e-05,
+      "loss": 2.0654,
+      "step": 7165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2675573910424672e-05,
+      "loss": 2.0623,
+      "step": 7166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.267355706743214e-05,
+      "loss": 2.0662,
+      "step": 7167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.267154004287735e-05,
+      "loss": 2.0798,
+      "step": 7168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2669522836813207e-05,
+      "loss": 2.0486,
+      "step": 7169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2667505449292604e-05,
+      "loss": 2.0916,
+      "step": 7170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2665487880368447e-05,
+      "loss": 2.0833,
+      "step": 7171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.266347013009364e-05,
+      "loss": 2.0825,
+      "step": 7172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2661452198521106e-05,
+      "loss": 2.0686,
+      "step": 7173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2659434085703747e-05,
+      "loss": 2.0862,
+      "step": 7174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2657415791694498e-05,
+      "loss": 2.0505,
+      "step": 7175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2655397316546284e-05,
+      "loss": 2.0676,
+      "step": 7176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2653378660312038e-05,
+      "loss": 2.0898,
+      "step": 7177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2651359823044692e-05,
+      "loss": 2.0925,
+      "step": 7178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2649340804797194e-05,
+      "loss": 2.0457,
+      "step": 7179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2647321605622486e-05,
+      "loss": 1.9746,
+      "step": 7180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2645302225573517e-05,
+      "loss": 2.082,
+      "step": 7181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2643282664703255e-05,
+      "loss": 2.0828,
+      "step": 7182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.264126292306465e-05,
+      "loss": 2.0732,
+      "step": 7183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2639243000710667e-05,
+      "loss": 2.0737,
+      "step": 7184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2637222897694273e-05,
+      "loss": 2.0667,
+      "step": 7185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2635202614068456e-05,
+      "loss": 2.0698,
+      "step": 7186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.263318214988619e-05,
+      "loss": 2.0542,
+      "step": 7187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2631161505200453e-05,
+      "loss": 2.0605,
+      "step": 7188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2629140680064237e-05,
+      "loss": 2.0779,
+      "step": 7189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2627119674530538e-05,
+      "loss": 2.053,
+      "step": 7190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2625098488652358e-05,
+      "loss": 2.0715,
+      "step": 7191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2623077122482693e-05,
+      "loss": 2.1033,
+      "step": 7192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2621055576074554e-05,
+      "loss": 2.0278,
+      "step": 7193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2619033849480952e-05,
+      "loss": 2.01,
+      "step": 7194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2617011942754908e-05,
+      "loss": 2.0723,
+      "step": 7195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2614989855949442e-05,
+      "loss": 2.0544,
+      "step": 7196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2612967589117583e-05,
+      "loss": 2.0522,
+      "step": 7197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.261094514231236e-05,
+      "loss": 2.0195,
+      "step": 7198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.260892251558681e-05,
+      "loss": 2.0964,
+      "step": 7199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2606899708993973e-05,
+      "loss": 2.093,
+      "step": 7200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2604876722586896e-05,
+      "loss": 2.0261,
+      "step": 7201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.260285355641863e-05,
+      "loss": 2.0266,
+      "step": 7202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.260083021054223e-05,
+      "loss": 2.0745,
+      "step": 7203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2598806685010755e-05,
+      "loss": 2.0627,
+      "step": 7204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.259678297987727e-05,
+      "loss": 2.0388,
+      "step": 7205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2594759095194845e-05,
+      "loss": 2.0618,
+      "step": 7206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2592735031016555e-05,
+      "loss": 2.0801,
+      "step": 7207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2590710787395473e-05,
+      "loss": 2.0603,
+      "step": 7208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2588686364384688e-05,
+      "loss": 2.0591,
+      "step": 7209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.258666176203729e-05,
+      "loss": 2.0715,
+      "step": 7210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2584636980406365e-05,
+      "loss": 2.033,
+      "step": 7211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2582612019545013e-05,
+      "loss": 2.0281,
+      "step": 7212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.258058687950634e-05,
+      "loss": 2.0425,
+      "step": 7213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2578561560343453e-05,
+      "loss": 2.0754,
+      "step": 7214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2576536062109455e-05,
+      "loss": 2.0625,
+      "step": 7215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.257451038485747e-05,
+      "loss": 2.0544,
+      "step": 7216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2572484528640622e-05,
+      "loss": 2.0942,
+      "step": 7217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.257045849351203e-05,
+      "loss": 2.1064,
+      "step": 7218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2568432279524823e-05,
+      "loss": 2.0239,
+      "step": 7219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.256640588673214e-05,
+      "loss": 2.0332,
+      "step": 7220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2564379315187122e-05,
+      "loss": 2.1028,
+      "step": 7221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2562352564942913e-05,
+      "loss": 2.0959,
+      "step": 7222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2560325636052663e-05,
+      "loss": 2.0576,
+      "step": 7223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.255829852856952e-05,
+      "loss": 2.0547,
+      "step": 7224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2556271242546648e-05,
+      "loss": 2.0574,
+      "step": 7225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2554243778037207e-05,
+      "loss": 2.0623,
+      "step": 7226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.255221613509437e-05,
+      "loss": 2.0876,
+      "step": 7227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.25501883137713e-05,
+      "loss": 2.0979,
+      "step": 7228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2548160314121185e-05,
+      "loss": 2.1143,
+      "step": 7229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2546132136197204e-05,
+      "loss": 2.0903,
+      "step": 7230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2544103780052538e-05,
+      "loss": 2.0974,
+      "step": 7231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.254207524574039e-05,
+      "loss": 2.1118,
+      "step": 7232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.254004653331394e-05,
+      "loss": 2.0308,
+      "step": 7233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.25380176428264e-05,
+      "loss": 1.9839,
+      "step": 7234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.253598857433097e-05,
+      "loss": 2.0444,
+      "step": 7235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2533959327880863e-05,
+      "loss": 2.0647,
+      "step": 7236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2531929903529296e-05,
+      "loss": 2.0215,
+      "step": 7237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2529900301329483e-05,
+      "loss": 2.002,
+      "step": 7238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.252787052133465e-05,
+      "loss": 2.0918,
+      "step": 7239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2525840563598023e-05,
+      "loss": 2.0994,
+      "step": 7240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.252381042817284e-05,
+      "loss": 2.0486,
+      "step": 7241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.252178011511234e-05,
+      "loss": 2.0532,
+      "step": 7242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2519749624469763e-05,
+      "loss": 1.9802,
+      "step": 7243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.251771895629835e-05,
+      "loss": 1.9988,
+      "step": 7244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2515688110651362e-05,
+      "loss": 2.0569,
+      "step": 7245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.251365708758205e-05,
+      "loss": 2.0447,
+      "step": 7246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.251162588714368e-05,
+      "loss": 2.0496,
+      "step": 7247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2509594509389513e-05,
+      "loss": 2.0837,
+      "step": 7248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2507562954372825e-05,
+      "loss": 2.0874,
+      "step": 7249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2505531222146884e-05,
+      "loss": 2.0991,
+      "step": 7250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2503499312764977e-05,
+      "loss": 2.0054,
+      "step": 7251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2501467226280384e-05,
+      "loss": 2.0251,
+      "step": 7252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2499434962746403e-05,
+      "loss": 2.0427,
+      "step": 7253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2497402522216308e-05,
+      "loss": 2.0339,
+      "step": 7254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2495369904743416e-05,
+      "loss": 2.0803,
+      "step": 7255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2493337110381022e-05,
+      "loss": 2.0881,
+      "step": 7256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2491304139182438e-05,
+      "loss": 2.0645,
+      "step": 7257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2489270991200973e-05,
+      "loss": 2.0601,
+      "step": 7258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2487237666489943e-05,
+      "loss": 2.0605,
+      "step": 7259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.248520416510267e-05,
+      "loss": 2.0493,
+      "step": 7260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2483170487092483e-05,
+      "loss": 2.0676,
+      "step": 7261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.248113663251271e-05,
+      "loss": 2.0676,
+      "step": 7262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2479102601416688e-05,
+      "loss": 2.0393,
+      "step": 7263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.247706839385776e-05,
+      "loss": 2.0337,
+      "step": 7264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2475034009889263e-05,
+      "loss": 2.0667,
+      "step": 7265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2472999449564553e-05,
+      "loss": 2.0676,
+      "step": 7266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.247096471293698e-05,
+      "loss": 2.0591,
+      "step": 7267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.246892980005991e-05,
+      "loss": 2.0776,
+      "step": 7268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2466894710986696e-05,
+      "loss": 2.0437,
+      "step": 7269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.246485944577071e-05,
+      "loss": 2.0259,
+      "step": 7270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2462824004465324e-05,
+      "loss": 2.0625,
+      "step": 7271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2460788387123924e-05,
+      "loss": 2.0574,
+      "step": 7272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.245875259379988e-05,
+      "loss": 2.0796,
+      "step": 7273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2456716624546576e-05,
+      "loss": 2.0784,
+      "step": 7274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2454680479417415e-05,
+      "loss": 2.0808,
+      "step": 7275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2452644158465786e-05,
+      "loss": 2.0781,
+      "step": 7276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2450607661745087e-05,
+      "loss": 2.0554,
+      "step": 7277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.244857098930873e-05,
+      "loss": 2.0337,
+      "step": 7278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.244653414121012e-05,
+      "loss": 2.0977,
+      "step": 7279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2444497117502667e-05,
+      "loss": 2.1074,
+      "step": 7280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2442459918239794e-05,
+      "loss": 2.0723,
+      "step": 7281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2440422543474923e-05,
+      "loss": 2.0295,
+      "step": 7282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2438384993261484e-05,
+      "loss": 2.0713,
+      "step": 7283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2436347267652907e-05,
+      "loss": 2.0825,
+      "step": 7284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.243430936670263e-05,
+      "loss": 2.0938,
+      "step": 7285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2432271290464094e-05,
+      "loss": 2.1008,
+      "step": 7286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.243023303899075e-05,
+      "loss": 2.1055,
+      "step": 7287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2428194612336042e-05,
+      "loss": 2.0996,
+      "step": 7288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.242615601055343e-05,
+      "loss": 2.0586,
+      "step": 7289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2424117233696367e-05,
+      "loss": 2.0581,
+      "step": 7290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2422078281818324e-05,
+      "loss": 2.0728,
+      "step": 7291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2420039154972767e-05,
+      "loss": 2.0786,
+      "step": 7292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2417999853213172e-05,
+      "loss": 2.0806,
+      "step": 7293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2415960376593018e-05,
+      "loss": 2.0686,
+      "step": 7294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2413920725165788e-05,
+      "loss": 2.0659,
+      "step": 7295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2411880898984964e-05,
+      "loss": 2.0486,
+      "step": 7296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.240984089810404e-05,
+      "loss": 2.1138,
+      "step": 7297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.240780072257652e-05,
+      "loss": 2.0864,
+      "step": 7298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2405760372455897e-05,
+      "loss": 2.1086,
+      "step": 7299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.240371984779568e-05,
+      "loss": 2.0781,
+      "step": 7300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2401679148649378e-05,
+      "loss": 2.0422,
+      "step": 7301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.239963827507051e-05,
+      "loss": 2.0637,
+      "step": 7302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.239759722711259e-05,
+      "loss": 2.1082,
+      "step": 7303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2395556004829147e-05,
+      "loss": 2.1179,
+      "step": 7304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2393514608273706e-05,
+      "loss": 2.0679,
+      "step": 7305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2391473037499802e-05,
+      "loss": 2.0801,
+      "step": 7306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2389431292560973e-05,
+      "loss": 2.0952,
+      "step": 7307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2387389373510757e-05,
+      "loss": 2.0496,
+      "step": 7308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.238534728040271e-05,
+      "loss": 2.105,
+      "step": 7309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2383305013290377e-05,
+      "loss": 2.0818,
+      "step": 7310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2381262572227315e-05,
+      "loss": 2.0815,
+      "step": 7311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2379219957267085e-05,
+      "loss": 2.0625,
+      "step": 7312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2377177168463252e-05,
+      "loss": 2.0962,
+      "step": 7313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2375134205869384e-05,
+      "loss": 2.0818,
+      "step": 7314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.237309106953906e-05,
+      "loss": 2.0398,
+      "step": 7315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2371047759525857e-05,
+      "loss": 2.0637,
+      "step": 7316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2369004275883357e-05,
+      "loss": 2.0918,
+      "step": 7317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2366960618665145e-05,
+      "loss": 2.1069,
+      "step": 7318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.236491678792482e-05,
+      "loss": 2.0288,
+      "step": 7319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2362872783715976e-05,
+      "loss": 2.0364,
+      "step": 7320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2360828606092217e-05,
+      "loss": 2.0496,
+      "step": 7321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2358784255107147e-05,
+      "loss": 2.0618,
+      "step": 7322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2356739730814374e-05,
+      "loss": 2.0972,
+      "step": 7323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2354695033267518e-05,
+      "loss": 2.0935,
+      "step": 7324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2352650162520195e-05,
+      "loss": 2.0896,
+      "step": 7325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2350605118626035e-05,
+      "loss": 2.1084,
+      "step": 7326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.234855990163866e-05,
+      "loss": 2.115,
+      "step": 7327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2346514511611707e-05,
+      "loss": 2.1082,
+      "step": 7328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2344468948598816e-05,
+      "loss": 2.0247,
+      "step": 7329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2342423212653626e-05,
+      "loss": 2.0264,
+      "step": 7330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2340377303829784e-05,
+      "loss": 2.0774,
+      "step": 7331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.233833122218094e-05,
+      "loss": 2.1121,
+      "step": 7332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2336284967760758e-05,
+      "loss": 2.061,
+      "step": 7333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2334238540622893e-05,
+      "loss": 2.085,
+      "step": 7334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.233219194082101e-05,
+      "loss": 2.0361,
+      "step": 7335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2330145168408776e-05,
+      "loss": 2.0486,
+      "step": 7336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.232809822343987e-05,
+      "loss": 2.0374,
+      "step": 7337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2326051105967974e-05,
+      "loss": 2.0603,
+      "step": 7338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2324003816046764e-05,
+      "loss": 2.0547,
+      "step": 7339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.232195635372993e-05,
+      "loss": 2.0537,
+      "step": 7340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2319908719071166e-05,
+      "loss": 2.1028,
+      "step": 7341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2317860912124166e-05,
+      "loss": 2.0894,
+      "step": 7342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2315812932942637e-05,
+      "loss": 2.0291,
+      "step": 7343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.231376478158028e-05,
+      "loss": 2.0625,
+      "step": 7344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2311716458090802e-05,
+      "loss": 2.075,
+      "step": 7345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2309667962527923e-05,
+      "loss": 2.0073,
+      "step": 7346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2307619294945365e-05,
+      "loss": 2.0908,
+      "step": 7347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2305570455396846e-05,
+      "loss": 2.043,
+      "step": 7348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.23035214439361e-05,
+      "loss": 2.0325,
+      "step": 7349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2301472260616852e-05,
+      "loss": 2.0322,
+      "step": 7350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.229942290549285e-05,
+      "loss": 2.0449,
+      "step": 7351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2297373378617822e-05,
+      "loss": 2.0781,
+      "step": 7352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.229532368004553e-05,
+      "loss": 2.0603,
+      "step": 7353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2293273809829717e-05,
+      "loss": 2.0427,
+      "step": 7354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2291223768024137e-05,
+      "loss": 2.0576,
+      "step": 7355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2289173554682554e-05,
+      "loss": 2.0513,
+      "step": 7356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.228712316985873e-05,
+      "loss": 2.103,
+      "step": 7357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2285072613606437e-05,
+      "loss": 2.1074,
+      "step": 7358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.228302188597944e-05,
+      "loss": 2.0686,
+      "step": 7359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.228097098703153e-05,
+      "loss": 2.0952,
+      "step": 7360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.227891991681648e-05,
+      "loss": 2.0676,
+      "step": 7361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.227686867538808e-05,
+      "loss": 2.0925,
+      "step": 7362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.227481726280012e-05,
+      "loss": 2.0498,
+      "step": 7363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2272765679106398e-05,
+      "loss": 2.1055,
+      "step": 7364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2270713924360712e-05,
+      "loss": 2.0806,
+      "step": 7365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2268661998616868e-05,
+      "loss": 2.0828,
+      "step": 7366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2266609901928674e-05,
+      "loss": 2.0894,
+      "step": 7367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.226455763434995e-05,
+      "loss": 2.1064,
+      "step": 7368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2262505195934505e-05,
+      "loss": 2.0686,
+      "step": 7369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.226045258673617e-05,
+      "loss": 2.1328,
+      "step": 7370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2258399806808763e-05,
+      "loss": 2.0342,
+      "step": 7371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2256346856206126e-05,
+      "loss": 2.0488,
+      "step": 7372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.225429373498209e-05,
+      "loss": 2.0461,
+      "step": 7373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.22522404431905e-05,
+      "loss": 2.05,
+      "step": 7374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2250186980885194e-05,
+      "loss": 2.0725,
+      "step": 7375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2248133348120026e-05,
+      "loss": 2.0627,
+      "step": 7376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.224607954494885e-05,
+      "loss": 2.0933,
+      "step": 7377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2244025571425524e-05,
+      "loss": 2.116,
+      "step": 7378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.224197142760391e-05,
+      "loss": 2.0715,
+      "step": 7379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.223991711353788e-05,
+      "loss": 2.0989,
+      "step": 7380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.22378626292813e-05,
+      "loss": 2.1177,
+      "step": 7381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2235807974888048e-05,
+      "loss": 2.1052,
+      "step": 7382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2233753150412006e-05,
+      "loss": 2.0793,
+      "step": 7383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2231698155907063e-05,
+      "loss": 2.0681,
+      "step": 7384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2229642991427107e-05,
+      "loss": 2.0437,
+      "step": 7385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2227587657026023e-05,
+      "loss": 2.063,
+      "step": 7386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2225532152757723e-05,
+      "loss": 2.1016,
+      "step": 7387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2223476478676102e-05,
+      "loss": 2.1074,
+      "step": 7388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2221420634835075e-05,
+      "loss": 2.1084,
+      "step": 7389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2219364621288544e-05,
+      "loss": 2.0852,
+      "step": 7390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2217308438090435e-05,
+      "loss": 2.0928,
+      "step": 7391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2215252085294662e-05,
+      "loss": 2.0801,
+      "step": 7392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.221319556295515e-05,
+      "loss": 2.0986,
+      "step": 7393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.221113887112584e-05,
+      "loss": 2.0745,
+      "step": 7394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2209082009860654e-05,
+      "loss": 2.0405,
+      "step": 7395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.220702497921354e-05,
+      "loss": 2.0566,
+      "step": 7396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2204967779238434e-05,
+      "loss": 2.0718,
+      "step": 7397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2202910409989287e-05,
+      "loss": 2.114,
+      "step": 7398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2200852871520048e-05,
+      "loss": 2.0593,
+      "step": 7399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.219879516388468e-05,
+      "loss": 2.0515,
+      "step": 7400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.219673728713714e-05,
+      "loss": 2.0872,
+      "step": 7401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.219467924133139e-05,
+      "loss": 2.071,
+      "step": 7402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2192621026521407e-05,
+      "loss": 2.1003,
+      "step": 7403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.219056264276116e-05,
+      "loss": 2.0911,
+      "step": 7404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.218850409010463e-05,
+      "loss": 2.0183,
+      "step": 7405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2186445368605796e-05,
+      "loss": 2.0173,
+      "step": 7406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2184386478318657e-05,
+      "loss": 2.0332,
+      "step": 7407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.218232741929719e-05,
+      "loss": 2.0374,
+      "step": 7408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.21802681915954e-05,
+      "loss": 2.0925,
+      "step": 7409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2178208795267293e-05,
+      "loss": 2.0647,
+      "step": 7410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.217614923036686e-05,
+      "loss": 2.0652,
+      "step": 7411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2174089496948123e-05,
+      "loss": 2.0317,
+      "step": 7412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2172029595065087e-05,
+      "loss": 2.0337,
+      "step": 7413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.216996952477178e-05,
+      "loss": 2.0095,
+      "step": 7414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.216790928612222e-05,
+      "loss": 2.1396,
+      "step": 7415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2165848879170433e-05,
+      "loss": 2.1282,
+      "step": 7416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.216378830397045e-05,
+      "loss": 2.0139,
+      "step": 7417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2161727560576314e-05,
+      "loss": 2.0144,
+      "step": 7418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2159666649042056e-05,
+      "loss": 2.0618,
+      "step": 7419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2157605569421734e-05,
+      "loss": 2.0435,
+      "step": 7420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2155544321769384e-05,
+      "loss": 2.0486,
+      "step": 7421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2153482906139064e-05,
+      "loss": 2.0696,
+      "step": 7422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2151421322584834e-05,
+      "loss": 2.0825,
+      "step": 7423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2149359571160757e-05,
+      "loss": 2.0479,
+      "step": 7424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2147297651920896e-05,
+      "loss": 2.0664,
+      "step": 7425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.214523556491933e-05,
+      "loss": 2.0437,
+      "step": 7426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.214317331021013e-05,
+      "loss": 2.0447,
+      "step": 7427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2141110887847377e-05,
+      "loss": 2.0391,
+      "step": 7428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2139048297885154e-05,
+      "loss": 2.0774,
+      "step": 7429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.213698554037755e-05,
+      "loss": 2.094,
+      "step": 7430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2134922615378666e-05,
+      "loss": 2.127,
+      "step": 7431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2132859522942587e-05,
+      "loss": 2.1184,
+      "step": 7432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2130796263123427e-05,
+      "loss": 2.0754,
+      "step": 7433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2128732835975286e-05,
+      "loss": 2.0308,
+      "step": 7434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2126669241552273e-05,
+      "loss": 2.1362,
+      "step": 7435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2124605479908514e-05,
+      "loss": 2.1208,
+      "step": 7436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2122541551098114e-05,
+      "loss": 2.0164,
+      "step": 7437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2120477455175204e-05,
+      "loss": 2.0554,
+      "step": 7438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.211841319219392e-05,
+      "loss": 2.0542,
+      "step": 7439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2116348762208383e-05,
+      "loss": 2.0408,
+      "step": 7440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.211428416527274e-05,
+      "loss": 2.0198,
+      "step": 7441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.211221940144112e-05,
+      "loss": 2.0564,
+      "step": 7442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2110154470767683e-05,
+      "loss": 2.0542,
+      "step": 7443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2108089373306572e-05,
+      "loss": 2.0461,
+      "step": 7444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2106024109111945e-05,
+      "loss": 2.0835,
+      "step": 7445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2103958678237962e-05,
+      "loss": 2.0928,
+      "step": 7446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.210189308073878e-05,
+      "loss": 2.0803,
+      "step": 7447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2099827316668572e-05,
+      "loss": 2.072,
+      "step": 7448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2097761386081507e-05,
+      "loss": 2.075,
+      "step": 7449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2095695289031764e-05,
+      "loss": 2.0876,
+      "step": 7450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2093629025573528e-05,
+      "loss": 2.0361,
+      "step": 7451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2091562595760976e-05,
+      "loss": 2.0779,
+      "step": 7452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2089495999648307e-05,
+      "loss": 2.0671,
+      "step": 7453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2087429237289704e-05,
+      "loss": 2.0605,
+      "step": 7454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2085362308739372e-05,
+      "loss": 2.0891,
+      "step": 7455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2083295214051516e-05,
+      "loss": 2.0969,
+      "step": 7456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2081227953280337e-05,
+      "loss": 2.0923,
+      "step": 7457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2079160526480055e-05,
+      "loss": 2.0828,
+      "step": 7458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.207709293370488e-05,
+      "loss": 2.0345,
+      "step": 7459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2075025175009024e-05,
+      "loss": 2.0747,
+      "step": 7460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.207295725044673e-05,
+      "loss": 2.0681,
+      "step": 7461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.207088916007221e-05,
+      "loss": 2.0906,
+      "step": 7462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2068820903939713e-05,
+      "loss": 2.0652,
+      "step": 7463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.206675248210346e-05,
+      "loss": 2.0798,
+      "step": 7464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.206468389461771e-05,
+      "loss": 2.0217,
+      "step": 7465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2062615141536697e-05,
+      "loss": 2.0574,
+      "step": 7466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2060546222914676e-05,
+      "loss": 2.063,
+      "step": 7467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.20584771388059e-05,
+      "loss": 2.0339,
+      "step": 7468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.205640788926463e-05,
+      "loss": 2.0581,
+      "step": 7469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.205433847434513e-05,
+      "loss": 2.0615,
+      "step": 7470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2052268894101668e-05,
+      "loss": 2.0654,
+      "step": 7471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2050199148588514e-05,
+      "loss": 2.0708,
+      "step": 7472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2048129237859943e-05,
+      "loss": 2.0415,
+      "step": 7473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2046059161970243e-05,
+      "loss": 2.0713,
+      "step": 7474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2043988920973694e-05,
+      "loss": 2.0781,
+      "step": 7475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2041918514924592e-05,
+      "loss": 2.0801,
+      "step": 7476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.203984794387722e-05,
+      "loss": 2.0442,
+      "step": 7477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2037777207885885e-05,
+      "loss": 2.084,
+      "step": 7478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2035706307004886e-05,
+      "loss": 2.0872,
+      "step": 7479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.203363524128853e-05,
+      "loss": 2.0632,
+      "step": 7480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2031564010791134e-05,
+      "loss": 2.0781,
+      "step": 7481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.202949261556701e-05,
+      "loss": 2.0938,
+      "step": 7482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2027421055670474e-05,
+      "loss": 2.0376,
+      "step": 7483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2025349331155853e-05,
+      "loss": 2.0752,
+      "step": 7484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2023277442077476e-05,
+      "loss": 2.021,
+      "step": 7485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2021205388489675e-05,
+      "loss": 2.0408,
+      "step": 7486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2019133170446794e-05,
+      "loss": 2.0913,
+      "step": 7487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.201706078800316e-05,
+      "loss": 2.0669,
+      "step": 7488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2014988241213132e-05,
+      "loss": 2.0413,
+      "step": 7489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2012915530131055e-05,
+      "loss": 2.0154,
+      "step": 7490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2010842654811286e-05,
+      "loss": 2.0793,
+      "step": 7491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2008769615308182e-05,
+      "loss": 2.1138,
+      "step": 7492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.20066964116761e-05,
+      "loss": 2.0486,
+      "step": 7493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2004623043969418e-05,
+      "loss": 2.0691,
+      "step": 7494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.20025495122425e-05,
+      "loss": 2.1091,
+      "step": 7495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2000475816549728e-05,
+      "loss": 2.092,
+      "step": 7496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.1998401956945473e-05,
+      "loss": 2.0493,
+      "step": 7497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.199632793348413e-05,
+      "loss": 2.0559,
+      "step": 7498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.199425374622008e-05,
+      "loss": 2.0918,
+      "step": 7499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1992179395207723e-05,
+      "loss": 2.0654,
+      "step": 7500
+    },
+    {
+      "Bleu_1": 0.7736231992292749,
+      "Bleu_2": 0.6125302563387922,
+      "Bleu_3": 0.4708661294396024,
+      "Bleu_4": 0.3576380778699701,
+      "CIDEr": 1.205359170158909,
+      "METEOR": 0.27884634798738794,
+      "ROUGE_L": 0.5600115374210025,
+      "epoch": 0.38,
+      "step": 7500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1990104880501454e-05,
+      "loss": 2.0767,
+      "step": 7501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1988030202155674e-05,
+      "loss": 2.0774,
+      "step": 7502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1985955360224786e-05,
+      "loss": 2.0583,
+      "step": 7503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1983880354763202e-05,
+      "loss": 2.1062,
+      "step": 7504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1981805185825338e-05,
+      "loss": 2.0598,
+      "step": 7505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1979729853465615e-05,
+      "loss": 2.0625,
+      "step": 7506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1977654357738455e-05,
+      "loss": 2.0962,
+      "step": 7507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.197557869869828e-05,
+      "loss": 2.1016,
+      "step": 7508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.197350287639953e-05,
+      "loss": 2.0198,
+      "step": 7509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1971426890896635e-05,
+      "loss": 2.0225,
+      "step": 7510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1969350742244035e-05,
+      "loss": 2.1057,
+      "step": 7511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.196727443049618e-05,
+      "loss": 2.0945,
+      "step": 7512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1965197955707514e-05,
+      "loss": 2.0662,
+      "step": 7513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1963121317932493e-05,
+      "loss": 2.0496,
+      "step": 7514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1961044517225577e-05,
+      "loss": 2.0532,
+      "step": 7515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1958967553641216e-05,
+      "loss": 2.0413,
+      "step": 7516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1956890427233894e-05,
+      "loss": 2.0376,
+      "step": 7517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1954813138058064e-05,
+      "loss": 2.0303,
+      "step": 7518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.195273568616821e-05,
+      "loss": 2.0205,
+      "step": 7519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1950658071618806e-05,
+      "loss": 2.0203,
+      "step": 7520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1948580294464344e-05,
+      "loss": 2.0818,
+      "step": 7521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.19465023547593e-05,
+      "loss": 2.0549,
+      "step": 7522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.194442425255817e-05,
+      "loss": 2.0386,
+      "step": 7523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1942345987915453e-05,
+      "loss": 2.0179,
+      "step": 7524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.194026756088564e-05,
+      "loss": 2.084,
+      "step": 7525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.193818897152325e-05,
+      "loss": 2.0928,
+      "step": 7526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.193611021988278e-05,
+      "loss": 2.0315,
+      "step": 7527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1934031306018745e-05,
+      "loss": 2.0627,
+      "step": 7528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1931952229985662e-05,
+      "loss": 2.064,
+      "step": 7529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1929872991838056e-05,
+      "loss": 2.0615,
+      "step": 7530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1927793591630448e-05,
+      "loss": 2.0417,
+      "step": 7531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.192571402941737e-05,
+      "loss": 2.0205,
+      "step": 7532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.192363430525336e-05,
+      "loss": 2.0793,
+      "step": 7533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1921554419192948e-05,
+      "loss": 2.0674,
+      "step": 7534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.191947437129068e-05,
+      "loss": 2.033,
+      "step": 7535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.191739416160111e-05,
+      "loss": 2.0601,
+      "step": 7536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.191531379017878e-05,
+      "loss": 2.0803,
+      "step": 7537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.191323325707825e-05,
+      "loss": 2.0303,
+      "step": 7538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1911152562354074e-05,
+      "loss": 2.0046,
+      "step": 7539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.190907170606082e-05,
+      "loss": 2.0073,
+      "step": 7540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.190699068825306e-05,
+      "loss": 2.0266,
+      "step": 7541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1904909508985356e-05,
+      "loss": 2.0718,
+      "step": 7542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1902828168312297e-05,
+      "loss": 2.0381,
+      "step": 7543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1900746666288454e-05,
+      "loss": 2.0247,
+      "step": 7544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1898665002968415e-05,
+      "loss": 2.0769,
+      "step": 7545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.189658317840677e-05,
+      "loss": 2.1067,
+      "step": 7546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.189450119265811e-05,
+      "loss": 2.009,
+      "step": 7547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1892419045777038e-05,
+      "loss": 2.0203,
+      "step": 7548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.189033673781815e-05,
+      "loss": 2.0452,
+      "step": 7549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1888254268836053e-05,
+      "loss": 2.0449,
+      "step": 7550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1886171638885362e-05,
+      "loss": 2.0654,
+      "step": 7551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.188408884802069e-05,
+      "loss": 2.0984,
+      "step": 7552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1882005896296653e-05,
+      "loss": 2.0862,
+      "step": 7553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1879922783767874e-05,
+      "loss": 2.071,
+      "step": 7554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.187783951048898e-05,
+      "loss": 2.0884,
+      "step": 7555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1875756076514605e-05,
+      "loss": 2.073,
+      "step": 7556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1873672481899382e-05,
+      "loss": 2.05,
+      "step": 7557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1871588726697954e-05,
+      "loss": 2.0603,
+      "step": 7558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1869504810964965e-05,
+      "loss": 2.0247,
+      "step": 7559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.186742073475506e-05,
+      "loss": 2.0029,
+      "step": 7560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1865336498122892e-05,
+      "loss": 2.0686,
+      "step": 7561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.186325210112312e-05,
+      "loss": 2.0942,
+      "step": 7562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1861167543810405e-05,
+      "loss": 2.1262,
+      "step": 7563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1859082826239413e-05,
+      "loss": 2.124,
+      "step": 7564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1856997948464807e-05,
+      "loss": 2.0942,
+      "step": 7565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1854912910541264e-05,
+      "loss": 2.0435,
+      "step": 7566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.185282771252346e-05,
+      "loss": 1.9941,
+      "step": 7567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1850742354466086e-05,
+      "loss": 2.0356,
+      "step": 7568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1848656836423818e-05,
+      "loss": 2.0984,
+      "step": 7569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1846571158451346e-05,
+      "loss": 2.0989,
+      "step": 7570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1844485320603376e-05,
+      "loss": 2.0427,
+      "step": 7571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.184239932293459e-05,
+      "loss": 2.0774,
+      "step": 7572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1840313165499708e-05,
+      "loss": 2.0771,
+      "step": 7573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1838226848353426e-05,
+      "loss": 2.0442,
+      "step": 7574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1836140371550455e-05,
+      "loss": 2.0623,
+      "step": 7575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1834053735145515e-05,
+      "loss": 2.0544,
+      "step": 7576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.183196693919332e-05,
+      "loss": 2.075,
+      "step": 7577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1829879983748607e-05,
+      "loss": 2.0862,
+      "step": 7578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.182779286886609e-05,
+      "loss": 2.0117,
+      "step": 7579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1825705594600507e-05,
+      "loss": 2.0154,
+      "step": 7580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1823618161006592e-05,
+      "loss": 2.0667,
+      "step": 7581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1821530568139085e-05,
+      "loss": 2.0579,
+      "step": 7582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1819442816052736e-05,
+      "loss": 2.073,
+      "step": 7583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1817354904802285e-05,
+      "loss": 2.0691,
+      "step": 7584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1815266834442495e-05,
+      "loss": 2.043,
+      "step": 7585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1813178605028116e-05,
+      "loss": 2.0735,
+      "step": 7586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1811090216613915e-05,
+      "loss": 2.0088,
+      "step": 7587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1809001669254657e-05,
+      "loss": 1.9897,
+      "step": 7588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1806912963005106e-05,
+      "loss": 2.0481,
+      "step": 7589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1804824097920035e-05,
+      "loss": 2.0662,
+      "step": 7590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1802735074054233e-05,
+      "loss": 2.0205,
+      "step": 7591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.180064589146247e-05,
+      "loss": 2.0229,
+      "step": 7592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1798556550199546e-05,
+      "loss": 2.103,
+      "step": 7593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1796467050320235e-05,
+      "loss": 2.1174,
+      "step": 7594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1794377391879345e-05,
+      "loss": 2.0593,
+      "step": 7595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.179228757493167e-05,
+      "loss": 2.0867,
+      "step": 7596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.179019759953201e-05,
+      "loss": 2.0876,
+      "step": 7597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.178810746573518e-05,
+      "loss": 2.1331,
+      "step": 7598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1786017173595985e-05,
+      "loss": 2.0959,
+      "step": 7599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.178392672316924e-05,
+      "loss": 2.1265,
+      "step": 7600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1781836114509765e-05,
+      "loss": 2.061,
+      "step": 7601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1779745347672385e-05,
+      "loss": 2.071,
+      "step": 7602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1777654422711932e-05,
+      "loss": 2.0811,
+      "step": 7603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1775563339683235e-05,
+      "loss": 2.0967,
+      "step": 7604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1773472098641128e-05,
+      "loss": 2.064,
+      "step": 7605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1771380699640454e-05,
+      "loss": 2.0474,
+      "step": 7606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1769289142736055e-05,
+      "loss": 2.0349,
+      "step": 7607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1767197427982782e-05,
+      "loss": 2.0327,
+      "step": 7608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1765105555435495e-05,
+      "loss": 2.1206,
+      "step": 7609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1763013525149033e-05,
+      "loss": 2.1296,
+      "step": 7610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1760921337178268e-05,
+      "loss": 2.1116,
+      "step": 7611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1758828991578065e-05,
+      "loss": 2.0989,
+      "step": 7612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1756736488403297e-05,
+      "loss": 2.0251,
+      "step": 7613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1754643827708833e-05,
+      "loss": 2.0369,
+      "step": 7614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1752551009549543e-05,
+      "loss": 2.0762,
+      "step": 7615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1750458033980326e-05,
+      "loss": 2.0614,
+      "step": 7616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.174836490105605e-05,
+      "loss": 2.083,
+      "step": 7617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1746271610831616e-05,
+      "loss": 2.0757,
+      "step": 7618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1744178163361916e-05,
+      "loss": 2.0581,
+      "step": 7619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1742084558701847e-05,
+      "loss": 2.0496,
+      "step": 7620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.173999079690631e-05,
+      "loss": 2.0322,
+      "step": 7621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.173789687803022e-05,
+      "loss": 2.0181,
+      "step": 7622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1735802802128473e-05,
+      "loss": 2.0784,
+      "step": 7623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1733708569255997e-05,
+      "loss": 2.0657,
+      "step": 7624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1731614179467706e-05,
+      "loss": 2.0361,
+      "step": 7625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.172951963281852e-05,
+      "loss": 2.0007,
+      "step": 7626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1727424929363368e-05,
+      "loss": 2.064,
+      "step": 7627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1725330069157186e-05,
+      "loss": 2.0688,
+      "step": 7628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1723235052254904e-05,
+      "loss": 2.0391,
+      "step": 7629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.172113987871146e-05,
+      "loss": 2.05,
+      "step": 7630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1719044548581797e-05,
+      "loss": 2.0371,
+      "step": 7631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.171694906192087e-05,
+      "loss": 2.0537,
+      "step": 7632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1714853418783622e-05,
+      "loss": 2.0803,
+      "step": 7633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.171275761922502e-05,
+      "loss": 2.0876,
+      "step": 7634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1710661663300014e-05,
+      "loss": 2.0786,
+      "step": 7635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1708565551063564e-05,
+      "loss": 2.0845,
+      "step": 7636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.170646928257065e-05,
+      "loss": 2.0742,
+      "step": 7637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.170437285787624e-05,
+      "loss": 2.0857,
+      "step": 7638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1702276277035314e-05,
+      "loss": 2.0754,
+      "step": 7639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1700179540102843e-05,
+      "loss": 2.1096,
+      "step": 7640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1698082647133814e-05,
+      "loss": 1.9153,
+      "step": 7641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1695985598183225e-05,
+      "loss": 1.9397,
+      "step": 7642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1693888393306057e-05,
+      "loss": 2.1128,
+      "step": 7643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1691791032557314e-05,
+      "loss": 2.0903,
+      "step": 7644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.168969351599199e-05,
+      "loss": 2.0725,
+      "step": 7645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.16875958436651e-05,
+      "loss": 2.095,
+      "step": 7646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.168549801563164e-05,
+      "loss": 2.063,
+      "step": 7647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1683400031946637e-05,
+      "loss": 2.074,
+      "step": 7648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.16813018926651e-05,
+      "loss": 2.1035,
+      "step": 7649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.167920359784205e-05,
+      "loss": 2.0901,
+      "step": 7650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1677105147532516e-05,
+      "loss": 2.1008,
+      "step": 7651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1675006541791524e-05,
+      "loss": 2.0957,
+      "step": 7652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.167290778067411e-05,
+      "loss": 2.1013,
+      "step": 7653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.167080886423531e-05,
+      "loss": 2.0942,
+      "step": 7654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1668709792530168e-05,
+      "loss": 2.0671,
+      "step": 7655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1666610565613722e-05,
+      "loss": 2.0547,
+      "step": 7656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1664511183541037e-05,
+      "loss": 2.1118,
+      "step": 7657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1662411646367154e-05,
+      "loss": 2.0933,
+      "step": 7658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1660311954147135e-05,
+      "loss": 2.0369,
+      "step": 7659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1658212106936038e-05,
+      "loss": 2.0283,
+      "step": 7660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.165611210478894e-05,
+      "loss": 2.0474,
+      "step": 7661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1654011947760896e-05,
+      "loss": 2.0178,
+      "step": 7662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1651911635906992e-05,
+      "loss": 2.0059,
+      "step": 7663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.16498111692823e-05,
+      "loss": 1.9951,
+      "step": 7664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.164771054794191e-05,
+      "loss": 2.0964,
+      "step": 7665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1645609771940898e-05,
+      "loss": 2.0798,
+      "step": 7666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.164350884133436e-05,
+      "loss": 2.1216,
+      "step": 7667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.164140775617739e-05,
+      "loss": 2.1096,
+      "step": 7668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1639306516525083e-05,
+      "loss": 2.0217,
+      "step": 7669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1637205122432555e-05,
+      "loss": 2.0708,
+      "step": 7670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.163510357395489e-05,
+      "loss": 2.02,
+      "step": 7671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1633001871147217e-05,
+      "loss": 2.0144,
+      "step": 7672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1630900014064647e-05,
+      "loss": 1.9944,
+      "step": 7673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1628798002762293e-05,
+      "loss": 2.0149,
+      "step": 7674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1626695837295284e-05,
+      "loss": 2.0322,
+      "step": 7675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1624593517718742e-05,
+      "loss": 2.0476,
+      "step": 7676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.16224910440878e-05,
+      "loss": 2.075,
+      "step": 7677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1620388416457593e-05,
+      "loss": 2.0869,
+      "step": 7678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1618285634883264e-05,
+      "loss": 2.0457,
+      "step": 7679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1616182699419955e-05,
+      "loss": 2.0039,
+      "step": 7680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1614079610122802e-05,
+      "loss": 2.0938,
+      "step": 7681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1611976367046966e-05,
+      "loss": 2.1106,
+      "step": 7682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1609872970247603e-05,
+      "loss": 2.0464,
+      "step": 7683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1607769419779868e-05,
+      "loss": 2.0659,
+      "step": 7684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1605665715698927e-05,
+      "loss": 2.0334,
+      "step": 7685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1603561858059945e-05,
+      "loss": 2.0525,
+      "step": 7686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1601457846918092e-05,
+      "loss": 2.0088,
+      "step": 7687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1599353682328543e-05,
+      "loss": 2.0203,
+      "step": 7688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1597249364346487e-05,
+      "loss": 2.0833,
+      "step": 7689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1595144893027097e-05,
+      "loss": 2.0813,
+      "step": 7690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1593040268425562e-05,
+      "loss": 2.0789,
+      "step": 7691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1590935490597074e-05,
+      "loss": 2.0413,
+      "step": 7692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.158883055959683e-05,
+      "loss": 2.125,
+      "step": 7693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1586725475480023e-05,
+      "loss": 2.1504,
+      "step": 7694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.1584620238301872e-05,
+      "loss": 2.0442,
+      "step": 7695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.158251484811757e-05,
+      "loss": 2.0164,
+      "step": 7696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.158040930498233e-05,
+      "loss": 2.0969,
+      "step": 7697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.157830360895137e-05,
+      "loss": 2.0815,
+      "step": 7698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.157619776007991e-05,
+      "loss": 2.061,
+      "step": 7699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.157409175842318e-05,
+      "loss": 2.0518,
+      "step": 7700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.157198560403639e-05,
+      "loss": 2.0793,
+      "step": 7701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1569879296974787e-05,
+      "loss": 2.0776,
+      "step": 7702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1567772837293597e-05,
+      "loss": 2.0308,
+      "step": 7703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1565666225048068e-05,
+      "loss": 2.042,
+      "step": 7704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.156355946029344e-05,
+      "loss": 2.0767,
+      "step": 7705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1561452543084954e-05,
+      "loss": 2.0403,
+      "step": 7706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.155934547347787e-05,
+      "loss": 2.0874,
+      "step": 7707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1557238251527438e-05,
+      "loss": 2.0671,
+      "step": 7708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1555130877288925e-05,
+      "loss": 2.0449,
+      "step": 7709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.155302335081759e-05,
+      "loss": 2.0552,
+      "step": 7710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1550915672168692e-05,
+      "loss": 2.0703,
+      "step": 7711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1548807841397517e-05,
+      "loss": 2.0952,
+      "step": 7712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.154669985855933e-05,
+      "loss": 2.1206,
+      "step": 7713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1544591723709415e-05,
+      "loss": 2.0823,
+      "step": 7714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1542483436903056e-05,
+      "loss": 2.0327,
+      "step": 7715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1540374998195538e-05,
+      "loss": 2.061,
+      "step": 7716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.153826640764215e-05,
+      "loss": 2.0757,
+      "step": 7717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1536157665298198e-05,
+      "loss": 2.0642,
+      "step": 7718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1534048771218967e-05,
+      "loss": 2.0662,
+      "step": 7719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.153193972545977e-05,
+      "loss": 2.0876,
+      "step": 7720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1529830528075912e-05,
+      "loss": 2.0203,
+      "step": 7721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1527721179122702e-05,
+      "loss": 2.0488,
+      "step": 7722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1525611678655455e-05,
+      "loss": 2.0354,
+      "step": 7723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1523502026729497e-05,
+      "loss": 2.0374,
+      "step": 7724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.152139222340014e-05,
+      "loss": 2.083,
+      "step": 7725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.151928226872272e-05,
+      "loss": 2.0613,
+      "step": 7726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.151717216275257e-05,
+      "loss": 2.0552,
+      "step": 7727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1515061905545014e-05,
+      "loss": 2.0508,
+      "step": 7728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1512951497155402e-05,
+      "loss": 2.0344,
+      "step": 7729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.151084093763907e-05,
+      "loss": 2.011,
+      "step": 7730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.150873022705137e-05,
+      "loss": 2.0879,
+      "step": 7731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1506619365447653e-05,
+      "loss": 2.106,
+      "step": 7732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1504508352883266e-05,
+      "loss": 2.1079,
+      "step": 7733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1502397189413577e-05,
+      "loss": 2.0776,
+      "step": 7734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.150028587509394e-05,
+      "loss": 2.1023,
+      "step": 7735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1498174409979738e-05,
+      "loss": 2.1025,
+      "step": 7736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1496062794126316e-05,
+      "loss": 2.0652,
+      "step": 7737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1493951027589073e-05,
+      "loss": 2.0725,
+      "step": 7738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1491839110423375e-05,
+      "loss": 2.0569,
+      "step": 7739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.148972704268461e-05,
+      "loss": 2.0518,
+      "step": 7740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1487614824428158e-05,
+      "loss": 2.0906,
+      "step": 7741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1485502455709416e-05,
+      "loss": 2.0569,
+      "step": 7742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1483389936583776e-05,
+      "loss": 2.0928,
+      "step": 7743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1481277267106633e-05,
+      "loss": 2.1008,
+      "step": 7744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1479164447333394e-05,
+      "loss": 2.0635,
+      "step": 7745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.147705147731947e-05,
+      "loss": 2.0437,
+      "step": 7746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.147493835712026e-05,
+      "loss": 2.0562,
+      "step": 7747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.147282508679118e-05,
+      "loss": 2.0808,
+      "step": 7748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1470711666387655e-05,
+      "loss": 2.0432,
+      "step": 7749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.14685980959651e-05,
+      "loss": 2.0459,
+      "step": 7750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.146648437557895e-05,
+      "loss": 2.0735,
+      "step": 7751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.146437050528463e-05,
+      "loss": 2.1108,
+      "step": 7752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1462256485137567e-05,
+      "loss": 2.0686,
+      "step": 7753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1460142315193204e-05,
+      "loss": 2.0701,
+      "step": 7754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.145802799550699e-05,
+      "loss": 2.0845,
+      "step": 7755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1455913526134357e-05,
+      "loss": 2.051,
+      "step": 7756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1453798907130767e-05,
+      "loss": 2.0728,
+      "step": 7757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1451684138551662e-05,
+      "loss": 2.0725,
+      "step": 7758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1449569220452505e-05,
+      "loss": 2.0461,
+      "step": 7759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1447454152888758e-05,
+      "loss": 2.0708,
+      "step": 7760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.144533893591589e-05,
+      "loss": 2.1221,
+      "step": 7761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1443223569589363e-05,
+      "loss": 2.1221,
+      "step": 7762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1441108053964654e-05,
+      "loss": 2.0552,
+      "step": 7763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1438992389097236e-05,
+      "loss": 2.0583,
+      "step": 7764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.143687657504259e-05,
+      "loss": 2.0769,
+      "step": 7765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.143476061185621e-05,
+      "loss": 2.0674,
+      "step": 7766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1432644499593577e-05,
+      "loss": 2.0869,
+      "step": 7767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1430528238310183e-05,
+      "loss": 2.0591,
+      "step": 7768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1428411828061523e-05,
+      "loss": 2.0295,
+      "step": 7769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1426295268903103e-05,
+      "loss": 2.0503,
+      "step": 7770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1424178560890426e-05,
+      "loss": 2.1006,
+      "step": 7771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1422061704079e-05,
+      "loss": 2.103,
+      "step": 7772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.141994469852433e-05,
+      "loss": 2.0532,
+      "step": 7773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.141782754428194e-05,
+      "loss": 2.0598,
+      "step": 7774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.141571024140735e-05,
+      "loss": 2.0764,
+      "step": 7775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1413592789956082e-05,
+      "loss": 2.0559,
+      "step": 7776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1411475189983664e-05,
+      "loss": 2.0422,
+      "step": 7777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1409357441545622e-05,
+      "loss": 2.0513,
+      "step": 7778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1407239544697504e-05,
+      "loss": 2.072,
+      "step": 7779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1405121499494838e-05,
+      "loss": 2.0549,
+      "step": 7780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1403003305993173e-05,
+      "loss": 2.0918,
+      "step": 7781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.140088496424805e-05,
+      "loss": 2.0898,
+      "step": 7782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1398766474315028e-05,
+      "loss": 2.1135,
+      "step": 7783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1396647836249662e-05,
+      "loss": 2.1099,
+      "step": 7784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.13945290501075e-05,
+      "loss": 2.0801,
+      "step": 7785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1392410115944116e-05,
+      "loss": 2.0913,
+      "step": 7786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1390291033815073e-05,
+      "loss": 2.043,
+      "step": 7787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.138817180377594e-05,
+      "loss": 2.0381,
+      "step": 7788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.138605242588229e-05,
+      "loss": 2.0671,
+      "step": 7789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1383932900189707e-05,
+      "loss": 2.1094,
+      "step": 7790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1381813226753767e-05,
+      "loss": 2.1182,
+      "step": 7791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1379693405630064e-05,
+      "loss": 2.0662,
+      "step": 7792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.137757343687418e-05,
+      "loss": 2.0562,
+      "step": 7793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.137545332054171e-05,
+      "loss": 2.0537,
+      "step": 7794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1373333056688254e-05,
+      "loss": 2.0969,
+      "step": 7795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1371212645369408e-05,
+      "loss": 2.1006,
+      "step": 7796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1369092086640784e-05,
+      "loss": 2.0886,
+      "step": 7797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.136697138055799e-05,
+      "loss": 2.0977,
+      "step": 7798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1364850527176636e-05,
+      "loss": 1.9797,
+      "step": 7799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.136272952655234e-05,
+      "loss": 1.9871,
+      "step": 7800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.136060837874073e-05,
+      "loss": 2.094,
+      "step": 7801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1358487083797423e-05,
+      "loss": 2.0781,
+      "step": 7802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.135636564177804e-05,
+      "loss": 2.0925,
+      "step": 7803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1354244052738233e-05,
+      "loss": 2.0886,
+      "step": 7804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.135212231673362e-05,
+      "loss": 2.0918,
+      "step": 7805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1350000433819847e-05,
+      "loss": 2.0632,
+      "step": 7806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1347878404052567e-05,
+      "loss": 2.0317,
+      "step": 7807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1345756227487416e-05,
+      "loss": 2.062,
+      "step": 7808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1343633904180048e-05,
+      "loss": 2.1118,
+      "step": 7809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1341511434186123e-05,
+      "loss": 2.1113,
+      "step": 7810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1339388817561295e-05,
+      "loss": 2.0425,
+      "step": 7811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.133726605436123e-05,
+      "loss": 2.063,
+      "step": 7812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1335143144641595e-05,
+      "loss": 2.0715,
+      "step": 7813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1333020088458058e-05,
+      "loss": 2.0925,
+      "step": 7814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.13308968858663e-05,
+      "loss": 2.0796,
+      "step": 7815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1328773536921992e-05,
+      "loss": 2.0923,
+      "step": 7816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1326650041680828e-05,
+      "loss": 2.0315,
+      "step": 7817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1324526400198477e-05,
+      "loss": 2.03,
+      "step": 7818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1322402612530642e-05,
+      "loss": 2.1091,
+      "step": 7819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1320278678733013e-05,
+      "loss": 2.1091,
+      "step": 7820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1318154598861285e-05,
+      "loss": 2.0334,
+      "step": 7821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1316030372971168e-05,
+      "loss": 2.0291,
+      "step": 7822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.131390600111836e-05,
+      "loss": 2.0872,
+      "step": 7823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.131178148335857e-05,
+      "loss": 2.0872,
+      "step": 7824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1309656819747512e-05,
+      "loss": 2.0596,
+      "step": 7825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1307532010340907e-05,
+      "loss": 2.0332,
+      "step": 7826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1305407055194473e-05,
+      "loss": 2.0461,
+      "step": 7827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.130328195436393e-05,
+      "loss": 2.0652,
+      "step": 7828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1301156707905012e-05,
+      "loss": 2.0898,
+      "step": 7829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.129903131587345e-05,
+      "loss": 2.0559,
+      "step": 7830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.129690577832498e-05,
+      "loss": 2.0967,
+      "step": 7831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1294780095315334e-05,
+      "loss": 2.0894,
+      "step": 7832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.129265426690027e-05,
+      "loss": 2.0735,
+      "step": 7833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1290528293135523e-05,
+      "loss": 2.0593,
+      "step": 7834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.128840217407685e-05,
+      "loss": 2.0488,
+      "step": 7835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.128627590978001e-05,
+      "loss": 2.0696,
+      "step": 7836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1284149500300758e-05,
+      "loss": 2.0725,
+      "step": 7837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1282022945694848e-05,
+      "loss": 2.0454,
+      "step": 7838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1279896246018058e-05,
+      "loss": 2.0552,
+      "step": 7839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1277769401326152e-05,
+      "loss": 2.0657,
+      "step": 7840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.127564241167491e-05,
+      "loss": 2.0234,
+      "step": 7841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1273515277120107e-05,
+      "loss": 2.0039,
+      "step": 7842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1271387997717522e-05,
+      "loss": 2.0903,
+      "step": 7843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.126926057352294e-05,
+      "loss": 2.0319,
+      "step": 7844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1267133004592154e-05,
+      "loss": 2.1045,
+      "step": 7845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1265005290980955e-05,
+      "loss": 2.0525,
+      "step": 7846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1262877432745138e-05,
+      "loss": 2.0854,
+      "step": 7847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.126074942994051e-05,
+      "loss": 2.0918,
+      "step": 7848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1258621282622867e-05,
+      "loss": 2.0393,
+      "step": 7849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1256492990848026e-05,
+      "loss": 2.0376,
+      "step": 7850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1254364554671793e-05,
+      "loss": 2.1218,
+      "step": 7851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1252235974149986e-05,
+      "loss": 2.1418,
+      "step": 7852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1250107249338423e-05,
+      "loss": 2.0818,
+      "step": 7853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1247978380292928e-05,
+      "loss": 2.082,
+      "step": 7854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1245849367069327e-05,
+      "loss": 2.0986,
+      "step": 7855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1243720209723453e-05,
+      "loss": 2.0994,
+      "step": 7856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.124159090831114e-05,
+      "loss": 2.0886,
+      "step": 7857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1239461462888227e-05,
+      "loss": 2.0986,
+      "step": 7858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1237331873510556e-05,
+      "loss": 2.0356,
+      "step": 7859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.123520214023397e-05,
+      "loss": 2.0273,
+      "step": 7860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1233072263114324e-05,
+      "loss": 2.0725,
+      "step": 7861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1230942242207464e-05,
+      "loss": 2.071,
+      "step": 7862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.122881207756926e-05,
+      "loss": 2.0481,
+      "step": 7863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1226681769255562e-05,
+      "loss": 2.0708,
+      "step": 7864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1224551317322235e-05,
+      "loss": 2.1035,
+      "step": 7865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1222420721825154e-05,
+      "loss": 2.0928,
+      "step": 7866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1220289982820186e-05,
+      "loss": 2.0557,
+      "step": 7867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1218159100363214e-05,
+      "loss": 2.0742,
+      "step": 7868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.121602807451011e-05,
+      "loss": 2.0784,
+      "step": 7869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.121389690531676e-05,
+      "loss": 2.1069,
+      "step": 7870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.121176559283905e-05,
+      "loss": 2.0884,
+      "step": 7871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1209634137132876e-05,
+      "loss": 2.0417,
+      "step": 7872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1207502538254136e-05,
+      "loss": 2.0815,
+      "step": 7873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.120537079625872e-05,
+      "loss": 2.0449,
+      "step": 7874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.120323891120253e-05,
+      "loss": 2.0474,
+      "step": 7875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1201106883141476e-05,
+      "loss": 1.9907,
+      "step": 7876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1198974712131472e-05,
+      "loss": 2.0454,
+      "step": 7877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1196842398228432e-05,
+      "loss": 2.0427,
+      "step": 7878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.119470994148826e-05,
+      "loss": 2.0925,
+      "step": 7879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1192577341966887e-05,
+      "loss": 2.058,
+      "step": 7880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.119044459972024e-05,
+      "loss": 2.0598,
+      "step": 7881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.118831171480425e-05,
+      "loss": 2.0752,
+      "step": 7882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1186178687274836e-05,
+      "loss": 2.1196,
+      "step": 7883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1184045517187942e-05,
+      "loss": 2.0967,
+      "step": 7884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.118191220459951e-05,
+      "loss": 2.0703,
+      "step": 7885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1179778749565486e-05,
+      "loss": 2.1025,
+      "step": 7886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.117764515214181e-05,
+      "loss": 2.0337,
+      "step": 7887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1175511412384442e-05,
+      "loss": 2.04,
+      "step": 7888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1173377530349327e-05,
+      "loss": 2.0757,
+      "step": 7889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1171243506092427e-05,
+      "loss": 2.0901,
+      "step": 7890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1169109339669707e-05,
+      "loss": 2.0112,
+      "step": 7891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1166975031137132e-05,
+      "loss": 2.0042,
+      "step": 7892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.116484058055067e-05,
+      "loss": 2.1174,
+      "step": 7893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1162705987966294e-05,
+      "loss": 2.0977,
+      "step": 7894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1160571253439985e-05,
+      "loss": 2.011,
+      "step": 7895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.115843637702772e-05,
+      "loss": 2.0251,
+      "step": 7896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.115630135878549e-05,
+      "loss": 2.0767,
+      "step": 7897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1154166198769276e-05,
+      "loss": 2.0405,
+      "step": 7898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.115203089703507e-05,
+      "loss": 2.1021,
+      "step": 7899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1149895453638875e-05,
+      "loss": 2.1028,
+      "step": 7900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1147759868636684e-05,
+      "loss": 2.0193,
+      "step": 7901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.11456241420845e-05,
+      "loss": 2.0039,
+      "step": 7902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1143488274038336e-05,
+      "loss": 2.0725,
+      "step": 7903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1141352264554196e-05,
+      "loss": 2.0869,
+      "step": 7904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1139216113688098e-05,
+      "loss": 2.0752,
+      "step": 7905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1137079821496063e-05,
+      "loss": 2.0649,
+      "step": 7906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1134943388034105e-05,
+      "loss": 2.082,
+      "step": 7907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1132806813358256e-05,
+      "loss": 2.0659,
+      "step": 7908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1130670097524547e-05,
+      "loss": 2.0696,
+      "step": 7909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1128533240589002e-05,
+      "loss": 2.062,
+      "step": 7910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1126396242607665e-05,
+      "loss": 1.9951,
+      "step": 7911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1124259103636573e-05,
+      "loss": 2.0066,
+      "step": 7912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.112212182373177e-05,
+      "loss": 2.031,
+      "step": 7913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1119984402949305e-05,
+      "loss": 2.0571,
+      "step": 7914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1117846841345228e-05,
+      "loss": 2.054,
+      "step": 7915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1115709138975596e-05,
+      "loss": 2.0789,
+      "step": 7916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1113571295896464e-05,
+      "loss": 2.0627,
+      "step": 7917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.11114333121639e-05,
+      "loss": 2.0493,
+      "step": 7918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.110929518783397e-05,
+      "loss": 2.0981,
+      "step": 7919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1107156922962732e-05,
+      "loss": 2.1108,
+      "step": 7920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1105018517606277e-05,
+      "loss": 2.1099,
+      "step": 7921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.110287997182067e-05,
+      "loss": 2.0974,
+      "step": 7922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1100741285661995e-05,
+      "loss": 2.0752,
+      "step": 7923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.109860245918634e-05,
+      "loss": 2.0916,
+      "step": 7924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.109646349244979e-05,
+      "loss": 2.0769,
+      "step": 7925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1094324385508436e-05,
+      "loss": 2.0837,
+      "step": 7926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1092185138418374e-05,
+      "loss": 2.0845,
+      "step": 7927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1090045751235706e-05,
+      "loss": 2.092,
+      "step": 7928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1087906224016536e-05,
+      "loss": 2.0549,
+      "step": 7929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1085766556816967e-05,
+      "loss": 2.0918,
+      "step": 7930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1083626749693104e-05,
+      "loss": 2.0754,
+      "step": 7931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1081486802701072e-05,
+      "loss": 2.0398,
+      "step": 7932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1079346715896983e-05,
+      "loss": 2.0288,
+      "step": 7933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1077206489336963e-05,
+      "loss": 2.0615,
+      "step": 7934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1075066123077128e-05,
+      "loss": 2.0754,
+      "step": 7935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1072925617173614e-05,
+      "loss": 2.0532,
+      "step": 7936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1070784971682547e-05,
+      "loss": 2.0171,
+      "step": 7937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.106864418666007e-05,
+      "loss": 2.0181,
+      "step": 7938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.106650326216232e-05,
+      "loss": 2.033,
+      "step": 7939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1064362198245436e-05,
+      "loss": 2.0581,
+      "step": 7940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.106222099496557e-05,
+      "loss": 2.0376,
+      "step": 7941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.106007965237887e-05,
+      "loss": 2.0273,
+      "step": 7942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1057938170541496e-05,
+      "loss": 2.1067,
+      "step": 7943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.10557965495096e-05,
+      "loss": 2.1084,
+      "step": 7944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1053654789339347e-05,
+      "loss": 2.0774,
+      "step": 7945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1051512890086896e-05,
+      "loss": 2.0718,
+      "step": 7946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.104937085180842e-05,
+      "loss": 2.0828,
+      "step": 7947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1047228674560092e-05,
+      "loss": 2.0576,
+      "step": 7948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1045086358398086e-05,
+      "loss": 2.094,
+      "step": 7949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1042943903378588e-05,
+      "loss": 2.0791,
+      "step": 7950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1040801309557772e-05,
+      "loss": 2.0251,
+      "step": 7951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.103865857699183e-05,
+      "loss": 2.0459,
+      "step": 7952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1036515705736954e-05,
+      "loss": 2.0452,
+      "step": 7953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.103437269584933e-05,
+      "loss": 2.0217,
+      "step": 7954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.103222954738517e-05,
+      "loss": 2.0442,
+      "step": 7955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1030086260400665e-05,
+      "loss": 2.0708,
+      "step": 7956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.102794283495203e-05,
+      "loss": 2.0791,
+      "step": 7957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1025799271095458e-05,
+      "loss": 2.093,
+      "step": 7958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1023655568887175e-05,
+      "loss": 2.0767,
+      "step": 7959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1021511728383393e-05,
+      "loss": 2.0879,
+      "step": 7960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.101936774964033e-05,
+      "loss": 2.0791,
+      "step": 7961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.101722363271421e-05,
+      "loss": 2.0747,
+      "step": 7962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.101507937766126e-05,
+      "loss": 2.0549,
+      "step": 7963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1012934984537718e-05,
+      "loss": 2.0588,
+      "step": 7964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1010790453399808e-05,
+      "loss": 2.0696,
+      "step": 7965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.100864578430377e-05,
+      "loss": 2.0603,
+      "step": 7966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1006500977305845e-05,
+      "loss": 2.0715,
+      "step": 7967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.1004356032462285e-05,
+      "loss": 2.0879,
+      "step": 7968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.100221094982933e-05,
+      "loss": 2.0737,
+      "step": 7969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.100006572946324e-05,
+      "loss": 2.0806,
+      "step": 7970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0997920371420263e-05,
+      "loss": 2.0916,
+      "step": 7971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0995774875756663e-05,
+      "loss": 2.1238,
+      "step": 7972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0993629242528702e-05,
+      "loss": 2.0215,
+      "step": 7973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.099148347179265e-05,
+      "loss": 2.0503,
+      "step": 7974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0989337563604773e-05,
+      "loss": 2.0137,
+      "step": 7975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0987191518021346e-05,
+      "loss": 2.0298,
+      "step": 7976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0985045335098648e-05,
+      "loss": 2.084,
+      "step": 7977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.098289901489296e-05,
+      "loss": 2.0869,
+      "step": 7978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0980752557460565e-05,
+      "loss": 2.0857,
+      "step": 7979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0978605962857753e-05,
+      "loss": 2.0964,
+      "step": 7980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0976459231140817e-05,
+      "loss": 2.0066,
+      "step": 7981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0974312362366048e-05,
+      "loss": 2.0186,
+      "step": 7982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.097216535658975e-05,
+      "loss": 2.0696,
+      "step": 7983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0970018213868223e-05,
+      "loss": 2.0479,
+      "step": 7984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0967870934257775e-05,
+      "loss": 2.0552,
+      "step": 7985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0965723517814718e-05,
+      "loss": 2.0344,
+      "step": 7986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0963575964595352e-05,
+      "loss": 2.0491,
+      "step": 7987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0961428274656016e-05,
+      "loss": 2.0688,
+      "step": 7988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0959280448053015e-05,
+      "loss": 2.0945,
+      "step": 7989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0957132484842676e-05,
+      "loss": 2.1028,
+      "step": 7990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.095498438508133e-05,
+      "loss": 2.0476,
+      "step": 7991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0952836148825307e-05,
+      "loss": 2.0254,
+      "step": 7992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.095068777613094e-05,
+      "loss": 2.0645,
+      "step": 7993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.094853926705457e-05,
+      "loss": 2.0674,
+      "step": 7994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0946390621652543e-05,
+      "loss": 2.0906,
+      "step": 7995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.09442418399812e-05,
+      "loss": 2.084,
+      "step": 7996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0942092922096887e-05,
+      "loss": 2.0835,
+      "step": 7997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.093994386805596e-05,
+      "loss": 2.1018,
+      "step": 7998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0937794677914774e-05,
+      "loss": 2.0854,
+      "step": 7999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.09356453517297e-05,
+      "loss": 2.0867,
+      "step": 8000
+    },
+    {
+      "Bleu_1": 0.7735826231943491,
+      "Bleu_2": 0.6128220773515669,
+      "Bleu_3": 0.4709373094437509,
+      "Bleu_4": 0.3574472879063426,
+      "CIDEr": 1.203493211894608,
+      "METEOR": 0.278494447154688,
+      "ROUGE_L": 0.5587908730995429,
+      "epoch": 0.4,
+      "step": 8000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0933495889557086e-05,
+      "loss": 2.0874,
+      "step": 8001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0931346291453307e-05,
+      "loss": 2.0845,
+      "step": 8002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.092919655747473e-05,
+      "loss": 2.1211,
+      "step": 8003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0927046687677736e-05,
+      "loss": 2.0796,
+      "step": 8004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0924896682118698e-05,
+      "loss": 2.0757,
+      "step": 8005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0922746540853998e-05,
+      "loss": 2.1155,
+      "step": 8006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0920596263940017e-05,
+      "loss": 2.0457,
+      "step": 8007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.091844585143315e-05,
+      "loss": 2.0654,
+      "step": 8008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0916295303389788e-05,
+      "loss": 2.0762,
+      "step": 8009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.091414461986632e-05,
+      "loss": 2.0979,
+      "step": 8010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0911993800919152e-05,
+      "loss": 2.0718,
+      "step": 8011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.090984284660469e-05,
+      "loss": 2.0732,
+      "step": 8012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.090769175697933e-05,
+      "loss": 2.1272,
+      "step": 8013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.090554053209949e-05,
+      "loss": 2.1311,
+      "step": 8014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0903389172021584e-05,
+      "loss": 2.0757,
+      "step": 8015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0901237676802022e-05,
+      "loss": 2.0569,
+      "step": 8016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0899086046497226e-05,
+      "loss": 2.0442,
+      "step": 8017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0896934281163625e-05,
+      "loss": 2.0493,
+      "step": 8018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.089478238085764e-05,
+      "loss": 2.0576,
+      "step": 8019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.089263034563571e-05,
+      "loss": 2.0881,
+      "step": 8020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0890478175554266e-05,
+      "loss": 2.0811,
+      "step": 8021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.088832587066974e-05,
+      "loss": 2.0894,
+      "step": 8022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.088617343103858e-05,
+      "loss": 2.0574,
+      "step": 8023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.088402085671723e-05,
+      "loss": 2.0583,
+      "step": 8024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.088186814776214e-05,
+      "loss": 2.0518,
+      "step": 8025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0879715304229763e-05,
+      "loss": 2.0479,
+      "step": 8026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0877562326176553e-05,
+      "loss": 2.0942,
+      "step": 8027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0875409213658967e-05,
+      "loss": 2.0796,
+      "step": 8028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.087325596673347e-05,
+      "loss": 2.0549,
+      "step": 8029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.087110258545653e-05,
+      "loss": 2.0742,
+      "step": 8030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0868949069884623e-05,
+      "loss": 2.064,
+      "step": 8031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0866795420074207e-05,
+      "loss": 2.0591,
+      "step": 8032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.086464163608177e-05,
+      "loss": 2.0435,
+      "step": 8033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0862487717963788e-05,
+      "loss": 2.0383,
+      "step": 8034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.086033366577675e-05,
+      "loss": 2.1208,
+      "step": 8035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0858179479577138e-05,
+      "loss": 2.1174,
+      "step": 8036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0856025159421443e-05,
+      "loss": 2.0667,
+      "step": 8037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0853870705366167e-05,
+      "loss": 2.0632,
+      "step": 8038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0851716117467796e-05,
+      "loss": 2.032,
+      "step": 8039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0849561395782844e-05,
+      "loss": 2.0718,
+      "step": 8040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.084740654036781e-05,
+      "loss": 2.0698,
+      "step": 8041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0845251551279204e-05,
+      "loss": 2.1003,
+      "step": 8042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0843096428573534e-05,
+      "loss": 2.0696,
+      "step": 8043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0840941172307316e-05,
+      "loss": 2.0281,
+      "step": 8044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0838785782537074e-05,
+      "loss": 2.0789,
+      "step": 8045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0836630259319332e-05,
+      "loss": 2.1157,
+      "step": 8046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.083447460271061e-05,
+      "loss": 2.0938,
+      "step": 8047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0832318812767435e-05,
+      "loss": 2.1069,
+      "step": 8048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0830162889546353e-05,
+      "loss": 2.054,
+      "step": 8049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.082800683310389e-05,
+      "loss": 2.0603,
+      "step": 8050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0825850643496595e-05,
+      "loss": 2.0613,
+      "step": 8051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0823694320780995e-05,
+      "loss": 2.0728,
+      "step": 8052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0821537865013653e-05,
+      "loss": 2.052,
+      "step": 8053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0819381276251113e-05,
+      "loss": 2.0764,
+      "step": 8054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.081722455454993e-05,
+      "loss": 2.0364,
+      "step": 8055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0815067699966665e-05,
+      "loss": 2.0195,
+      "step": 8056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0812910712557874e-05,
+      "loss": 2.0852,
+      "step": 8057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0810753592380117e-05,
+      "loss": 2.0693,
+      "step": 8058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0808596339489973e-05,
+      "loss": 2.052,
+      "step": 8059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0806438953944008e-05,
+      "loss": 2.0823,
+      "step": 8060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.08042814357988e-05,
+      "loss": 2.0293,
+      "step": 8061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.080212378511092e-05,
+      "loss": 2.011,
+      "step": 8062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0799966001936955e-05,
+      "loss": 2.0295,
+      "step": 8063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0797808086333496e-05,
+      "loss": 2.0154,
+      "step": 8064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.079565003835712e-05,
+      "loss": 2.0911,
+      "step": 8065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.079349185806443e-05,
+      "loss": 2.073,
+      "step": 8066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0791333545512016e-05,
+      "loss": 2.0645,
+      "step": 8067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0789175100756476e-05,
+      "loss": 2.0747,
+      "step": 8068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0787016523854414e-05,
+      "loss": 2.0911,
+      "step": 8069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.078485781486244e-05,
+      "loss": 2.082,
+      "step": 8070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0782698973837162e-05,
+      "loss": 2.0166,
+      "step": 8071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.078054000083519e-05,
+      "loss": 2.0291,
+      "step": 8072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.077838089591314e-05,
+      "loss": 2.1333,
+      "step": 8073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.077622165912764e-05,
+      "loss": 2.146,
+      "step": 8074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0774062290535304e-05,
+      "loss": 2.0034,
+      "step": 8075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0771902790192766e-05,
+      "loss": 2.0059,
+      "step": 8076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.076974315815665e-05,
+      "loss": 2.064,
+      "step": 8077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0767583394483594e-05,
+      "loss": 2.0557,
+      "step": 8078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0765423499230237e-05,
+      "loss": 2.0984,
+      "step": 8079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.076326347245321e-05,
+      "loss": 2.1062,
+      "step": 8080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.076110331420917e-05,
+      "loss": 2.0637,
+      "step": 8081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.075894302455476e-05,
+      "loss": 2.0259,
+      "step": 8082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0756782603546627e-05,
+      "loss": 2.0227,
+      "step": 8083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.075462205124143e-05,
+      "loss": 2.0161,
+      "step": 8084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0752461367695825e-05,
+      "loss": 2.032,
+      "step": 8085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.075030055296648e-05,
+      "loss": 2.0962,
+      "step": 8086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0748139607110047e-05,
+      "loss": 2.0583,
+      "step": 8087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.07459785301832e-05,
+      "loss": 2.0649,
+      "step": 8088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0743817322242615e-05,
+      "loss": 2.0432,
+      "step": 8089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0741655983344965e-05,
+      "loss": 2.0413,
+      "step": 8090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0739494513546925e-05,
+      "loss": 2.0889,
+      "step": 8091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0737332912905185e-05,
+      "loss": 2.1394,
+      "step": 8092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.073517118147642e-05,
+      "loss": 2.0476,
+      "step": 8093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0733009319317327e-05,
+      "loss": 2.0681,
+      "step": 8094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0730847326484594e-05,
+      "loss": 2.1023,
+      "step": 8095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.072868520303492e-05,
+      "loss": 2.0715,
+      "step": 8096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0726522949025e-05,
+      "loss": 2.0691,
+      "step": 8097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0724360564511537e-05,
+      "loss": 2.0808,
+      "step": 8098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.0722198049551242e-05,
+      "loss": 2.1057,
+      "step": 8099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.072003540420082e-05,
+      "loss": 2.1201,
+      "step": 8100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.071787262851699e-05,
+      "loss": 2.1038,
+      "step": 8101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0715709722556462e-05,
+      "loss": 2.1023,
+      "step": 8102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0713546686375957e-05,
+      "loss": 2.0188,
+      "step": 8103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0711383520032195e-05,
+      "loss": 2.0029,
+      "step": 8104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0709220223581908e-05,
+      "loss": 2.0549,
+      "step": 8105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0707056797081822e-05,
+      "loss": 2.0498,
+      "step": 8106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0704893240588677e-05,
+      "loss": 2.0771,
+      "step": 8107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0702729554159204e-05,
+      "loss": 2.0974,
+      "step": 8108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0700565737850143e-05,
+      "loss": 2.0759,
+      "step": 8109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0698401791718234e-05,
+      "loss": 2.1038,
+      "step": 8110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0696237715820234e-05,
+      "loss": 2.0774,
+      "step": 8111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.069407351021289e-05,
+      "loss": 2.0957,
+      "step": 8112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.069190917495295e-05,
+      "loss": 2.0422,
+      "step": 8113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0689744710097176e-05,
+      "loss": 2.0564,
+      "step": 8114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0687580115702322e-05,
+      "loss": 2.0791,
+      "step": 8115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0685415391825164e-05,
+      "loss": 2.0801,
+      "step": 8116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0683250538522466e-05,
+      "loss": 2.0415,
+      "step": 8117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0681085555850992e-05,
+      "loss": 2.0266,
+      "step": 8118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.067892044386752e-05,
+      "loss": 2.021,
+      "step": 8119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0676755202628826e-05,
+      "loss": 1.9951,
+      "step": 8120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0674589832191693e-05,
+      "loss": 2.0791,
+      "step": 8121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.067242433261291e-05,
+      "loss": 2.1204,
+      "step": 8122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.067025870394925e-05,
+      "loss": 2.043,
+      "step": 8123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.066809294625752e-05,
+      "loss": 2.051,
+      "step": 8124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0665927059594506e-05,
+      "loss": 2.1172,
+      "step": 8125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.066376104401701e-05,
+      "loss": 2.1052,
+      "step": 8126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0661594899581834e-05,
+      "loss": 2.0476,
+      "step": 8127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0659428626345774e-05,
+      "loss": 2.0605,
+      "step": 8128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.065726222436564e-05,
+      "loss": 2.0557,
+      "step": 8129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0655095693698257e-05,
+      "loss": 2.0703,
+      "step": 8130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.065292903440042e-05,
+      "loss": 2.0554,
+      "step": 8131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0650762246528965e-05,
+      "loss": 2.0593,
+      "step": 8132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0648595330140704e-05,
+      "loss": 2.0627,
+      "step": 8133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.064642828529246e-05,
+      "loss": 2.0601,
+      "step": 8134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0644261112041068e-05,
+      "loss": 2.0493,
+      "step": 8135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.064209381044335e-05,
+      "loss": 2.074,
+      "step": 8136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.063992638055616e-05,
+      "loss": 2.0833,
+      "step": 8137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0637758822436313e-05,
+      "loss": 2.0981,
+      "step": 8138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0635591136140663e-05,
+      "loss": 2.0781,
+      "step": 8139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0633423321726052e-05,
+      "loss": 2.0415,
+      "step": 8140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0631255379249333e-05,
+      "loss": 2.0803,
+      "step": 8141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0629087308767355e-05,
+      "loss": 2.0708,
+      "step": 8142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0626919110336974e-05,
+      "loss": 2.0801,
+      "step": 8143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.062475078401504e-05,
+      "loss": 2.0435,
+      "step": 8144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0622582329858427e-05,
+      "loss": 2.0317,
+      "step": 8145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0620413747923996e-05,
+      "loss": 2.0623,
+      "step": 8146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0618245038268615e-05,
+      "loss": 2.0676,
+      "step": 8147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0616076200949153e-05,
+      "loss": 2.0439,
+      "step": 8148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0613907236022483e-05,
+      "loss": 2.0627,
+      "step": 8149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0611738143545496e-05,
+      "loss": 2.0508,
+      "step": 8150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0609568923575062e-05,
+      "loss": 2.0664,
+      "step": 8151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0607399576168073e-05,
+      "loss": 2.0479,
+      "step": 8152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0605230101381417e-05,
+      "loss": 2.0598,
+      "step": 8153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0603060499271983e-05,
+      "loss": 2.074,
+      "step": 8154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0600890769896666e-05,
+      "loss": 2.0471,
+      "step": 8155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0598720913312365e-05,
+      "loss": 2.0659,
+      "step": 8156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0596550929575988e-05,
+      "loss": 2.1018,
+      "step": 8157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.059438081874443e-05,
+      "loss": 2.1086,
+      "step": 8158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0592210580874612e-05,
+      "loss": 2.0859,
+      "step": 8159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.059004021602343e-05,
+      "loss": 2.0857,
+      "step": 8160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0587869724247815e-05,
+      "loss": 2.064,
+      "step": 8161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.058569910560468e-05,
+      "loss": 2.0681,
+      "step": 8162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0583528360150942e-05,
+      "loss": 2.054,
+      "step": 8163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.058135748794353e-05,
+      "loss": 2.0547,
+      "step": 8164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0579186489039374e-05,
+      "loss": 2.0659,
+      "step": 8165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0577015363495405e-05,
+      "loss": 2.0645,
+      "step": 8166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0574844111368555e-05,
+      "loss": 2.0361,
+      "step": 8167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.057267273271577e-05,
+      "loss": 2.0464,
+      "step": 8168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0570501227593984e-05,
+      "loss": 2.0625,
+      "step": 8169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0568329596060148e-05,
+      "loss": 2.0815,
+      "step": 8170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0566157838171206e-05,
+      "loss": 2.0686,
+      "step": 8171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0563985953984115e-05,
+      "loss": 2.0806,
+      "step": 8172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0561813943555827e-05,
+      "loss": 2.0876,
+      "step": 8173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.05596418069433e-05,
+      "loss": 2.0754,
+      "step": 8174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0557469544203493e-05,
+      "loss": 2.0332,
+      "step": 8175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0555297155393376e-05,
+      "loss": 2.021,
+      "step": 8176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0553124640569917e-05,
+      "loss": 2.0466,
+      "step": 8177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0550951999790087e-05,
+      "loss": 2.0122,
+      "step": 8178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0548779233110858e-05,
+      "loss": 2.0576,
+      "step": 8179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0546606340589214e-05,
+      "loss": 2.0664,
+      "step": 8180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.054443332228213e-05,
+      "loss": 2.0928,
+      "step": 8181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0542260178246597e-05,
+      "loss": 2.0964,
+      "step": 8182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0540086908539603e-05,
+      "loss": 1.991,
+      "step": 8183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0537913513218128e-05,
+      "loss": 2.021,
+      "step": 8184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.053573999233918e-05,
+      "loss": 1.9849,
+      "step": 8185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0533566345959755e-05,
+      "loss": 2.0049,
+      "step": 8186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0531392574136848e-05,
+      "loss": 1.9761,
+      "step": 8187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0529218676927473e-05,
+      "loss": 2.0293,
+      "step": 8188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0527044654388628e-05,
+      "loss": 2.0073,
+      "step": 8189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.052487050657733e-05,
+      "loss": 2.0239,
+      "step": 8190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0522696233550594e-05,
+      "loss": 2.0422,
+      "step": 8191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0520521835365433e-05,
+      "loss": 2.0959,
+      "step": 8192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0518347312078876e-05,
+      "loss": 2.0925,
+      "step": 8193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.051617266374794e-05,
+      "loss": 2.0845,
+      "step": 8194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0513997890429656e-05,
+      "loss": 2.0811,
+      "step": 8195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.051182299218105e-05,
+      "loss": 2.0779,
+      "step": 8196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0509647969059166e-05,
+      "loss": 2.0542,
+      "step": 8197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0507472821121037e-05,
+      "loss": 2.031,
+      "step": 8198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0505297548423696e-05,
+      "loss": 2.0615,
+      "step": 8199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0503122151024196e-05,
+      "loss": 2.114,
+      "step": 8200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.050094662897958e-05,
+      "loss": 2.0232,
+      "step": 8201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0498770982346903e-05,
+      "loss": 2.0254,
+      "step": 8202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0496595211183216e-05,
+      "loss": 2.0759,
+      "step": 8203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0494419315545574e-05,
+      "loss": 2.0845,
+      "step": 8204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0492243295491043e-05,
+      "loss": 2.0896,
+      "step": 8205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0490067151076676e-05,
+      "loss": 2.1099,
+      "step": 8206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0487890882359554e-05,
+      "loss": 2.0811,
+      "step": 8207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0485714489396742e-05,
+      "loss": 2.0588,
+      "step": 8208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0483537972245307e-05,
+      "loss": 2.0754,
+      "step": 8209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.048136133096233e-05,
+      "loss": 2.0764,
+      "step": 8210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0479184565604894e-05,
+      "loss": 2.0879,
+      "step": 8211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0477007676230075e-05,
+      "loss": 2.0723,
+      "step": 8212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0474830662894965e-05,
+      "loss": 2.0637,
+      "step": 8213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0472653525656656e-05,
+      "loss": 2.0532,
+      "step": 8214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.047047626457223e-05,
+      "loss": 2.1033,
+      "step": 8215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0468298879698796e-05,
+      "loss": 2.0886,
+      "step": 8216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0466121371093447e-05,
+      "loss": 2.0447,
+      "step": 8217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0463943738813284e-05,
+      "loss": 2.0481,
+      "step": 8218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0461765982915415e-05,
+      "loss": 2.0122,
+      "step": 8219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0459588103456953e-05,
+      "loss": 2.0786,
+      "step": 8220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0457410100495003e-05,
+      "loss": 2.1033,
+      "step": 8221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.045523197408668e-05,
+      "loss": 2.0439,
+      "step": 8222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0453053724289115e-05,
+      "loss": 2.0051,
+      "step": 8223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0450875351159425e-05,
+      "loss": 2.0457,
+      "step": 8224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0448696854754725e-05,
+      "loss": 2.0444,
+      "step": 8225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0446518235132152e-05,
+      "loss": 2.0391,
+      "step": 8226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0444339492348838e-05,
+      "loss": 2.043,
+      "step": 8227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0442160626461915e-05,
+      "loss": 2.0525,
+      "step": 8228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.043998163752853e-05,
+      "loss": 2.0659,
+      "step": 8229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0437802525605812e-05,
+      "loss": 2.0903,
+      "step": 8230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.043562329075091e-05,
+      "loss": 2.0771,
+      "step": 8231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.043344393302097e-05,
+      "loss": 2.0581,
+      "step": 8232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0431264452473156e-05,
+      "loss": 2.0168,
+      "step": 8233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0429084849164607e-05,
+      "loss": 2.0625,
+      "step": 8234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0426905123152485e-05,
+      "loss": 2.0476,
+      "step": 8235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0424725274493952e-05,
+      "loss": 2.0413,
+      "step": 8236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0422545303246173e-05,
+      "loss": 2.093,
+      "step": 8237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0420365209466314e-05,
+      "loss": 2.0642,
+      "step": 8238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0418184993211545e-05,
+      "loss": 2.0461,
+      "step": 8239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.041600465453904e-05,
+      "loss": 2.0593,
+      "step": 8240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0413824193505978e-05,
+      "loss": 2.0789,
+      "step": 8241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0411643610169535e-05,
+      "loss": 2.0996,
+      "step": 8242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0409462904586896e-05,
+      "loss": 2.0391,
+      "step": 8243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0407282076815254e-05,
+      "loss": 2.0486,
+      "step": 8244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0405101126911785e-05,
+      "loss": 2.0144,
+      "step": 8245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.040292005493369e-05,
+      "loss": 2.0422,
+      "step": 8246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0400738860938172e-05,
+      "loss": 2.0586,
+      "step": 8247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0398557544982415e-05,
+      "loss": 2.0911,
+      "step": 8248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0396376107123635e-05,
+      "loss": 2.082,
+      "step": 8249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0394194547419026e-05,
+      "loss": 2.0986,
+      "step": 8250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0392012865925804e-05,
+      "loss": 2.0979,
+      "step": 8251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0389831062701184e-05,
+      "loss": 2.0886,
+      "step": 8252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0387649137802372e-05,
+      "loss": 2.0549,
+      "step": 8253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0385467091286594e-05,
+      "loss": 2.041,
+      "step": 8254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0383284923211067e-05,
+      "loss": 2.0254,
+      "step": 8255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.038110263363302e-05,
+      "loss": 2.0393,
+      "step": 8256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0378920222609678e-05,
+      "loss": 2.0798,
+      "step": 8257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0376737690198274e-05,
+      "loss": 2.0532,
+      "step": 8258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0374555036456044e-05,
+      "loss": 2.1152,
+      "step": 8259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.037237226144022e-05,
+      "loss": 2.0962,
+      "step": 8260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.037018936520805e-05,
+      "loss": 2.0154,
+      "step": 8261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.036800634781677e-05,
+      "loss": 2.0461,
+      "step": 8262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.036582320932363e-05,
+      "loss": 2.0674,
+      "step": 8263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0363639949785887e-05,
+      "loss": 2.064,
+      "step": 8264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0361456569260782e-05,
+      "loss": 1.9907,
+      "step": 8265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.035927306780558e-05,
+      "loss": 2.0229,
+      "step": 8266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0357089445477543e-05,
+      "loss": 2.0969,
+      "step": 8267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0354905702333924e-05,
+      "loss": 2.0867,
+      "step": 8268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0352721838432002e-05,
+      "loss": 2.0549,
+      "step": 8269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0350537853829037e-05,
+      "loss": 2.0479,
+      "step": 8270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.03483537485823e-05,
+      "loss": 2.0889,
+      "step": 8271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0346169522749074e-05,
+      "loss": 2.1162,
+      "step": 8272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.034398517638663e-05,
+      "loss": 2.02,
+      "step": 8273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0341800709552262e-05,
+      "loss": 2.0071,
+      "step": 8274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.033961612230324e-05,
+      "loss": 2.0413,
+      "step": 8275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0337431414696863e-05,
+      "loss": 2.0383,
+      "step": 8276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.033524658679042e-05,
+      "loss": 2.1143,
+      "step": 8277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.03330616386412e-05,
+      "loss": 2.1152,
+      "step": 8278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.033087657030651e-05,
+      "loss": 2.0166,
+      "step": 8279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0328691381843648e-05,
+      "loss": 1.9944,
+      "step": 8280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.032650607330991e-05,
+      "loss": 2.0505,
+      "step": 8281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0324320644762614e-05,
+      "loss": 2.0281,
+      "step": 8282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0322135096259066e-05,
+      "loss": 2.0325,
+      "step": 8283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0319949427856578e-05,
+      "loss": 2.0186,
+      "step": 8284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0317763639612474e-05,
+      "loss": 2.0327,
+      "step": 8285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0315577731584057e-05,
+      "loss": 2.0654,
+      "step": 8286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0313391703828668e-05,
+      "loss": 2.0901,
+      "step": 8287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.031120555640362e-05,
+      "loss": 2.085,
+      "step": 8288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0309019289366252e-05,
+      "loss": 2.0513,
+      "step": 8289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.030683290277389e-05,
+      "loss": 2.0557,
+      "step": 8290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.030464639668387e-05,
+      "loss": 2.0288,
+      "step": 8291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0302459771153536e-05,
+      "loss": 2.0344,
+      "step": 8292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0300273026240226e-05,
+      "loss": 2.0283,
+      "step": 8293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0298086162001285e-05,
+      "loss": 2.0518,
+      "step": 8294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.029589917849406e-05,
+      "loss": 2.021,
+      "step": 8295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.02937120757759e-05,
+      "loss": 2.0378,
+      "step": 8296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0291524853904163e-05,
+      "loss": 1.9438,
+      "step": 8297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.0289337512936204e-05,
+      "loss": 1.9456,
+      "step": 8298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.028715005292939e-05,
+      "loss": 2.0625,
+      "step": 8299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.028496247394108e-05,
+      "loss": 2.0833,
+      "step": 8300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0282774776028632e-05,
+      "loss": 2.0259,
+      "step": 8301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.028058695924943e-05,
+      "loss": 2.0139,
+      "step": 8302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0278399023660838e-05,
+      "loss": 2.1033,
+      "step": 8303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.027621096932024e-05,
+      "loss": 2.074,
+      "step": 8304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0274022796285007e-05,
+      "loss": 2.0493,
+      "step": 8305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0271834504612524e-05,
+      "loss": 2.0771,
+      "step": 8306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0269646094360177e-05,
+      "loss": 2.1316,
+      "step": 8307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0267457565585354e-05,
+      "loss": 2.1226,
+      "step": 8308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.026526891834545e-05,
+      "loss": 2.0337,
+      "step": 8309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0263080152697856e-05,
+      "loss": 2.0635,
+      "step": 8310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0260891268699966e-05,
+      "loss": 2.105,
+      "step": 8311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0258702266409194e-05,
+      "loss": 2.064,
+      "step": 8312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0256513145882926e-05,
+      "loss": 2.0613,
+      "step": 8313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0254323907178588e-05,
+      "loss": 2.0508,
+      "step": 8314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0252134550353578e-05,
+      "loss": 2.0938,
+      "step": 8315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0249945075465307e-05,
+      "loss": 2.1052,
+      "step": 8316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0247755482571204e-05,
+      "loss": 2.0715,
+      "step": 8317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0245565771728675e-05,
+      "loss": 2.0674,
+      "step": 8318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0243375942995155e-05,
+      "loss": 2.0505,
+      "step": 8319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0241185996428062e-05,
+      "loss": 2.0505,
+      "step": 8320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0238995932084823e-05,
+      "loss": 2.0662,
+      "step": 8321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.023680575002288e-05,
+      "loss": 2.0791,
+      "step": 8322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0234615450299658e-05,
+      "loss": 2.0173,
+      "step": 8323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.02324250329726e-05,
+      "loss": 2.0486,
+      "step": 8324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.023023449809914e-05,
+      "loss": 2.0762,
+      "step": 8325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0228043845736734e-05,
+      "loss": 2.0432,
+      "step": 8326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0225853075942816e-05,
+      "loss": 2.0588,
+      "step": 8327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.022366218877485e-05,
+      "loss": 2.0613,
+      "step": 8328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.022147118429028e-05,
+      "loss": 2.0471,
+      "step": 8329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0219280062546566e-05,
+      "loss": 2.0588,
+      "step": 8330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0217088823601174e-05,
+      "loss": 2.093,
+      "step": 8331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.021489746751155e-05,
+      "loss": 2.0967,
+      "step": 8332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0212705994335176e-05,
+      "loss": 2.073,
+      "step": 8333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0210514404129513e-05,
+      "loss": 2.1052,
+      "step": 8334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0208322696952032e-05,
+      "loss": 2.0325,
+      "step": 8335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0206130872860217e-05,
+      "loss": 2.0867,
+      "step": 8336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0203938931911536e-05,
+      "loss": 2.0659,
+      "step": 8337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.020174687416347e-05,
+      "loss": 2.0526,
+      "step": 8338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0199554699673507e-05,
+      "loss": 2.0947,
+      "step": 8339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.019736240849914e-05,
+      "loss": 2.0989,
+      "step": 8340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0195170000697847e-05,
+      "loss": 2.0728,
+      "step": 8341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.019297747632713e-05,
+      "loss": 2.0586,
+      "step": 8342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0190784835444482e-05,
+      "loss": 2.0671,
+      "step": 8343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0188592078107405e-05,
+      "loss": 2.0657,
+      "step": 8344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.01863992043734e-05,
+      "loss": 2.053,
+      "step": 8345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.018420621429997e-05,
+      "loss": 2.0593,
+      "step": 8346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0182013107944632e-05,
+      "loss": 2.072,
+      "step": 8347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0179819885364886e-05,
+      "loss": 2.0923,
+      "step": 8348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0177626546618255e-05,
+      "loss": 2.1008,
+      "step": 8349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0175433091762254e-05,
+      "loss": 2.0874,
+      "step": 8350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.017323952085441e-05,
+      "loss": 2.02,
+      "step": 8351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.017104583395223e-05,
+      "loss": 2.0161,
+      "step": 8352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.016885203111326e-05,
+      "loss": 2.0955,
+      "step": 8353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0166658112395017e-05,
+      "loss": 2.1455,
+      "step": 8354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.016446407785504e-05,
+      "loss": 2.0559,
+      "step": 8355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.016226992755087e-05,
+      "loss": 2.0798,
+      "step": 8356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0160075661540032e-05,
+      "loss": 2.0596,
+      "step": 8357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.015788127988008e-05,
+      "loss": 2.0698,
+      "step": 8358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0155686782628552e-05,
+      "loss": 2.0515,
+      "step": 8359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0153492169843002e-05,
+      "loss": 2.033,
+      "step": 8360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.015129744158098e-05,
+      "loss": 2.0715,
+      "step": 8361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0149102597900034e-05,
+      "loss": 2.0886,
+      "step": 8362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0146907638857725e-05,
+      "loss": 2.0898,
+      "step": 8363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.014471256451162e-05,
+      "loss": 2.0723,
+      "step": 8364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0142517374919272e-05,
+      "loss": 2.0225,
+      "step": 8365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0140322070138256e-05,
+      "loss": 2.0332,
+      "step": 8366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0138126650226133e-05,
+      "loss": 2.0605,
+      "step": 8367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.013593111524048e-05,
+      "loss": 2.0393,
+      "step": 8368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0133735465238873e-05,
+      "loss": 2.0271,
+      "step": 8369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0131539700278887e-05,
+      "loss": 2.0432,
+      "step": 8370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0129343820418106e-05,
+      "loss": 2.0754,
+      "step": 8371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.012714782571412e-05,
+      "loss": 2.1101,
+      "step": 8372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.01249517162245e-05,
+      "loss": 2.0706,
+      "step": 8373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.012275549200685e-05,
+      "loss": 2.0322,
+      "step": 8374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0120559153118762e-05,
+      "loss": 2.1357,
+      "step": 8375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0118362699617833e-05,
+      "loss": 2.1101,
+      "step": 8376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0116166131561658e-05,
+      "loss": 2.1179,
+      "step": 8377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.011396944900784e-05,
+      "loss": 2.0808,
+      "step": 8378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0111772652013985e-05,
+      "loss": 2.0447,
+      "step": 8379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.01095757406377e-05,
+      "loss": 2.0652,
+      "step": 8380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0107378714936604e-05,
+      "loss": 2.0371,
+      "step": 8381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0105181574968306e-05,
+      "loss": 2.0479,
+      "step": 8382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0102984320790422e-05,
+      "loss": 2.0408,
+      "step": 8383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0100786952460574e-05,
+      "loss": 2.0208,
+      "step": 8384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0098589470036388e-05,
+      "loss": 2.0659,
+      "step": 8385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.009639187357549e-05,
+      "loss": 2.0923,
+      "step": 8386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.00941941631355e-05,
+      "loss": 2.0713,
+      "step": 8387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0091996338774062e-05,
+      "loss": 2.0969,
+      "step": 8388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0089798400548806e-05,
+      "loss": 2.0874,
+      "step": 8389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0087600348517372e-05,
+      "loss": 2.0933,
+      "step": 8390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.00854021827374e-05,
+      "loss": 2.0308,
+      "step": 8391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.008320390326654e-05,
+      "loss": 2.051,
+      "step": 8392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.008100551016243e-05,
+      "loss": 2.0803,
+      "step": 8393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0078807003482726e-05,
+      "loss": 2.0691,
+      "step": 8394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.007660838328508e-05,
+      "loss": 2.0564,
+      "step": 8395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.007440964962715e-05,
+      "loss": 2.0405,
+      "step": 8396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0072210802566592e-05,
+      "loss": 2.043,
+      "step": 8397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.007001184216107e-05,
+      "loss": 2.0369,
+      "step": 8398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.006781276846825e-05,
+      "loss": 2.0129,
+      "step": 8399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0065613581545797e-05,
+      "loss": 2.0115,
+      "step": 8400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0063414281451386e-05,
+      "loss": 2.1001,
+      "step": 8401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0061214868242695e-05,
+      "loss": 2.127,
+      "step": 8402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.005901534197739e-05,
+      "loss": 2.0596,
+      "step": 8403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0056815702713155e-05,
+      "loss": 2.0481,
+      "step": 8404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.005461595050768e-05,
+      "loss": 2.0815,
+      "step": 8405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.005241608541864e-05,
+      "loss": 2.093,
+      "step": 8406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0050216107503735e-05,
+      "loss": 2.0625,
+      "step": 8407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0048016016820647e-05,
+      "loss": 2.0679,
+      "step": 8408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0045815813427076e-05,
+      "loss": 2.0457,
+      "step": 8409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0043615497380718e-05,
+      "loss": 2.0449,
+      "step": 8410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0041415068739277e-05,
+      "loss": 2.0591,
+      "step": 8411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0039214527560456e-05,
+      "loss": 2.0527,
+      "step": 8412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0037013873901958e-05,
+      "loss": 2.0767,
+      "step": 8413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0034813107821493e-05,
+      "loss": 2.0771,
+      "step": 8414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0032612229376774e-05,
+      "loss": 2.1318,
+      "step": 8415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.003041123862552e-05,
+      "loss": 2.1243,
+      "step": 8416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0028210135625448e-05,
+      "loss": 2.0759,
+      "step": 8417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0026008920434274e-05,
+      "loss": 2.0881,
+      "step": 8418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.002380759310973e-05,
+      "loss": 2.1072,
+      "step": 8419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.002160615370954e-05,
+      "loss": 2.0969,
+      "step": 8420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.001940460229143e-05,
+      "loss": 2.0813,
+      "step": 8421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0017202938913146e-05,
+      "loss": 2.05,
+      "step": 8422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.001500116363241e-05,
+      "loss": 2.0918,
+      "step": 8423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0012799276506967e-05,
+      "loss": 2.0723,
+      "step": 8424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0010597277594557e-05,
+      "loss": 2.0105,
+      "step": 8425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0008395166952928e-05,
+      "loss": 2.0046,
+      "step": 8426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0006192944639826e-05,
+      "loss": 2.1055,
+      "step": 8427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0003990610713e-05,
+      "loss": 2.1084,
+      "step": 8428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.0001788165230208e-05,
+      "loss": 2.0708,
+      "step": 8429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9999585608249202e-05,
+      "loss": 2.085,
+      "step": 8430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9997382939827745e-05,
+      "loss": 2.0874,
+      "step": 8431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9995180160023592e-05,
+      "loss": 2.0521,
+      "step": 8432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.999297726889452e-05,
+      "loss": 2.0315,
+      "step": 8433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9990774266498294e-05,
+      "loss": 2.0554,
+      "step": 8434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9988571152892674e-05,
+      "loss": 2.0381,
+      "step": 8435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.998636792813545e-05,
+      "loss": 2.0359,
+      "step": 8436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.998416459228439e-05,
+      "loss": 2.0854,
+      "step": 8437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9981961145397277e-05,
+      "loss": 2.0981,
+      "step": 8438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.997975758753189e-05,
+      "loss": 2.0903,
+      "step": 8439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9977553918746017e-05,
+      "loss": 2.0889,
+      "step": 8440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9975350139097445e-05,
+      "loss": 2.084,
+      "step": 8441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9973146248643973e-05,
+      "loss": 2.0737,
+      "step": 8442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.997094224744339e-05,
+      "loss": 2.0984,
+      "step": 8443
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9968738135553486e-05,
+      "loss": 2.1077,
+      "step": 8444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9966533913032074e-05,
+      "loss": 2.0986,
+      "step": 8445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.996432957993695e-05,
+      "loss": 2.1211,
+      "step": 8446
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.996212513632593e-05,
+      "loss": 2.0662,
+      "step": 8447
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9959920582256804e-05,
+      "loss": 1.9917,
+      "step": 8448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9957715917787396e-05,
+      "loss": 2.0588,
+      "step": 8449
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.995551114297552e-05,
+      "loss": 2.0469,
+      "step": 8450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9953306257879e-05,
+      "loss": 2.061,
+      "step": 8451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9951101262555642e-05,
+      "loss": 2.062,
+      "step": 8452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.994889615706328e-05,
+      "loss": 2.0564,
+      "step": 8453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9946690941459736e-05,
+      "loss": 2.0452,
+      "step": 8454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9944485615802843e-05,
+      "loss": 2.0645,
+      "step": 8455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.994228018015043e-05,
+      "loss": 2.061,
+      "step": 8456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.994007463456033e-05,
+      "loss": 2.0522,
+      "step": 8457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.993786897909039e-05,
+      "loss": 2.0525,
+      "step": 8458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.993566321379844e-05,
+      "loss": 2.0154,
+      "step": 8459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.993345733874233e-05,
+      "loss": 2.0378,
+      "step": 8460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9931251353979903e-05,
+      "loss": 2.0278,
+      "step": 8461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.992904525956901e-05,
+      "loss": 2.031,
+      "step": 8462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9926839055567503e-05,
+      "loss": 2.0906,
+      "step": 8463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.992463274203324e-05,
+      "loss": 2.0698,
+      "step": 8464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9922426319024076e-05,
+      "loss": 2.1055,
+      "step": 8465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9920219786597866e-05,
+      "loss": 2.0879,
+      "step": 8466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9918013144812487e-05,
+      "loss": 2.1074,
+      "step": 8467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9915806393725796e-05,
+      "loss": 2.0798,
+      "step": 8468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9913599533395664e-05,
+      "loss": 2.0535,
+      "step": 8469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9911392563879965e-05,
+      "loss": 2.0642,
+      "step": 8470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9909185485236578e-05,
+      "loss": 2.0837,
+      "step": 8471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.990697829752337e-05,
+      "loss": 2.0754,
+      "step": 8472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9904771000798234e-05,
+      "loss": 2.0581,
+      "step": 8473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9902563595119047e-05,
+      "loss": 2.0642,
+      "step": 8474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.99003560805437e-05,
+      "loss": 2.0608,
+      "step": 8475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9898148457130075e-05,
+      "loss": 2.0427,
+      "step": 8476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9895940724936068e-05,
+      "loss": 2.0876,
+      "step": 8477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.989373288401958e-05,
+      "loss": 2.0901,
+      "step": 8478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.98915249344385e-05,
+      "loss": 2.0864,
+      "step": 8479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.988931687625074e-05,
+      "loss": 2.0779,
+      "step": 8480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.988710870951419e-05,
+      "loss": 2.1108,
+      "step": 8481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9884900434286765e-05,
+      "loss": 2.1187,
+      "step": 8482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9882692050626373e-05,
+      "loss": 2.0674,
+      "step": 8483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.988048355859093e-05,
+      "loss": 2.0725,
+      "step": 8484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9878274958238342e-05,
+      "loss": 2.0732,
+      "step": 8485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9876066249626537e-05,
+      "loss": 2.1011,
+      "step": 8486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.987385743281343e-05,
+      "loss": 2.0981,
+      "step": 8487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9871648507856946e-05,
+      "loss": 2.0757,
+      "step": 8488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.986943947481501e-05,
+      "loss": 2.0979,
+      "step": 8489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9867230333745554e-05,
+      "loss": 2.0869,
+      "step": 8490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9865021084706508e-05,
+      "loss": 2.0686,
+      "step": 8491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.986281172775581e-05,
+      "loss": 2.0872,
+      "step": 8492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9860602262951393e-05,
+      "loss": 1.9807,
+      "step": 8493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9858392690351206e-05,
+      "loss": 2.0264,
+      "step": 8494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.985618301001318e-05,
+      "loss": 2.0789,
+      "step": 8495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9853973221995273e-05,
+      "loss": 2.062,
+      "step": 8496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9851763326355422e-05,
+      "loss": 2.0918,
+      "step": 8497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9849553323151596e-05,
+      "loss": 2.0642,
+      "step": 8498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9847343212441738e-05,
+      "loss": 2.02,
+      "step": 8499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9845132994283806e-05,
+      "loss": 2.0486,
+      "step": 8500
+    },
+    {
+      "Bleu_1": 0.7753973075015191,
+      "Bleu_2": 0.615742427089791,
+      "Bleu_3": 0.47454626194640787,
+      "Bleu_4": 0.36179950921019755,
+      "CIDEr": 1.2086605522519491,
+      "METEOR": 0.27913546498075253,
+      "ROUGE_L": 0.5619148757201342,
+      "epoch": 0.42,
+      "step": 8500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.984292266873576e-05,
+      "loss": 2.0991,
+      "step": 8501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9840712235855566e-05,
+      "loss": 2.0835,
+      "step": 8502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.983850169570119e-05,
+      "loss": 2.0808,
+      "step": 8503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9836291048330596e-05,
+      "loss": 2.0911,
+      "step": 8504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9834080293801766e-05,
+      "loss": 2.0388,
+      "step": 8505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9831869432172663e-05,
+      "loss": 2.0688,
+      "step": 8506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9829658463501276e-05,
+      "loss": 2.0659,
+      "step": 8507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.982744738784558e-05,
+      "loss": 2.0859,
+      "step": 8508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.982523620526355e-05,
+      "loss": 2.04,
+      "step": 8509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9823024915813177e-05,
+      "loss": 2.0117,
+      "step": 8510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9820813519552457e-05,
+      "loss": 2.0562,
+      "step": 8511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9818602016539368e-05,
+      "loss": 2.0203,
+      "step": 8512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.981639040683192e-05,
+      "loss": 2.0122,
+      "step": 8513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9814178690488096e-05,
+      "loss": 2.0234,
+      "step": 8514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.98119668675659e-05,
+      "loss": 2.1118,
+      "step": 8515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.980975493812334e-05,
+      "loss": 2.1079,
+      "step": 8516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9807542902218412e-05,
+      "loss": 2.0427,
+      "step": 8517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9805330759909127e-05,
+      "loss": 2.0928,
+      "step": 8518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.98031185112535e-05,
+      "loss": 2.0007,
+      "step": 8519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9800906156309548e-05,
+      "loss": 2.0063,
+      "step": 8520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9798693695135272e-05,
+      "loss": 2.0806,
+      "step": 8521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9796481127788704e-05,
+      "loss": 2.0713,
+      "step": 8522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.979426845432786e-05,
+      "loss": 2.0381,
+      "step": 8523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9792055674810773e-05,
+      "loss": 2.0308,
+      "step": 8524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9789842789295464e-05,
+      "loss": 2.0334,
+      "step": 8525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9787629797839963e-05,
+      "loss": 2.0752,
+      "step": 8526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9785416700502304e-05,
+      "loss": 2.1121,
+      "step": 8527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9783203497340527e-05,
+      "loss": 2.0801,
+      "step": 8528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9780990188412666e-05,
+      "loss": 2.0879,
+      "step": 8529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.977877677377676e-05,
+      "loss": 2.105,
+      "step": 8530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.977656325349086e-05,
+      "loss": 2.1248,
+      "step": 8531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9774349627613008e-05,
+      "loss": 2.0908,
+      "step": 8532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.977213589620126e-05,
+      "loss": 2.0745,
+      "step": 8533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9769922059313664e-05,
+      "loss": 2.0735,
+      "step": 8534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9767708117008273e-05,
+      "loss": 2.0603,
+      "step": 8535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9765494069343148e-05,
+      "loss": 2.092,
+      "step": 8536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9763279916376347e-05,
+      "loss": 2.0583,
+      "step": 8537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.976106565816594e-05,
+      "loss": 2.072,
+      "step": 8538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9758851294769987e-05,
+      "loss": 2.0369,
+      "step": 8539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.975663682624656e-05,
+      "loss": 2.0923,
+      "step": 8540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9754422252653728e-05,
+      "loss": 2.0271,
+      "step": 8541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9752207574049573e-05,
+      "loss": 2.0234,
+      "step": 8542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9749992790492164e-05,
+      "loss": 2.0474,
+      "step": 8543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9747777902039587e-05,
+      "loss": 2.082,
+      "step": 8544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.974556290874992e-05,
+      "loss": 2.0881,
+      "step": 8545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9743347810681254e-05,
+      "loss": 2.0667,
+      "step": 8546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9741132607891667e-05,
+      "loss": 2.04,
+      "step": 8547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9738917300439264e-05,
+      "loss": 2.0759,
+      "step": 8548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9736701888382134e-05,
+      "loss": 2.0796,
+      "step": 8549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.973448637177837e-05,
+      "loss": 2.0601,
+      "step": 8550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.973227075068607e-05,
+      "loss": 2.0991,
+      "step": 8551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9730055025163346e-05,
+      "loss": 2.0745,
+      "step": 8552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.972783919526829e-05,
+      "loss": 2.0562,
+      "step": 8553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9725623261059023e-05,
+      "loss": 2.0354,
+      "step": 8554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.972340722259364e-05,
+      "loss": 2.0176,
+      "step": 8555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9721191079930266e-05,
+      "loss": 2.0405,
+      "step": 8556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9718974833127012e-05,
+      "loss": 2.0298,
+      "step": 8557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9716758482241995e-05,
+      "loss": 2.0369,
+      "step": 8558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9714542027333344e-05,
+      "loss": 2.104,
+      "step": 8559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9712325468459174e-05,
+      "loss": 2.0837,
+      "step": 8560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.971010880567762e-05,
+      "loss": 2.0825,
+      "step": 8561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9707892039046802e-05,
+      "loss": 2.095,
+      "step": 8562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9705675168624858e-05,
+      "loss": 2.0989,
+      "step": 8563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9703458194469925e-05,
+      "loss": 2.0889,
+      "step": 8564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9701241116640132e-05,
+      "loss": 2.0857,
+      "step": 8565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9699023935193626e-05,
+      "loss": 2.0823,
+      "step": 8566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.969680665018855e-05,
+      "loss": 2.1267,
+      "step": 8567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.969458926168305e-05,
+      "loss": 2.1445,
+      "step": 8568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.969237176973527e-05,
+      "loss": 2.1008,
+      "step": 8569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9690154174403363e-05,
+      "loss": 2.134,
+      "step": 8570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9687936475745488e-05,
+      "loss": 2.0659,
+      "step": 8571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.968571867381979e-05,
+      "loss": 2.0605,
+      "step": 8572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9683500768684445e-05,
+      "loss": 2.043,
+      "step": 8573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9681282760397602e-05,
+      "loss": 2.0608,
+      "step": 8574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9679064649017424e-05,
+      "loss": 2.0635,
+      "step": 8575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9676846434602092e-05,
+      "loss": 2.0601,
+      "step": 8576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9674628117209763e-05,
+      "loss": 2.0942,
+      "step": 8577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9672409696898613e-05,
+      "loss": 2.0842,
+      "step": 8578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9670191173726822e-05,
+      "loss": 2.0261,
+      "step": 8579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.966797254775257e-05,
+      "loss": 2.062,
+      "step": 8580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9665753819034028e-05,
+      "loss": 2.0527,
+      "step": 8581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9663534987629383e-05,
+      "loss": 2.0479,
+      "step": 8582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9661316053596824e-05,
+      "loss": 2.0171,
+      "step": 8583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9659097016994542e-05,
+      "loss": 2.0671,
+      "step": 8584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9656877877880727e-05,
+      "loss": 2.0256,
+      "step": 8585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.965465863631357e-05,
+      "loss": 2.0237,
+      "step": 8586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9652439292351265e-05,
+      "loss": 2.0488,
+      "step": 8587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9650219846052024e-05,
+      "loss": 2.0771,
+      "step": 8588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9648000297474036e-05,
+      "loss": 2.0806,
+      "step": 8589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.964578064667552e-05,
+      "loss": 2.0864,
+      "step": 8590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.964356089371467e-05,
+      "loss": 2.0325,
+      "step": 8591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9641341038649707e-05,
+      "loss": 2.0684,
+      "step": 8592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.963912108153883e-05,
+      "loss": 2.1101,
+      "step": 8593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9636901022440276e-05,
+      "loss": 2.1577,
+      "step": 8594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9634680861412245e-05,
+      "loss": 2.0444,
+      "step": 8595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9632460598512972e-05,
+      "loss": 2.0664,
+      "step": 8596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9630240233800667e-05,
+      "loss": 2.1211,
+      "step": 8597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.962801976733357e-05,
+      "loss": 2.1255,
+      "step": 8598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9625799199169892e-05,
+      "loss": 2.0693,
+      "step": 8599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.962357852936789e-05,
+      "loss": 2.0718,
+      "step": 8600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9621357757985777e-05,
+      "loss": 2.1174,
+      "step": 8601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9619136885081798e-05,
+      "loss": 2.1199,
+      "step": 8602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9616915910714195e-05,
+      "loss": 2.0845,
+      "step": 8603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9614694834941207e-05,
+      "loss": 2.0432,
+      "step": 8604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9612473657821083e-05,
+      "loss": 2.0483,
+      "step": 8605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9610252379412067e-05,
+      "loss": 2.0847,
+      "step": 8606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.960803099977241e-05,
+      "loss": 2.0107,
+      "step": 8607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9605809518960366e-05,
+      "loss": 2.011,
+      "step": 8608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9603587937034187e-05,
+      "loss": 2.0889,
+      "step": 8609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.960136625405214e-05,
+      "loss": 2.0862,
+      "step": 8610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9599144470072478e-05,
+      "loss": 2.0557,
+      "step": 8611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.959692258515347e-05,
+      "loss": 2.0442,
+      "step": 8612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.959470059935338e-05,
+      "loss": 2.0535,
+      "step": 8613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9592478512730475e-05,
+      "loss": 2.041,
+      "step": 8614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.959025632534303e-05,
+      "loss": 2.0461,
+      "step": 8615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.958803403724932e-05,
+      "loss": 2.0505,
+      "step": 8616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.958581164850762e-05,
+      "loss": 2.0361,
+      "step": 8617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9583589159176203e-05,
+      "loss": 2.0469,
+      "step": 8618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9581366569313364e-05,
+      "loss": 2.0466,
+      "step": 8619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.957914387897738e-05,
+      "loss": 2.0779,
+      "step": 8620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.957692108822654e-05,
+      "loss": 2.1067,
+      "step": 8621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9574698197119134e-05,
+      "loss": 2.0803,
+      "step": 8622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.957247520571345e-05,
+      "loss": 2.0735,
+      "step": 8623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9570252114067796e-05,
+      "loss": 2.064,
+      "step": 8624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.956802892224046e-05,
+      "loss": 2.0552,
+      "step": 8625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9565805630289744e-05,
+      "loss": 2.0916,
+      "step": 8626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.956358223827395e-05,
+      "loss": 2.0779,
+      "step": 8627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.956135874625139e-05,
+      "loss": 2.1116,
+      "step": 8628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9559135154280366e-05,
+      "loss": 2.0662,
+      "step": 8629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9556911462419192e-05,
+      "loss": 2.1086,
+      "step": 8630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9554687670726182e-05,
+      "loss": 2.0513,
+      "step": 8631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.955246377925965e-05,
+      "loss": 2.063,
+      "step": 8632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.955023978807792e-05,
+      "loss": 2.0208,
+      "step": 8633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.954801569723931e-05,
+      "loss": 2.0125,
+      "step": 8634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.954579150680214e-05,
+      "loss": 2.0691,
+      "step": 8635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9543567216824744e-05,
+      "loss": 2.1072,
+      "step": 8636
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.954134282736545e-05,
+      "loss": 2.0464,
+      "step": 8637
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9539118338482585e-05,
+      "loss": 2.0425,
+      "step": 8638
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9536893750234486e-05,
+      "loss": 2.0637,
+      "step": 8639
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.95346690626795e-05,
+      "loss": 2.0247,
+      "step": 8640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9532444275875954e-05,
+      "loss": 2.0776,
+      "step": 8641
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9530219389882195e-05,
+      "loss": 2.0918,
+      "step": 8642
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.952799440475657e-05,
+      "loss": 2.0334,
+      "step": 8643
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.952576932055742e-05,
+      "loss": 2.0088,
+      "step": 8644
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9523544137343103e-05,
+      "loss": 2.0283,
+      "step": 8645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9521318855171966e-05,
+      "loss": 2.0645,
+      "step": 8646
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.951909347410237e-05,
+      "loss": 2.1121,
+      "step": 8647
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.951686799419267e-05,
+      "loss": 2.0986,
+      "step": 8648
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9514642415501224e-05,
+      "loss": 2.0127,
+      "step": 8649
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9512416738086404e-05,
+      "loss": 2.0029,
+      "step": 8650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9510190962006565e-05,
+      "loss": 2.0471,
+      "step": 8651
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9507965087320082e-05,
+      "loss": 2.0139,
+      "step": 8652
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.950573911408532e-05,
+      "loss": 2.0801,
+      "step": 8653
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9503513042360666e-05,
+      "loss": 2.0918,
+      "step": 8654
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.950128687220448e-05,
+      "loss": 2.0249,
+      "step": 8655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.949906060367515e-05,
+      "loss": 2.0149,
+      "step": 8656
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9496834236831055e-05,
+      "loss": 2.0261,
+      "step": 8657
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9494607771730577e-05,
+      "loss": 2.0137,
+      "step": 8658
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9492381208432106e-05,
+      "loss": 2.0564,
+      "step": 8659
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.949015454699403e-05,
+      "loss": 2.0457,
+      "step": 8660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.948792778747474e-05,
+      "loss": 2.0625,
+      "step": 8661
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9485700929932626e-05,
+      "loss": 2.0288,
+      "step": 8662
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.948347397442609e-05,
+      "loss": 1.9797,
+      "step": 8663
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.948124692101353e-05,
+      "loss": 2.0027,
+      "step": 8664
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9479019769753354e-05,
+      "loss": 2.0596,
+      "step": 8665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.947679252070396e-05,
+      "loss": 2.0991,
+      "step": 8666
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.947456517392375e-05,
+      "loss": 1.9823,
+      "step": 8667
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9472337729471145e-05,
+      "loss": 2.02,
+      "step": 8668
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9470110187404545e-05,
+      "loss": 2.0818,
+      "step": 8669
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9467882547782378e-05,
+      "loss": 2.1045,
+      "step": 8670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.946565481066305e-05,
+      "loss": 2.0479,
+      "step": 8671
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.946342697610499e-05,
+      "loss": 2.0413,
+      "step": 8672
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.946119904416661e-05,
+      "loss": 2.0732,
+      "step": 8673
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.945897101490634e-05,
+      "loss": 2.0708,
+      "step": 8674
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.945674288838261e-05,
+      "loss": 2.0964,
+      "step": 8675
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.945451466465385e-05,
+      "loss": 2.1016,
+      "step": 8676
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9452286343778488e-05,
+      "loss": 2.0498,
+      "step": 8677
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.945005792581496e-05,
+      "loss": 2.0974,
+      "step": 8678
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9447829410821706e-05,
+      "loss": 2.0466,
+      "step": 8679
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9445600798857168e-05,
+      "loss": 2.1162,
+      "step": 8680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9443372089979786e-05,
+      "loss": 2.0061,
+      "step": 8681
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9441143284248004e-05,
+      "loss": 2.0117,
+      "step": 8682
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9438914381720276e-05,
+      "loss": 2.0405,
+      "step": 8683
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9436685382455043e-05,
+      "loss": 2.0698,
+      "step": 8684
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9434456286510763e-05,
+      "loss": 2.0723,
+      "step": 8685
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9432227093945895e-05,
+      "loss": 2.0566,
+      "step": 8686
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9429997804818892e-05,
+      "loss": 2.0474,
+      "step": 8687
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9427768419188216e-05,
+      "loss": 2.0496,
+      "step": 8688
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9425538937112326e-05,
+      "loss": 2.0686,
+      "step": 8689
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9423309358649693e-05,
+      "loss": 2.0823,
+      "step": 8690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9421079683858788e-05,
+      "loss": 2.082,
+      "step": 8691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9418849912798075e-05,
+      "loss": 2.0579,
+      "step": 8692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9416620045526026e-05,
+      "loss": 2.1147,
+      "step": 8693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9414390082101122e-05,
+      "loss": 2.0789,
+      "step": 8694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9412160022581838e-05,
+      "loss": 2.0674,
+      "step": 8695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.940992986702666e-05,
+      "loss": 2.0867,
+      "step": 8696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9407699615494063e-05,
+      "loss": 2.0486,
+      "step": 8697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.940546926804254e-05,
+      "loss": 2.0669,
+      "step": 8698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9403238824730575e-05,
+      "loss": 2.0544,
+      "step": 8699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9401008285616655e-05,
+      "loss": 2.071,
+      "step": 8700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9398777650759284e-05,
+      "loss": 2.0698,
+      "step": 8701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9396546920216955e-05,
+      "loss": 2.0481,
+      "step": 8702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9394316094048157e-05,
+      "loss": 2.0786,
+      "step": 8703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.93920851723114e-05,
+      "loss": 2.0837,
+      "step": 8704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9389854155065183e-05,
+      "loss": 2.105,
+      "step": 8705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9387623042368016e-05,
+      "loss": 2.0942,
+      "step": 8706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9385391834278404e-05,
+      "loss": 2.0244,
+      "step": 8707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.938316053085486e-05,
+      "loss": 2.054,
+      "step": 8708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9380929132155892e-05,
+      "loss": 2.0251,
+      "step": 8709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.937869763824002e-05,
+      "loss": 2.0461,
+      "step": 8710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9376466049165763e-05,
+      "loss": 2.1169,
+      "step": 8711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9374234364991645e-05,
+      "loss": 2.1187,
+      "step": 8712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9372002585776184e-05,
+      "loss": 2.0811,
+      "step": 8713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9369770711577905e-05,
+      "loss": 2.1006,
+      "step": 8714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.936753874245534e-05,
+      "loss": 2.0254,
+      "step": 8715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9365306678467013e-05,
+      "loss": 1.9889,
+      "step": 8716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9363074519671467e-05,
+      "loss": 2.072,
+      "step": 8717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9360842266127233e-05,
+      "loss": 2.0874,
+      "step": 8718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.935860991789285e-05,
+      "loss": 2.0496,
+      "step": 8719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9356377475026855e-05,
+      "loss": 2.0779,
+      "step": 8720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9354144937587798e-05,
+      "loss": 2.0349,
+      "step": 8721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9351912305634224e-05,
+      "loss": 2.0439,
+      "step": 8722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9349679579224675e-05,
+      "loss": 2.0503,
+      "step": 8723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9347446758417705e-05,
+      "loss": 2.0571,
+      "step": 8724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9345213843271865e-05,
+      "loss": 2.0376,
+      "step": 8725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9342980833845714e-05,
+      "loss": 2.053,
+      "step": 8726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9340747730197815e-05,
+      "loss": 2.0415,
+      "step": 8727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9338514532386716e-05,
+      "loss": 2.0405,
+      "step": 8728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9336281240470988e-05,
+      "loss": 2.126,
+      "step": 8729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9334047854509196e-05,
+      "loss": 2.127,
+      "step": 8730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9331814374559906e-05,
+      "loss": 1.9993,
+      "step": 8731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9329580800681686e-05,
+      "loss": 1.9819,
+      "step": 8732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9327347132933116e-05,
+      "loss": 2.0759,
+      "step": 8733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9325113371372764e-05,
+      "loss": 2.0605,
+      "step": 8734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9322879516059218e-05,
+      "loss": 2.1021,
+      "step": 8735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9320645567051044e-05,
+      "loss": 2.1079,
+      "step": 8736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.931841152440684e-05,
+      "loss": 2.1067,
+      "step": 8737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9316177388185178e-05,
+      "loss": 2.1255,
+      "step": 8738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9313943158444654e-05,
+      "loss": 2.0479,
+      "step": 8739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9311708835243853e-05,
+      "loss": 2.0393,
+      "step": 8740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.930947441864137e-05,
+      "loss": 2.0374,
+      "step": 8741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9307239908695804e-05,
+      "loss": 2.03,
+      "step": 8742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9305005305465745e-05,
+      "loss": 2.064,
+      "step": 8743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9302770609009796e-05,
+      "loss": 2.0393,
+      "step": 8744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9300535819386556e-05,
+      "loss": 2.0967,
+      "step": 8745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9298300936654638e-05,
+      "loss": 2.1138,
+      "step": 8746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9296065960872645e-05,
+      "loss": 2.0381,
+      "step": 8747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9293830892099183e-05,
+      "loss": 2.0449,
+      "step": 8748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.929159573039287e-05,
+      "loss": 2.0151,
+      "step": 8749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9289360475812314e-05,
+      "loss": 2.075,
+      "step": 8750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9287125128416136e-05,
+      "loss": 2.0662,
+      "step": 8751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9284889688262956e-05,
+      "loss": 2.0725,
+      "step": 8752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9282654155411394e-05,
+      "loss": 2.0632,
+      "step": 8753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9280418529920077e-05,
+      "loss": 2.0769,
+      "step": 8754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9278182811847627e-05,
+      "loss": 2.0786,
+      "step": 8755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.927594700125268e-05,
+      "loss": 2.0637,
+      "step": 8756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9273711098193858e-05,
+      "loss": 2.1045,
+      "step": 8757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9271475102729808e-05,
+      "loss": 2.1128,
+      "step": 8758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9269239014919152e-05,
+      "loss": 2.1067,
+      "step": 8759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9267002834820537e-05,
+      "loss": 2.0972,
+      "step": 8760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9264766562492607e-05,
+      "loss": 2.0603,
+      "step": 8761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9262530197993995e-05,
+      "loss": 2.0632,
+      "step": 8762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.926029374138336e-05,
+      "loss": 2.0562,
+      "step": 8763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.925805719271934e-05,
+      "loss": 2.075,
+      "step": 8764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.925582055206059e-05,
+      "loss": 2.0502,
+      "step": 8765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9253583819465764e-05,
+      "loss": 2.0991,
+      "step": 8766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9251346994993515e-05,
+      "loss": 1.9651,
+      "step": 8767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9249110078702502e-05,
+      "loss": 2.019,
+      "step": 8768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.924687307065139e-05,
+      "loss": 2.0525,
+      "step": 8769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9244635970898835e-05,
+      "loss": 2.0498,
+      "step": 8770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9242398779503508e-05,
+      "loss": 2.0309,
+      "step": 8771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9240161496524075e-05,
+      "loss": 2.0503,
+      "step": 8772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.923792412201921e-05,
+      "loss": 2.0449,
+      "step": 8773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9235686656047575e-05,
+      "loss": 2.0459,
+      "step": 8774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9233449098667855e-05,
+      "loss": 2.0779,
+      "step": 8775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9231211449938718e-05,
+      "loss": 2.0706,
+      "step": 8776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9228973709918857e-05,
+      "loss": 2.0652,
+      "step": 8777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9226735878666944e-05,
+      "loss": 2.061,
+      "step": 8778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.922449795624166e-05,
+      "loss": 2.0264,
+      "step": 8779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9222259942701705e-05,
+      "loss": 2.0349,
+      "step": 8780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.922002183810576e-05,
+      "loss": 2.0962,
+      "step": 8781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9217783642512513e-05,
+      "loss": 2.1045,
+      "step": 8782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.921554535598067e-05,
+      "loss": 2.0117,
+      "step": 8783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9213306978568918e-05,
+      "loss": 2.0103,
+      "step": 8784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9211068510335954e-05,
+      "loss": 2.0994,
+      "step": 8785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.920882995134049e-05,
+      "loss": 2.1079,
+      "step": 8786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9206591301641218e-05,
+      "loss": 2.0896,
+      "step": 8787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.920435256129685e-05,
+      "loss": 2.0781,
+      "step": 8788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9202113730366094e-05,
+      "loss": 2.075,
+      "step": 8789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.919987480890766e-05,
+      "loss": 2.0876,
+      "step": 8790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9197635796980262e-05,
+      "loss": 1.9438,
+      "step": 8791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9195396694642618e-05,
+      "loss": 1.9817,
+      "step": 8792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9193157501953443e-05,
+      "loss": 2.0569,
+      "step": 8793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.919091821897145e-05,
+      "loss": 2.0696,
+      "step": 8794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9188678845755376e-05,
+      "loss": 2.0007,
+      "step": 8795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9186439382363933e-05,
+      "loss": 2.0637,
+      "step": 8796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9184199828855856e-05,
+      "loss": 2.0393,
+      "step": 8797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9181960185289874e-05,
+      "loss": 2.0618,
+      "step": 8798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9179720451724714e-05,
+      "loss": 2.0613,
+      "step": 8799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.917748062821912e-05,
+      "loss": 2.1006,
+      "step": 8800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9175240714831814e-05,
+      "loss": 2.041,
+      "step": 8801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.917300071162155e-05,
+      "loss": 2.0337,
+      "step": 8802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.917076061864706e-05,
+      "loss": 2.0278,
+      "step": 8803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.916852043596709e-05,
+      "loss": 2.0359,
+      "step": 8804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9166280163640394e-05,
+      "loss": 2.0618,
+      "step": 8805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.916403980172571e-05,
+      "loss": 2.1021,
+      "step": 8806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9161799350281796e-05,
+      "loss": 1.9631,
+      "step": 8807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9159558809367402e-05,
+      "loss": 1.9998,
+      "step": 8808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.915731817904128e-05,
+      "loss": 2.1155,
+      "step": 8809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9155077459362192e-05,
+      "loss": 2.1211,
+      "step": 8810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9152836650388898e-05,
+      "loss": 2.061,
+      "step": 8811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.915059575218016e-05,
+      "loss": 2.0303,
+      "step": 8812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.914835476479475e-05,
+      "loss": 2.0891,
+      "step": 8813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9146113688291426e-05,
+      "loss": 2.0835,
+      "step": 8814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9143872522728953e-05,
+      "loss": 2.1038,
+      "step": 8815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9141631268166116e-05,
+      "loss": 2.1025,
+      "step": 8816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9139389924661682e-05,
+      "loss": 2.0383,
+      "step": 8817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9137148492274436e-05,
+      "loss": 2.0532,
+      "step": 8818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9134906971063143e-05,
+      "loss": 2.0591,
+      "step": 8819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9132665361086596e-05,
+      "loss": 2.0579,
+      "step": 8820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9130423662403572e-05,
+      "loss": 2.063,
+      "step": 8821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9128181875072854e-05,
+      "loss": 2.0266,
+      "step": 8822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.912593999915324e-05,
+      "loss": 2.093,
+      "step": 8823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9123698034703518e-05,
+      "loss": 2.0547,
+      "step": 8824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9121455981782476e-05,
+      "loss": 2.0693,
+      "step": 8825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.911921384044891e-05,
+      "loss": 2.0391,
+      "step": 8826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9116971610761625e-05,
+      "loss": 2.051,
+      "step": 8827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9114729292779413e-05,
+      "loss": 2.0623,
+      "step": 8828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9112486886561084e-05,
+      "loss": 2.061,
+      "step": 8829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.911024439216543e-05,
+      "loss": 2.072,
+      "step": 8830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9108001809651266e-05,
+      "loss": 2.0525,
+      "step": 8831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.91057591390774e-05,
+      "loss": 2.061,
+      "step": 8832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9103516380502648e-05,
+      "loss": 2.0854,
+      "step": 8833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9101273533985822e-05,
+      "loss": 2.0774,
+      "step": 8834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9099030599585726e-05,
+      "loss": 2.0168,
+      "step": 8835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9096787577361192e-05,
+      "loss": 2.0635,
+      "step": 8836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9094544467371037e-05,
+      "loss": 2.053,
+      "step": 8837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9092301269674084e-05,
+      "loss": 2.0398,
+      "step": 8838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9090057984329156e-05,
+      "loss": 2.0759,
+      "step": 8839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9087814611395082e-05,
+      "loss": 2.0708,
+      "step": 8840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.908557115093069e-05,
+      "loss": 2.0642,
+      "step": 8841
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9083327602994818e-05,
+      "loss": 2.0381,
+      "step": 8842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9081083967646295e-05,
+      "loss": 2.0378,
+      "step": 8843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9078840244943965e-05,
+      "loss": 2.0125,
+      "step": 8844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9076596434946653e-05,
+      "loss": 2.0715,
+      "step": 8845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9074352537713212e-05,
+      "loss": 2.031,
+      "step": 8846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9072108553302482e-05,
+      "loss": 2.1079,
+      "step": 8847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9069864481773313e-05,
+      "loss": 2.1206,
+      "step": 8848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9067620323184545e-05,
+      "loss": 2.0247,
+      "step": 8849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9065376077595035e-05,
+      "loss": 2.0574,
+      "step": 8850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.906313174506363e-05,
+      "loss": 2.0427,
+      "step": 8851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9060887325649193e-05,
+      "loss": 2.0541,
+      "step": 8852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9058642819410575e-05,
+      "loss": 2.0459,
+      "step": 8853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.905639822640664e-05,
+      "loss": 2.0781,
+      "step": 8854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9054153546696246e-05,
+      "loss": 2.0713,
+      "step": 8855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9051908780338252e-05,
+      "loss": 2.0808,
+      "step": 8856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.904966392739154e-05,
+      "loss": 2.0662,
+      "step": 8857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9047418987914963e-05,
+      "loss": 2.0369,
+      "step": 8858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9045173961967398e-05,
+      "loss": 2.0796,
+      "step": 8859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9042928849607724e-05,
+      "loss": 2.075,
+      "step": 8860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9040683650894807e-05,
+      "loss": 2.0625,
+      "step": 8861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.903843836588753e-05,
+      "loss": 2.0706,
+      "step": 8862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9036192994644773e-05,
+      "loss": 2.0906,
+      "step": 8863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9033947537225418e-05,
+      "loss": 2.0916,
+      "step": 8864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.903170199368835e-05,
+      "loss": 2.0591,
+      "step": 8865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9029456364092448e-05,
+      "loss": 2.0769,
+      "step": 8866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.902721064849661e-05,
+      "loss": 2.0085,
+      "step": 8867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9024964846959726e-05,
+      "loss": 2.0488,
+      "step": 8868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.902271895954069e-05,
+      "loss": 2.0994,
+      "step": 8869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9020472986298393e-05,
+      "loss": 2.1069,
+      "step": 8870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9018226927291736e-05,
+      "loss": 2.0894,
+      "step": 8871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.901598078257962e-05,
+      "loss": 2.1182,
+      "step": 8872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9013734552220948e-05,
+      "loss": 2.0659,
+      "step": 8873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9011488236274623e-05,
+      "loss": 2.0398,
+      "step": 8874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.900924183479955e-05,
+      "loss": 2.0576,
+      "step": 8875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9006995347854645e-05,
+      "loss": 2.0464,
+      "step": 8876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.9004748775498812e-05,
+      "loss": 2.0842,
+      "step": 8877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.900250211779097e-05,
+      "loss": 2.0781,
+      "step": 8878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.900025537479003e-05,
+      "loss": 2.042,
+      "step": 8879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.899800854655492e-05,
+      "loss": 2.0352,
+      "step": 8880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8995761633144553e-05,
+      "loss": 2.0588,
+      "step": 8881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.899351463461785e-05,
+      "loss": 2.0393,
+      "step": 8882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8991267551033735e-05,
+      "loss": 2.0145,
+      "step": 8883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8989020382451143e-05,
+      "loss": 2.0642,
+      "step": 8884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8986773128929e-05,
+      "loss": 2.1021,
+      "step": 8885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8984525790526237e-05,
+      "loss": 2.095,
+      "step": 8886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8982278367301783e-05,
+      "loss": 2.0408,
+      "step": 8887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8980030859314583e-05,
+      "loss": 2.0005,
+      "step": 8888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897778326662357e-05,
+      "loss": 2.0947,
+      "step": 8889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897553558928769e-05,
+      "loss": 2.1199,
+      "step": 8890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8973287827365873e-05,
+      "loss": 2.061,
+      "step": 8891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8971039980917074e-05,
+      "loss": 2.0632,
+      "step": 8892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896879205000024e-05,
+      "loss": 2.0869,
+      "step": 8893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8966544034674316e-05,
+      "loss": 2.1025,
+      "step": 8894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896429593499826e-05,
+      "loss": 2.0505,
+      "step": 8895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896204775103102e-05,
+      "loss": 2.0491,
+      "step": 8896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8959799482831553e-05,
+      "loss": 2.0725,
+      "step": 8897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895755113045882e-05,
+      "loss": 2.0786,
+      "step": 8898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895530269397178e-05,
+      "loss": 2.0542,
+      "step": 8899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8953054173429397e-05,
+      "loss": 2.0657,
+      "step": 8900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8950805568890633e-05,
+      "loss": 2.0022,
+      "step": 8901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8948556880414455e-05,
+      "loss": 1.9834,
+      "step": 8902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8946308108059832e-05,
+      "loss": 1.9473,
+      "step": 8903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8944059251885737e-05,
+      "loss": 1.9988,
+      "step": 8904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8941810311951148e-05,
+      "loss": 2.0903,
+      "step": 8905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8939561288315034e-05,
+      "loss": 2.075,
+      "step": 8906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8937312181036374e-05,
+      "loss": 2.106,
+      "step": 8907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893506299017415e-05,
+      "loss": 2.0703,
+      "step": 8908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8932813715787345e-05,
+      "loss": 2.0604,
+      "step": 8909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8930564357934944e-05,
+      "loss": 2.0955,
+      "step": 8910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892831491667593e-05,
+      "loss": 2.0044,
+      "step": 8911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8926065392069297e-05,
+      "loss": 2.0371,
+      "step": 8912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892381578417403e-05,
+      "loss": 2.0015,
+      "step": 8913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892156609304913e-05,
+      "loss": 2.0356,
+      "step": 8914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891931631875359e-05,
+      "loss": 2.0483,
+      "step": 8915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.89170664613464e-05,
+      "loss": 2.0264,
+      "step": 8916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8914816520886575e-05,
+      "loss": 2.0964,
+      "step": 8917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8912566497433098e-05,
+      "loss": 2.1008,
+      "step": 8918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891031639104499e-05,
+      "loss": 2.0356,
+      "step": 8919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890806620178125e-05,
+      "loss": 2.054,
+      "step": 8920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890581592970089e-05,
+      "loss": 2.0537,
+      "step": 8921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8903565574862912e-05,
+      "loss": 2.0703,
+      "step": 8922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8901315137326337e-05,
+      "loss": 2.0454,
+      "step": 8923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8899064617150182e-05,
+      "loss": 2.0698,
+      "step": 8924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889681401439346e-05,
+      "loss": 2.0752,
+      "step": 8925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889456332911519e-05,
+      "loss": 2.0713,
+      "step": 8926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8892312561374394e-05,
+      "loss": 2.0439,
+      "step": 8927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8890061711230097e-05,
+      "loss": 2.0579,
+      "step": 8928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8887810778741322e-05,
+      "loss": 2.064,
+      "step": 8929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8885559763967105e-05,
+      "loss": 2.0542,
+      "step": 8930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.888330866696647e-05,
+      "loss": 2.0742,
+      "step": 8931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8881057487798445e-05,
+      "loss": 2.0903,
+      "step": 8932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.887880622652207e-05,
+      "loss": 2.1174,
+      "step": 8933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8876554883196388e-05,
+      "loss": 2.1079,
+      "step": 8934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8874303457880432e-05,
+      "loss": 2.0784,
+      "step": 8935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8872051950633236e-05,
+      "loss": 2.0999,
+      "step": 8936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8869800361513852e-05,
+      "loss": 2.1121,
+      "step": 8937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.886754869058132e-05,
+      "loss": 2.0798,
+      "step": 8938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8865296937894693e-05,
+      "loss": 2.0544,
+      "step": 8939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.886304510351302e-05,
+      "loss": 2.0508,
+      "step": 8940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.886079318749535e-05,
+      "loss": 2.0786,
+      "step": 8941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8858541189900736e-05,
+      "loss": 2.0701,
+      "step": 8942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.885628911078823e-05,
+      "loss": 2.0186,
+      "step": 8943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8854036950216906e-05,
+      "loss": 2.0393,
+      "step": 8944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8851784708245807e-05,
+      "loss": 2.0588,
+      "step": 8945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8849532384934005e-05,
+      "loss": 2.0615,
+      "step": 8946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8847279980340562e-05,
+      "loss": 2.0581,
+      "step": 8947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8845027494524544e-05,
+      "loss": 2.0078,
+      "step": 8948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.884277492754502e-05,
+      "loss": 1.9773,
+      "step": 8949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.884052227946106e-05,
+      "loss": 1.9841,
+      "step": 8950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8838269550331747e-05,
+      "loss": 2.0596,
+      "step": 8951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8836016740216137e-05,
+      "loss": 2.0435,
+      "step": 8952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8833763849173323e-05,
+      "loss": 2.1587,
+      "step": 8953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8831510877262382e-05,
+      "loss": 2.1409,
+      "step": 8954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8829257824542388e-05,
+      "loss": 2.0295,
+      "step": 8955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8827004691072435e-05,
+      "loss": 2.0266,
+      "step": 8956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8824751476911603e-05,
+      "loss": 2.0825,
+      "step": 8957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8822498182118978e-05,
+      "loss": 2.0745,
+      "step": 8958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.882024480675365e-05,
+      "loss": 2.0884,
+      "step": 8959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.881799135087472e-05,
+      "loss": 2.0872,
+      "step": 8960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8815737814541276e-05,
+      "loss": 2.0642,
+      "step": 8961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.881348419781241e-05,
+      "loss": 2.0376,
+      "step": 8962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8811230500747226e-05,
+      "loss": 2.0413,
+      "step": 8963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8808976723404823e-05,
+      "loss": 2.0391,
+      "step": 8964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8806722865844306e-05,
+      "loss": 2.075,
+      "step": 8965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8804468928124776e-05,
+      "loss": 2.0681,
+      "step": 8966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8802214910305342e-05,
+      "loss": 2.0613,
+      "step": 8967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8799960812445113e-05,
+      "loss": 2.1003,
+      "step": 8968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.87977066346032e-05,
+      "loss": 2.0569,
+      "step": 8969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.879545237683872e-05,
+      "loss": 2.0876,
+      "step": 8970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8793198039210784e-05,
+      "loss": 2.0681,
+      "step": 8971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8790943621778505e-05,
+      "loss": 2.0974,
+      "step": 8972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8788689124601008e-05,
+      "loss": 1.9996,
+      "step": 8973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8786434547737416e-05,
+      "loss": 2.0527,
+      "step": 8974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8784179891246852e-05,
+      "loss": 2.064,
+      "step": 8975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8781925155188442e-05,
+      "loss": 2.0903,
+      "step": 8976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.877967033962131e-05,
+      "loss": 2.05,
+      "step": 8977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8777415444604584e-05,
+      "loss": 2.0796,
+      "step": 8978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8775160470197405e-05,
+      "loss": 2.0781,
+      "step": 8979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.87729054164589e-05,
+      "loss": 2.0715,
+      "step": 8980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8770650283448205e-05,
+      "loss": 2.0654,
+      "step": 8981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8768395071224464e-05,
+      "loss": 2.0891,
+      "step": 8982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8766139779846812e-05,
+      "loss": 2.085,
+      "step": 8983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8763884409374396e-05,
+      "loss": 2.0874,
+      "step": 8984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8761628959866355e-05,
+      "loss": 2.1211,
+      "step": 8985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8759373431381843e-05,
+      "loss": 2.1216,
+      "step": 8986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.875711782398e-05,
+      "loss": 2.0544,
+      "step": 8987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8754862137719983e-05,
+      "loss": 2.0374,
+      "step": 8988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8752606372660938e-05,
+      "loss": 2.0498,
+      "step": 8989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8750350528862028e-05,
+      "loss": 2.0669,
+      "step": 8990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8748094606382404e-05,
+      "loss": 2.0459,
+      "step": 8991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8745838605281233e-05,
+      "loss": 2.0637,
+      "step": 8992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.874358252561766e-05,
+      "loss": 2.1025,
+      "step": 8993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8741326367450866e-05,
+      "loss": 2.0652,
+      "step": 8994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8739070130840003e-05,
+      "loss": 2.0259,
+      "step": 8995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8736813815844248e-05,
+      "loss": 2.0496,
+      "step": 8996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8734557422522766e-05,
+      "loss": 2.0825,
+      "step": 8997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8732300950934725e-05,
+      "loss": 2.0989,
+      "step": 8998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8730044401139302e-05,
+      "loss": 2.0544,
+      "step": 8999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8727787773195674e-05,
+      "loss": 2.0854,
+      "step": 9000
+    },
+    {
+      "Bleu_1": 0.7726820910874161,
+      "Bleu_2": 0.613509068663122,
+      "Bleu_3": 0.4722607498900242,
+      "Bleu_4": 0.3596511499401268,
+      "CIDEr": 1.2079796761019475,
+      "METEOR": 0.27879505939808324,
+      "ROUGE_L": 0.5606656485815291,
+      "epoch": 0.45,
+      "step": 9000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8725531067163008e-05,
+      "loss": 2.0857,
+      "step": 9001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.87232742831005e-05,
+      "loss": 2.1252,
+      "step": 9002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.872101742106732e-05,
+      "loss": 2.0688,
+      "step": 9003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8718760481122658e-05,
+      "loss": 2.0691,
+      "step": 9004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8716503463325695e-05,
+      "loss": 2.0669,
+      "step": 9005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.871424636773562e-05,
+      "loss": 2.074,
+      "step": 9006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.871198919441163e-05,
+      "loss": 2.1169,
+      "step": 9007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.87097319434129e-05,
+      "loss": 2.0789,
+      "step": 9008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8707474614798633e-05,
+      "loss": 2.0801,
+      "step": 9009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8705217208628028e-05,
+      "loss": 2.1077,
+      "step": 9010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8702959724960278e-05,
+      "loss": 2.0806,
+      "step": 9011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8700702163854594e-05,
+      "loss": 2.0771,
+      "step": 9012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.869844452537016e-05,
+      "loss": 1.991,
+      "step": 9013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.869618680956619e-05,
+      "loss": 2.0,
+      "step": 9014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8693929016501886e-05,
+      "loss": 2.0979,
+      "step": 9015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8691671146236464e-05,
+      "loss": 2.0957,
+      "step": 9016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8689413198829126e-05,
+      "loss": 2.082,
+      "step": 9017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.868715517433909e-05,
+      "loss": 2.0479,
+      "step": 9018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8684897072825565e-05,
+      "loss": 2.0396,
+      "step": 9019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8682638894347768e-05,
+      "loss": 2.0393,
+      "step": 9020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8680380638964918e-05,
+      "loss": 2.0361,
+      "step": 9021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.867812230673624e-05,
+      "loss": 2.0383,
+      "step": 9022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8675863897720948e-05,
+      "loss": 2.1082,
+      "step": 9023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8673605411978273e-05,
+      "loss": 2.0815,
+      "step": 9024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.867134684956743e-05,
+      "loss": 2.0671,
+      "step": 9025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8669088210547662e-05,
+      "loss": 2.0527,
+      "step": 9026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8666829494978192e-05,
+      "loss": 2.0048,
+      "step": 9027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.866457070291825e-05,
+      "loss": 2.0305,
+      "step": 9028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.866231183442707e-05,
+      "loss": 2.0181,
+      "step": 9029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8660052889563894e-05,
+      "loss": 2.042,
+      "step": 9030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.865779386838796e-05,
+      "loss": 2.0269,
+      "step": 9031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.86555347709585e-05,
+      "loss": 2.0815,
+      "step": 9032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8653275597334764e-05,
+      "loss": 2.083,
+      "step": 9033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.865101634757599e-05,
+      "loss": 2.1057,
+      "step": 9034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.864875702174143e-05,
+      "loss": 2.0942,
+      "step": 9035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.864649761989033e-05,
+      "loss": 2.0945,
+      "step": 9036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8644238142081944e-05,
+      "loss": 2.0981,
+      "step": 9037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8641978588375515e-05,
+      "loss": 2.0935,
+      "step": 9038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.86397189588303e-05,
+      "loss": 1.9944,
+      "step": 9039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8637459253505564e-05,
+      "loss": 2.0264,
+      "step": 9040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.863519947246056e-05,
+      "loss": 2.1042,
+      "step": 9041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8632939615754547e-05,
+      "loss": 2.1077,
+      "step": 9042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8630679683446782e-05,
+      "loss": 2.0686,
+      "step": 9043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8628419675596536e-05,
+      "loss": 2.0811,
+      "step": 9044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.862615959226307e-05,
+      "loss": 2.0271,
+      "step": 9045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8623899433505662e-05,
+      "loss": 2.0679,
+      "step": 9046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8621639199383575e-05,
+      "loss": 2.0359,
+      "step": 9047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8619378889956077e-05,
+      "loss": 2.052,
+      "step": 9048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.861711850528245e-05,
+      "loss": 2.0786,
+      "step": 9049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8614858045421964e-05,
+      "loss": 2.0898,
+      "step": 9050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.86125975104339e-05,
+      "loss": 2.1228,
+      "step": 9051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.861033690037754e-05,
+      "loss": 2.1323,
+      "step": 9052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8608076215312157e-05,
+      "loss": 2.106,
+      "step": 9053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8605815455297043e-05,
+      "loss": 2.0837,
+      "step": 9054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8603554620391484e-05,
+      "loss": 2.0686,
+      "step": 9055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8601293710654763e-05,
+      "loss": 2.0957,
+      "step": 9056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8599032726146177e-05,
+      "loss": 2.0182,
+      "step": 9057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.859677166692501e-05,
+      "loss": 2.0571,
+      "step": 9058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8594510533050556e-05,
+      "loss": 2.0725,
+      "step": 9059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.859224932458212e-05,
+      "loss": 2.0549,
+      "step": 9060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8589988041578984e-05,
+      "loss": 2.0237,
+      "step": 9061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8587726684100464e-05,
+      "loss": 2.0537,
+      "step": 9062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.858546525220585e-05,
+      "loss": 2.0256,
+      "step": 9063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.858320374595445e-05,
+      "loss": 2.0491,
+      "step": 9064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.858094216540557e-05,
+      "loss": 2.0603,
+      "step": 9065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8578680510618512e-05,
+      "loss": 2.0645,
+      "step": 9066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8576418781652588e-05,
+      "loss": 2.1111,
+      "step": 9067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8574156978567117e-05,
+      "loss": 2.1028,
+      "step": 9068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8571895101421405e-05,
+      "loss": 2.0425,
+      "step": 9069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8569633150274764e-05,
+      "loss": 2.0425,
+      "step": 9070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8567371125186515e-05,
+      "loss": 2.0496,
+      "step": 9071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8565109026215975e-05,
+      "loss": 2.0464,
+      "step": 9072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8562846853422473e-05,
+      "loss": 2.0232,
+      "step": 9073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8560584606865323e-05,
+      "loss": 2.0618,
+      "step": 9074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8558322286603845e-05,
+      "loss": 2.0659,
+      "step": 9075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.855605989269738e-05,
+      "loss": 2.0789,
+      "step": 9076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.855379742520525e-05,
+      "loss": 2.0198,
+      "step": 9077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8551534884186784e-05,
+      "loss": 2.0583,
+      "step": 9078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8549272269701317e-05,
+      "loss": 2.0598,
+      "step": 9079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8547009581808183e-05,
+      "loss": 2.1035,
+      "step": 9080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8544746820566716e-05,
+      "loss": 2.0308,
+      "step": 9081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8542483986036252e-05,
+      "loss": 2.0798,
+      "step": 9082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8540221078276144e-05,
+      "loss": 2.0194,
+      "step": 9083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.853795809734572e-05,
+      "loss": 2.0894,
+      "step": 9084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8535695043304325e-05,
+      "loss": 2.0959,
+      "step": 9085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8533431916211313e-05,
+      "loss": 2.0881,
+      "step": 9086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8531168716126027e-05,
+      "loss": 2.021,
+      "step": 9087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.852890544310782e-05,
+      "loss": 2.0659,
+      "step": 9088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.852664209721604e-05,
+      "loss": 1.9672,
+      "step": 9089
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8524378678510042e-05,
+      "loss": 2.0293,
+      "step": 9090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8522115187049182e-05,
+      "loss": 2.0481,
+      "step": 9091
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8519851622892818e-05,
+      "loss": 2.0674,
+      "step": 9092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8517587986100312e-05,
+      "loss": 1.9934,
+      "step": 9093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8515324276731018e-05,
+      "loss": 1.9917,
+      "step": 9094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.85130604948443e-05,
+      "loss": 2.0706,
+      "step": 9095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.851079664049953e-05,
+      "loss": 2.0679,
+      "step": 9096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8508532713756072e-05,
+      "loss": 2.0466,
+      "step": 9097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8506268714673294e-05,
+      "loss": 2.0374,
+      "step": 9098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8504004643310565e-05,
+      "loss": 2.0732,
+      "step": 9099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8501740499727256e-05,
+      "loss": 2.1011,
+      "step": 9100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8499476283982746e-05,
+      "loss": 2.1121,
+      "step": 9101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.849721199613641e-05,
+      "loss": 2.1404,
+      "step": 9102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.849494763624763e-05,
+      "loss": 2.1067,
+      "step": 9103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8492683204375782e-05,
+      "loss": 2.0645,
+      "step": 9104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8490418700580244e-05,
+      "loss": 2.0776,
+      "step": 9105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.848815412492041e-05,
+      "loss": 2.0671,
+      "step": 9106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8485889477455656e-05,
+      "loss": 2.1184,
+      "step": 9107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.848362475824538e-05,
+      "loss": 2.1077,
+      "step": 9108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8481359967348963e-05,
+      "loss": 2.0625,
+      "step": 9109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.84790951048258e-05,
+      "loss": 2.0769,
+      "step": 9110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8476830170735284e-05,
+      "loss": 2.0569,
+      "step": 9111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8474565165136807e-05,
+      "loss": 2.0562,
+      "step": 9112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8472300088089776e-05,
+      "loss": 2.0698,
+      "step": 9113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8470034939653578e-05,
+      "loss": 2.0984,
+      "step": 9114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.846776971988762e-05,
+      "loss": 2.0664,
+      "step": 9115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8465504428851306e-05,
+      "loss": 2.0464,
+      "step": 9116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.846323906660404e-05,
+      "loss": 2.073,
+      "step": 9117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.846097363320522e-05,
+      "loss": 2.072,
+      "step": 9118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.845870812871427e-05,
+      "loss": 1.9641,
+      "step": 9119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8456442553190587e-05,
+      "loss": 1.9475,
+      "step": 9120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8454176906693592e-05,
+      "loss": 2.0786,
+      "step": 9121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.845191118928269e-05,
+      "loss": 2.0935,
+      "step": 9122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8449645401017303e-05,
+      "loss": 2.0889,
+      "step": 9123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8447379541956848e-05,
+      "loss": 2.0627,
+      "step": 9124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.844511361216074e-05,
+      "loss": 2.0757,
+      "step": 9125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8442847611688408e-05,
+      "loss": 2.0613,
+      "step": 9126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8440581540599272e-05,
+      "loss": 2.092,
+      "step": 9127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8438315398952752e-05,
+      "loss": 2.1038,
+      "step": 9128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8436049186808283e-05,
+      "loss": 2.0923,
+      "step": 9129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.843378290422529e-05,
+      "loss": 2.1077,
+      "step": 9130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8431516551263206e-05,
+      "loss": 2.0049,
+      "step": 9131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8429250127981457e-05,
+      "loss": 2.0359,
+      "step": 9132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8426983634439484e-05,
+      "loss": 2.0742,
+      "step": 9133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8424717070696724e-05,
+      "loss": 2.0769,
+      "step": 9134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8422450436812606e-05,
+      "loss": 2.0759,
+      "step": 9135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8420183732846577e-05,
+      "loss": 2.0918,
+      "step": 9136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.841791695885808e-05,
+      "loss": 2.0188,
+      "step": 9137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8415650114906552e-05,
+      "loss": 2.0056,
+      "step": 9138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8413383201051448e-05,
+      "loss": 2.0955,
+      "step": 9139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8411116217352203e-05,
+      "loss": 2.0886,
+      "step": 9140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8408849163868274e-05,
+      "loss": 2.0398,
+      "step": 9141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.840658204065911e-05,
+      "loss": 2.0464,
+      "step": 9142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8404314847784165e-05,
+      "loss": 2.0485,
+      "step": 9143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8402047585302897e-05,
+      "loss": 2.0715,
+      "step": 9144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.839978025327475e-05,
+      "loss": 2.0725,
+      "step": 9145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8397512851759192e-05,
+      "loss": 2.1089,
+      "step": 9146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8395245380815682e-05,
+      "loss": 1.9852,
+      "step": 9147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.839297784050368e-05,
+      "loss": 2.021,
+      "step": 9148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8390710230882656e-05,
+      "loss": 2.0549,
+      "step": 9149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8388442552012063e-05,
+      "loss": 2.0496,
+      "step": 9150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8386174803951377e-05,
+      "loss": 2.0747,
+      "step": 9151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8383906986760063e-05,
+      "loss": 2.1072,
+      "step": 9152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8381639100497597e-05,
+      "loss": 2.0259,
+      "step": 9153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.837937114522345e-05,
+      "loss": 2.0376,
+      "step": 9154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.837710312099709e-05,
+      "loss": 2.0828,
+      "step": 9155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8374835027878e-05,
+      "loss": 2.105,
+      "step": 9156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.837256686592566e-05,
+      "loss": 2.0319,
+      "step": 9157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8370298635199542e-05,
+      "loss": 2.061,
+      "step": 9158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8368030335759134e-05,
+      "loss": 2.0706,
+      "step": 9159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8365761967663916e-05,
+      "loss": 2.0598,
+      "step": 9160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.836349353097338e-05,
+      "loss": 2.0347,
+      "step": 9161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8361225025747e-05,
+      "loss": 2.0518,
+      "step": 9162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.835895645204428e-05,
+      "loss": 2.0291,
+      "step": 9163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8356687809924703e-05,
+      "loss": 2.0217,
+      "step": 9164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.835441909944776e-05,
+      "loss": 2.0781,
+      "step": 9165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8352150320672945e-05,
+      "loss": 2.0798,
+      "step": 9166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8349881473659764e-05,
+      "loss": 2.0554,
+      "step": 9167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.83476125584677e-05,
+      "loss": 2.083,
+      "step": 9168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8345343575156265e-05,
+      "loss": 2.0881,
+      "step": 9169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8343074523784954e-05,
+      "loss": 2.0842,
+      "step": 9170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.834080540441327e-05,
+      "loss": 2.0891,
+      "step": 9171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8338536217100723e-05,
+      "loss": 2.0916,
+      "step": 9172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8336266961906813e-05,
+      "loss": 2.0383,
+      "step": 9173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8333997638891063e-05,
+      "loss": 2.0647,
+      "step": 9174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8331728248112958e-05,
+      "loss": 2.0598,
+      "step": 9175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.832945878963203e-05,
+      "loss": 2.0493,
+      "step": 9176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.832718926350779e-05,
+      "loss": 2.0825,
+      "step": 9177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.832491966979975e-05,
+      "loss": 2.0991,
+      "step": 9178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.832265000856743e-05,
+      "loss": 2.0374,
+      "step": 9179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.832038027987035e-05,
+      "loss": 2.0518,
+      "step": 9180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8318110483768022e-05,
+      "loss": 2.0103,
+      "step": 9181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8315840620319982e-05,
+      "loss": 2.0093,
+      "step": 9182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8313570689585746e-05,
+      "loss": 2.0896,
+      "step": 9183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8311300691624843e-05,
+      "loss": 2.085,
+      "step": 9184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8309030626496805e-05,
+      "loss": 2.1096,
+      "step": 9185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.830676049426115e-05,
+      "loss": 2.1067,
+      "step": 9186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8304490294977425e-05,
+      "loss": 2.0981,
+      "step": 9187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8302220028705147e-05,
+      "loss": 2.0999,
+      "step": 9188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8299949695503864e-05,
+      "loss": 2.062,
+      "step": 9189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8297679295433113e-05,
+      "loss": 2.053,
+      "step": 9190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8295408828552423e-05,
+      "loss": 2.0142,
+      "step": 9191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8293138294921338e-05,
+      "loss": 2.0396,
+      "step": 9192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.82908676945994e-05,
+      "loss": 2.1003,
+      "step": 9193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.828859702764616e-05,
+      "loss": 2.1489,
+      "step": 9194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.828632629412116e-05,
+      "loss": 2.0969,
+      "step": 9195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.828405549408394e-05,
+      "loss": 2.1052,
+      "step": 9196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8281784627594054e-05,
+      "loss": 2.0901,
+      "step": 9197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8279513694711055e-05,
+      "loss": 2.0959,
+      "step": 9198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8277242695494494e-05,
+      "loss": 2.0952,
+      "step": 9199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8274971630003927e-05,
+      "loss": 2.0823,
+      "step": 9200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8272700498298907e-05,
+      "loss": 2.0793,
+      "step": 9201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.827042930043899e-05,
+      "loss": 2.0708,
+      "step": 9202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.826815803648374e-05,
+      "loss": 2.0271,
+      "step": 9203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8265886706492718e-05,
+      "loss": 2.0864,
+      "step": 9204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8263615310525487e-05,
+      "loss": 2.0454,
+      "step": 9205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.826134384864161e-05,
+      "loss": 2.0491,
+      "step": 9206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8259072320900654e-05,
+      "loss": 2.0693,
+      "step": 9207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.825680072736219e-05,
+      "loss": 2.0679,
+      "step": 9208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8254529068085778e-05,
+      "loss": 2.0204,
+      "step": 9209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8252257343131e-05,
+      "loss": 2.073,
+      "step": 9210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8249985552557425e-05,
+      "loss": 2.04,
+      "step": 9211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.824771369642463e-05,
+      "loss": 2.0784,
+      "step": 9212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.824544177479219e-05,
+      "loss": 2.0571,
+      "step": 9213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8243169787719686e-05,
+      "loss": 2.074,
+      "step": 9214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.82408977352667e-05,
+      "loss": 2.1094,
+      "step": 9215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.823862561749281e-05,
+      "loss": 2.0725,
+      "step": 9216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8236353434457593e-05,
+      "loss": 2.1152,
+      "step": 9217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8234081186220648e-05,
+      "loss": 2.1018,
+      "step": 9218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8231808872841552e-05,
+      "loss": 2.0862,
+      "step": 9219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.82295364943799e-05,
+      "loss": 2.0688,
+      "step": 9220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.822726405089528e-05,
+      "loss": 2.0371,
+      "step": 9221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8224991542447283e-05,
+      "loss": 2.0688,
+      "step": 9222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8222718969095504e-05,
+      "loss": 2.0454,
+      "step": 9223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.822044633089954e-05,
+      "loss": 2.0647,
+      "step": 9224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.821817362791899e-05,
+      "loss": 2.1079,
+      "step": 9225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8215900860213447e-05,
+      "loss": 2.0908,
+      "step": 9226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.821362802784251e-05,
+      "loss": 2.0442,
+      "step": 9227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.821135513086579e-05,
+      "loss": 2.0442,
+      "step": 9228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8209082169342888e-05,
+      "loss": 1.9932,
+      "step": 9229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8206809143333405e-05,
+      "loss": 2.0249,
+      "step": 9230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8204536052896957e-05,
+      "loss": 2.1079,
+      "step": 9231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8202262898093147e-05,
+      "loss": 2.1094,
+      "step": 9232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.819998967898159e-05,
+      "loss": 2.0432,
+      "step": 9233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.819771639562189e-05,
+      "loss": 2.0652,
+      "step": 9234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8195443048073673e-05,
+      "loss": 2.0166,
+      "step": 9235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8193169636396544e-05,
+      "loss": 1.9917,
+      "step": 9236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.819089616065013e-05,
+      "loss": 2.0898,
+      "step": 9237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8188622620894043e-05,
+      "loss": 2.0557,
+      "step": 9238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8186349017187908e-05,
+      "loss": 2.0146,
+      "step": 9239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8184075349591347e-05,
+      "loss": 2.0366,
+      "step": 9240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8181801618163983e-05,
+      "loss": 2.0303,
+      "step": 9241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8179527822965446e-05,
+      "loss": 2.0605,
+      "step": 9242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8177253964055354e-05,
+      "loss": 2.0297,
+      "step": 9243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8174980041493345e-05,
+      "loss": 2.064,
+      "step": 9244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.817270605533905e-05,
+      "loss": 2.0532,
+      "step": 9245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8170432005652097e-05,
+      "loss": 2.0532,
+      "step": 9246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8168157892492124e-05,
+      "loss": 2.0596,
+      "step": 9247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8165883715918764e-05,
+      "loss": 2.0803,
+      "step": 9248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8163609475991657e-05,
+      "loss": 2.1003,
+      "step": 9249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.816133517277044e-05,
+      "loss": 2.0647,
+      "step": 9250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8159060806314763e-05,
+      "loss": 1.9971,
+      "step": 9251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8156786376684253e-05,
+      "loss": 2.0535,
+      "step": 9252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8154511883938563e-05,
+      "loss": 2.075,
+      "step": 9253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.815223732813734e-05,
+      "loss": 2.1226,
+      "step": 9254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8149962709340234e-05,
+      "loss": 2.1035,
+      "step": 9255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8147688027606887e-05,
+      "loss": 2.1394,
+      "step": 9256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8145413282996953e-05,
+      "loss": 2.0312,
+      "step": 9257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.814313847557008e-05,
+      "loss": 2.0386,
+      "step": 9258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.814086360538593e-05,
+      "loss": 2.0562,
+      "step": 9259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8138588672504156e-05,
+      "loss": 2.0583,
+      "step": 9260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8136313676984422e-05,
+      "loss": 2.0752,
+      "step": 9261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8134038618886375e-05,
+      "loss": 2.0703,
+      "step": 9262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8131763498269676e-05,
+      "loss": 2.0059,
+      "step": 9263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8129488315193998e-05,
+      "loss": 2.0442,
+      "step": 9264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8127213069718998e-05,
+      "loss": 2.1047,
+      "step": 9265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8124937761904338e-05,
+      "loss": 2.1035,
+      "step": 9266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8122662391809697e-05,
+      "loss": 2.0229,
+      "step": 9267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8120386959494737e-05,
+      "loss": 2.0684,
+      "step": 9268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8118111465019125e-05,
+      "loss": 2.0918,
+      "step": 9269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8115835908442537e-05,
+      "loss": 2.0911,
+      "step": 9270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8113560289824652e-05,
+      "loss": 2.0381,
+      "step": 9271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8111284609225137e-05,
+      "loss": 2.0603,
+      "step": 9272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8109008866703676e-05,
+      "loss": 1.9895,
+      "step": 9273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8106733062319942e-05,
+      "loss": 2.0239,
+      "step": 9274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8104457196133615e-05,
+      "loss": 2.0344,
+      "step": 9275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8102181268204386e-05,
+      "loss": 2.0447,
+      "step": 9276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8099905278591924e-05,
+      "loss": 2.0708,
+      "step": 9277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8097629227355925e-05,
+      "loss": 2.0999,
+      "step": 9278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8095353114556074e-05,
+      "loss": 2.0825,
+      "step": 9279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.809307694025206e-05,
+      "loss": 2.0879,
+      "step": 9280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8090800704503567e-05,
+      "loss": 2.0251,
+      "step": 9281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8088524407370295e-05,
+      "loss": 2.0793,
+      "step": 9282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8086248048911932e-05,
+      "loss": 2.0354,
+      "step": 9283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8083971629188175e-05,
+      "loss": 2.0408,
+      "step": 9284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8081695148258718e-05,
+      "loss": 2.0947,
+      "step": 9285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8079418606183263e-05,
+      "loss": 2.0996,
+      "step": 9286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8077142003021507e-05,
+      "loss": 2.05,
+      "step": 9287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.807486533883315e-05,
+      "loss": 2.0613,
+      "step": 9288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.80725886136779e-05,
+      "loss": 2.0303,
+      "step": 9289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.807031182761545e-05,
+      "loss": 2.0312,
+      "step": 9290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8068034980705524e-05,
+      "loss": 2.0381,
+      "step": 9291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8065758073007817e-05,
+      "loss": 2.0132,
+      "step": 9292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.806348110458204e-05,
+      "loss": 2.0098,
+      "step": 9293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.80612040754879e-05,
+      "loss": 2.0122,
+      "step": 9294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.805892698578512e-05,
+      "loss": 2.0461,
+      "step": 9295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.805664983553341e-05,
+      "loss": 2.0225,
+      "step": 9296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8054372624792482e-05,
+      "loss": 2.0608,
+      "step": 9297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8052095353622055e-05,
+      "loss": 2.0596,
+      "step": 9298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8049818022081848e-05,
+      "loss": 2.04,
+      "step": 9299
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.804754063023158e-05,
+      "loss": 2.0637,
+      "step": 9300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8045263178130975e-05,
+      "loss": 2.054,
+      "step": 9301
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8042985665839758e-05,
+      "loss": 2.0791,
+      "step": 9302
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8040708093417648e-05,
+      "loss": 2.0742,
+      "step": 9303
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8038430460924378e-05,
+      "loss": 2.0786,
+      "step": 9304
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8036152768419674e-05,
+      "loss": 2.0051,
+      "step": 9305
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8033875015963266e-05,
+      "loss": 2.0342,
+      "step": 9306
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8031597203614886e-05,
+      "loss": 2.0762,
+      "step": 9307
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8029319331434267e-05,
+      "loss": 2.0857,
+      "step": 9308
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.802704139948114e-05,
+      "loss": 2.0332,
+      "step": 9309
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8024763407815247e-05,
+      "loss": 2.0452,
+      "step": 9310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.802248535649632e-05,
+      "loss": 2.0803,
+      "step": 9311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8020207245584104e-05,
+      "loss": 2.0959,
+      "step": 9312
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8017929075138333e-05,
+      "loss": 2.1072,
+      "step": 9313
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8015650845218753e-05,
+      "loss": 2.0815,
+      "step": 9314
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.801337255588511e-05,
+      "loss": 1.9727,
+      "step": 9315
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.801109420719715e-05,
+      "loss": 2.0205,
+      "step": 9316
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8008815799214614e-05,
+      "loss": 2.1006,
+      "step": 9317
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8006537331997253e-05,
+      "loss": 2.0955,
+      "step": 9318
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8004258805604818e-05,
+      "loss": 2.1182,
+      "step": 9319
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8001980220097063e-05,
+      "loss": 2.0994,
+      "step": 9320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.799970157553374e-05,
+      "loss": 2.0657,
+      "step": 9321
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7997422871974607e-05,
+      "loss": 2.0593,
+      "step": 9322
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.799514410947941e-05,
+      "loss": 2.1233,
+      "step": 9323
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7992865288107916e-05,
+      "loss": 2.1113,
+      "step": 9324
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.799058640791988e-05,
+      "loss": 2.0718,
+      "step": 9325
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7988307468975066e-05,
+      "loss": 2.0847,
+      "step": 9326
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7986028471333235e-05,
+      "loss": 2.1106,
+      "step": 9327
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7983749415054156e-05,
+      "loss": 2.1211,
+      "step": 9328
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7981470300197582e-05,
+      "loss": 2.0054,
+      "step": 9329
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7979191126823293e-05,
+      "loss": 2.0601,
+      "step": 9330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.797691189499105e-05,
+      "loss": 2.0667,
+      "step": 9331
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7974632604760633e-05,
+      "loss": 2.0715,
+      "step": 9332
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.79723532561918e-05,
+      "loss": 2.0771,
+      "step": 9333
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.797007384934433e-05,
+      "loss": 2.0786,
+      "step": 9334
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7967794384278002e-05,
+      "loss": 2.0618,
+      "step": 9335
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7965514861052587e-05,
+      "loss": 2.0747,
+      "step": 9336
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7963235279727867e-05,
+      "loss": 2.0752,
+      "step": 9337
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7960955640363617e-05,
+      "loss": 2.0593,
+      "step": 9338
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7958675943019622e-05,
+      "loss": 2.022,
+      "step": 9339
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7956396187755663e-05,
+      "loss": 2.0381,
+      "step": 9340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7954116374631524e-05,
+      "loss": 2.0259,
+      "step": 9341
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7951836503706992e-05,
+      "loss": 2.0774,
+      "step": 9342
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7949556575041847e-05,
+      "loss": 2.0303,
+      "step": 9343
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7947276588695885e-05,
+      "loss": 2.0469,
+      "step": 9344
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.794499654472889e-05,
+      "loss": 2.0627,
+      "step": 9345
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7942716443200663e-05,
+      "loss": 2.0518,
+      "step": 9346
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7940436284170992e-05,
+      "loss": 2.0264,
+      "step": 9347
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7938156067699667e-05,
+      "loss": 2.0505,
+      "step": 9348
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.793587579384649e-05,
+      "loss": 2.0657,
+      "step": 9349
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7933595462671253e-05,
+      "loss": 2.0688,
+      "step": 9350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7931315074233757e-05,
+      "loss": 2.0293,
+      "step": 9351
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7929034628593806e-05,
+      "loss": 2.0679,
+      "step": 9352
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7926754125811203e-05,
+      "loss": 2.1064,
+      "step": 9353
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7924473565945742e-05,
+      "loss": 2.0889,
+      "step": 9354
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7922192949057236e-05,
+      "loss": 2.0537,
+      "step": 9355
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7919912275205494e-05,
+      "loss": 2.0732,
+      "step": 9356
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.791763154445032e-05,
+      "loss": 2.0747,
+      "step": 9357
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.791535075685152e-05,
+      "loss": 2.1191,
+      "step": 9358
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7913069912468907e-05,
+      "loss": 2.1006,
+      "step": 9359
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7910789011362298e-05,
+      "loss": 2.1021,
+      "step": 9360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7908508053591505e-05,
+      "loss": 2.084,
+      "step": 9361
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.790622703921634e-05,
+      "loss": 2.0901,
+      "step": 9362
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7903945968296626e-05,
+      "loss": 2.0713,
+      "step": 9363
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7901664840892175e-05,
+      "loss": 2.0581,
+      "step": 9364
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7899383657062807e-05,
+      "loss": 2.0383,
+      "step": 9365
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7897102416868346e-05,
+      "loss": 2.083,
+      "step": 9366
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.789482112036862e-05,
+      "loss": 1.9955,
+      "step": 9367
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7892539767623448e-05,
+      "loss": 2.054,
+      "step": 9368
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7890258358692655e-05,
+      "loss": 2.0151,
+      "step": 9369
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7887976893636066e-05,
+      "loss": 2.0127,
+      "step": 9370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.788569537251352e-05,
+      "loss": 1.9171,
+      "step": 9371
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7883413795384833e-05,
+      "loss": 1.9673,
+      "step": 9372
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7881132162309847e-05,
+      "loss": 2.0283,
+      "step": 9373
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7878850473348393e-05,
+      "loss": 2.0386,
+      "step": 9374
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7876568728560306e-05,
+      "loss": 2.0627,
+      "step": 9375
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.787428692800542e-05,
+      "loss": 2.0676,
+      "step": 9376
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7872005071743573e-05,
+      "loss": 2.0847,
+      "step": 9377
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7869723159834608e-05,
+      "loss": 2.0859,
+      "step": 9378
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.786744119233836e-05,
+      "loss": 2.0303,
+      "step": 9379
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7865159169314674e-05,
+      "loss": 2.0049,
+      "step": 9380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.786287709082339e-05,
+      "loss": 2.0823,
+      "step": 9381
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.786059495692436e-05,
+      "loss": 2.073,
+      "step": 9382
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7858312767677423e-05,
+      "loss": 2.0537,
+      "step": 9383
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.785603052314243e-05,
+      "loss": 2.0547,
+      "step": 9384
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.785374822337923e-05,
+      "loss": 2.014,
+      "step": 9385
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7851465868447673e-05,
+      "loss": 2.0298,
+      "step": 9386
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7849183458407613e-05,
+      "loss": 2.0911,
+      "step": 9387
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7846900993318895e-05,
+      "loss": 2.0959,
+      "step": 9388
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7844618473241388e-05,
+      "loss": 2.0591,
+      "step": 9389
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.784233589823494e-05,
+      "loss": 2.0667,
+      "step": 9390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.784005326835941e-05,
+      "loss": 2.0444,
+      "step": 9391
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7837770583674658e-05,
+      "loss": 2.0339,
+      "step": 9392
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7835487844240547e-05,
+      "loss": 2.0554,
+      "step": 9393
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7833205050116935e-05,
+      "loss": 2.074,
+      "step": 9394
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7830922201363685e-05,
+      "loss": 2.0408,
+      "step": 9395
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7828639298040666e-05,
+      "loss": 2.0566,
+      "step": 9396
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7826356340207743e-05,
+      "loss": 2.0891,
+      "step": 9397
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7824073327924788e-05,
+      "loss": 2.0808,
+      "step": 9398
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.782179026125166e-05,
+      "loss": 2.0947,
+      "step": 9399
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.781950714024824e-05,
+      "loss": 2.0991,
+      "step": 9400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7817223964974393e-05,
+      "loss": 2.0754,
+      "step": 9401
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7814940735489998e-05,
+      "loss": 2.0835,
+      "step": 9402
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.781265745185493e-05,
+      "loss": 2.0229,
+      "step": 9403
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7810374114129057e-05,
+      "loss": 2.0562,
+      "step": 9404
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7808090722372265e-05,
+      "loss": 2.0059,
+      "step": 9405
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.780580727664443e-05,
+      "loss": 2.0303,
+      "step": 9406
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.780352377700544e-05,
+      "loss": 2.0923,
+      "step": 9407
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7801240223515164e-05,
+      "loss": 2.0867,
+      "step": 9408
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7798956616233498e-05,
+      "loss": 2.0728,
+      "step": 9409
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.779667295522032e-05,
+      "loss": 2.0989,
+      "step": 9410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7794389240535517e-05,
+      "loss": 2.0439,
+      "step": 9411
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.779210547223898e-05,
+      "loss": 2.0686,
+      "step": 9412
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7789821650390597e-05,
+      "loss": 2.0161,
+      "step": 9413
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7787537775050258e-05,
+      "loss": 2.0681,
+      "step": 9414
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.778525384627785e-05,
+      "loss": 2.0886,
+      "step": 9415
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7782969864133276e-05,
+      "loss": 2.1201,
+      "step": 9416
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.778068582867642e-05,
+      "loss": 2.0522,
+      "step": 9417
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7778401739967192e-05,
+      "loss": 2.0359,
+      "step": 9418
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7776117598065475e-05,
+      "loss": 2.0366,
+      "step": 9419
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7773833403031178e-05,
+      "loss": 2.0359,
+      "step": 9420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7771549154924194e-05,
+      "loss": 2.0869,
+      "step": 9421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.776926485380443e-05,
+      "loss": 2.0813,
+      "step": 9422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.776698049973179e-05,
+      "loss": 2.0696,
+      "step": 9423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7764696092766173e-05,
+      "loss": 2.085,
+      "step": 9424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7762411632967492e-05,
+      "loss": 2.0791,
+      "step": 9425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7760127120395647e-05,
+      "loss": 2.0432,
+      "step": 9426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7757842555110553e-05,
+      "loss": 1.9883,
+      "step": 9427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7755557937172115e-05,
+      "loss": 1.9934,
+      "step": 9428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7753273266640254e-05,
+      "loss": 2.0747,
+      "step": 9429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.775098854357487e-05,
+      "loss": 2.1035,
+      "step": 9430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.774870376803588e-05,
+      "loss": 2.0515,
+      "step": 9431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7746418940083205e-05,
+      "loss": 2.0952,
+      "step": 9432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7744134059776756e-05,
+      "loss": 2.0298,
+      "step": 9433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7741849127176465e-05,
+      "loss": 2.0273,
+      "step": 9434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7739564142342234e-05,
+      "loss": 2.0381,
+      "step": 9435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.773727910533399e-05,
+      "loss": 2.0469,
+      "step": 9436
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7734994016211657e-05,
+      "loss": 2.0142,
+      "step": 9437
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7732708875035164e-05,
+      "loss": 2.0151,
+      "step": 9438
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7730423681864428e-05,
+      "loss": 2.0701,
+      "step": 9439
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7728138436759378e-05,
+      "loss": 2.0825,
+      "step": 9440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7725853139779942e-05,
+      "loss": 2.0913,
+      "step": 9441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.772356779098605e-05,
+      "loss": 2.0627,
+      "step": 9442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.772128239043763e-05,
+      "loss": 2.0447,
+      "step": 9443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.771899693819462e-05,
+      "loss": 2.0835,
+      "step": 9444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.771671143431695e-05,
+      "loss": 2.052,
+      "step": 9445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.771442587886455e-05,
+      "loss": 2.0571,
+      "step": 9446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.771214027189736e-05,
+      "loss": 2.0977,
+      "step": 9447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7709854613475324e-05,
+      "loss": 2.1094,
+      "step": 9448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7707568903658372e-05,
+      "loss": 2.0369,
+      "step": 9449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7705283142506444e-05,
+      "loss": 2.0554,
+      "step": 9450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7702997330079485e-05,
+      "loss": 2.0071,
+      "step": 9451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7700711466437437e-05,
+      "loss": 2.0261,
+      "step": 9452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7698425551640244e-05,
+      "loss": 2.0808,
+      "step": 9453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7696139585747852e-05,
+      "loss": 2.0884,
+      "step": 9454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.769385356882021e-05,
+      "loss": 2.082,
+      "step": 9455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7691567500917255e-05,
+      "loss": 2.1011,
+      "step": 9456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.768928138209895e-05,
+      "loss": 2.0173,
+      "step": 9457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7686995212425236e-05,
+      "loss": 2.0518,
+      "step": 9458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7684708991956075e-05,
+      "loss": 2.0383,
+      "step": 9459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7682422720751413e-05,
+      "loss": 2.0598,
+      "step": 9460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.768013639887121e-05,
+      "loss": 2.0066,
+      "step": 9461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7677850026375412e-05,
+      "loss": 2.0225,
+      "step": 9462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.767556360332399e-05,
+      "loss": 2.0312,
+      "step": 9463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.76732771297769e-05,
+      "loss": 2.0496,
+      "step": 9464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7670990605794088e-05,
+      "loss": 2.073,
+      "step": 9465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7668704031435534e-05,
+      "loss": 2.0728,
+      "step": 9466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7666417406761187e-05,
+      "loss": 2.0708,
+      "step": 9467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.766413073183102e-05,
+      "loss": 2.0715,
+      "step": 9468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7661844006705e-05,
+      "loss": 2.1008,
+      "step": 9469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.765955723144308e-05,
+      "loss": 2.0962,
+      "step": 9470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7657270406105243e-05,
+      "loss": 2.1011,
+      "step": 9471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7654983530751448e-05,
+      "loss": 2.0896,
+      "step": 9472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.765269660544167e-05,
+      "loss": 2.0659,
+      "step": 9473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7650409630235885e-05,
+      "loss": 2.0864,
+      "step": 9474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7648122605194058e-05,
+      "loss": 2.0554,
+      "step": 9475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7645835530376174e-05,
+      "loss": 2.0762,
+      "step": 9476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7643548405842192e-05,
+      "loss": 2.1033,
+      "step": 9477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7641261231652107e-05,
+      "loss": 2.1187,
+      "step": 9478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.763897400786589e-05,
+      "loss": 2.0491,
+      "step": 9479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7636686734543518e-05,
+      "loss": 2.0779,
+      "step": 9480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7634399411744978e-05,
+      "loss": 2.0493,
+      "step": 9481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7632112039530248e-05,
+      "loss": 1.9912,
+      "step": 9482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7629824617959315e-05,
+      "loss": 2.0774,
+      "step": 9483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.762753714709216e-05,
+      "loss": 2.0693,
+      "step": 9484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7625249626988773e-05,
+      "loss": 2.0264,
+      "step": 9485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.762296205770914e-05,
+      "loss": 2.0542,
+      "step": 9486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.762067443931325e-05,
+      "loss": 2.0879,
+      "step": 9487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7618386771861093e-05,
+      "loss": 2.0859,
+      "step": 9488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7616099055412658e-05,
+      "loss": 2.0486,
+      "step": 9489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7613811290027948e-05,
+      "loss": 2.0754,
+      "step": 9490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.761152347576694e-05,
+      "loss": 2.1084,
+      "step": 9491
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.760923561268964e-05,
+      "loss": 2.1006,
+      "step": 9492
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7606947700856048e-05,
+      "loss": 2.0044,
+      "step": 9493
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7604659740326155e-05,
+      "loss": 2.0354,
+      "step": 9494
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7602371731159958e-05,
+      "loss": 1.9988,
+      "step": 9495
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7600083673417463e-05,
+      "loss": 2.0134,
+      "step": 9496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.759779556715867e-05,
+      "loss": 2.0701,
+      "step": 9497
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7595507412443585e-05,
+      "loss": 2.0674,
+      "step": 9498
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7593219209332204e-05,
+      "loss": 2.0579,
+      "step": 9499
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.7590930957884542e-05,
+      "loss": 2.0623,
+      "step": 9500
+    },
+    {
+      "Bleu_1": 0.7751985836130648,
+      "Bleu_2": 0.6153043605240345,
+      "Bleu_3": 0.47371371569694143,
+      "Bleu_4": 0.3606923970198063,
+      "CIDEr": 1.2074462000305026,
+      "METEOR": 0.2795237599942428,
+      "ROUGE_L": 0.5609078626377818,
+      "epoch": 0.47,
+      "step": 9500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.75886426581606e-05,
+      "loss": 2.0338,
+      "step": 9501
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7586354310220388e-05,
+      "loss": 2.0879,
+      "step": 9502
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.758406591412391e-05,
+      "loss": 2.0618,
+      "step": 9503
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7581777469931188e-05,
+      "loss": 2.0667,
+      "step": 9504
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7579488977702223e-05,
+      "loss": 2.0508,
+      "step": 9505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7577200437497035e-05,
+      "loss": 2.0642,
+      "step": 9506
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7574911849375634e-05,
+      "loss": 2.0642,
+      "step": 9507
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7572623213398033e-05,
+      "loss": 2.0645,
+      "step": 9508
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7570334529624257e-05,
+      "loss": 2.0554,
+      "step": 9509
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.756804579811432e-05,
+      "loss": 2.0925,
+      "step": 9510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7565757018928238e-05,
+      "loss": 2.0266,
+      "step": 9511
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7563468192126037e-05,
+      "loss": 2.0544,
+      "step": 9512
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7561179317767734e-05,
+      "loss": 2.0669,
+      "step": 9513
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7558890395913352e-05,
+      "loss": 2.0725,
+      "step": 9514
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7556601426622926e-05,
+      "loss": 1.9861,
+      "step": 9515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.755431240995647e-05,
+      "loss": 2.0259,
+      "step": 9516
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7552023345974013e-05,
+      "loss": 2.0571,
+      "step": 9517
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7549734234735583e-05,
+      "loss": 2.0459,
+      "step": 9518
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7547445076301213e-05,
+      "loss": 2.1094,
+      "step": 9519
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7545155870730933e-05,
+      "loss": 2.116,
+      "step": 9520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.754286661808477e-05,
+      "loss": 2.0668,
+      "step": 9521
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.754057731842276e-05,
+      "loss": 2.0896,
+      "step": 9522
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7538287971804935e-05,
+      "loss": 2.073,
+      "step": 9523
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7535998578291335e-05,
+      "loss": 2.075,
+      "step": 9524
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7533709137941998e-05,
+      "loss": 2.0984,
+      "step": 9525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7531419650816953e-05,
+      "loss": 2.0952,
+      "step": 9526
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.752913011697625e-05,
+      "loss": 2.0774,
+      "step": 9527
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7526840536479917e-05,
+      "loss": 2.0825,
+      "step": 9528
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7524550909388004e-05,
+      "loss": 2.0381,
+      "step": 9529
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7522261235760552e-05,
+      "loss": 2.0503,
+      "step": 9530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7519971515657604e-05,
+      "loss": 2.0269,
+      "step": 9531
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.751768174913921e-05,
+      "loss": 2.0452,
+      "step": 9532
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.751539193626541e-05,
+      "loss": 2.1045,
+      "step": 9533
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7513102077096257e-05,
+      "loss": 2.1104,
+      "step": 9534
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7510812171691798e-05,
+      "loss": 2.0117,
+      "step": 9535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.750852222011208e-05,
+      "loss": 2.0674,
+      "step": 9536
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7506232222417157e-05,
+      "loss": 2.0457,
+      "step": 9537
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7503942178667084e-05,
+      "loss": 2.0479,
+      "step": 9538
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.750165208892191e-05,
+      "loss": 2.0627,
+      "step": 9539
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.749936195324169e-05,
+      "loss": 2.0933,
+      "step": 9540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7497071771686486e-05,
+      "loss": 2.0481,
+      "step": 9541
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.749478154431635e-05,
+      "loss": 2.0681,
+      "step": 9542
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.749249127119134e-05,
+      "loss": 2.0593,
+      "step": 9543
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.749020095237152e-05,
+      "loss": 2.0798,
+      "step": 9544
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.748791058791695e-05,
+      "loss": 2.0422,
+      "step": 9545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7485620177887692e-05,
+      "loss": 2.0598,
+      "step": 9546
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7483329722343804e-05,
+      "loss": 2.0933,
+      "step": 9547
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.748103922134536e-05,
+      "loss": 2.0481,
+      "step": 9548
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.747874867495241e-05,
+      "loss": 2.0293,
+      "step": 9549
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.747645808322504e-05,
+      "loss": 2.0293,
+      "step": 9550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7474167446223312e-05,
+      "loss": 2.0461,
+      "step": 9551
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.747187676400729e-05,
+      "loss": 2.0366,
+      "step": 9552
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7469586036637046e-05,
+      "loss": 2.0676,
+      "step": 9553
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7467295264172653e-05,
+      "loss": 2.0381,
+      "step": 9554
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.746500444667419e-05,
+      "loss": 2.0454,
+      "step": 9555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.746271358420172e-05,
+      "loss": 2.0308,
+      "step": 9556
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7460422676815324e-05,
+      "loss": 2.0232,
+      "step": 9557
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7458131724575076e-05,
+      "loss": 2.0464,
+      "step": 9558
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7455840727541053e-05,
+      "loss": 2.0322,
+      "step": 9559
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7453549685773343e-05,
+      "loss": 1.9937,
+      "step": 9560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7451258599332014e-05,
+      "loss": 2.1021,
+      "step": 9561
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7448967468277155e-05,
+      "loss": 2.0881,
+      "step": 9562
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7446676292668844e-05,
+      "loss": 2.0986,
+      "step": 9563
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7444385072567163e-05,
+      "loss": 2.0764,
+      "step": 9564
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.74420938080322e-05,
+      "loss": 2.0588,
+      "step": 9565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.743980249912404e-05,
+      "loss": 2.0569,
+      "step": 9566
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.743751114590277e-05,
+      "loss": 1.9993,
+      "step": 9567
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.743521974842848e-05,
+      "loss": 2.0693,
+      "step": 9568
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7432928306761253e-05,
+      "loss": 2.083,
+      "step": 9569
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7430636820961186e-05,
+      "loss": 2.072,
+      "step": 9570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.742834529108837e-05,
+      "loss": 2.0046,
+      "step": 9571
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7426053717202895e-05,
+      "loss": 2.0276,
+      "step": 9572
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7423762099364856e-05,
+      "loss": 2.1167,
+      "step": 9573
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.742147043763434e-05,
+      "loss": 2.1211,
+      "step": 9574
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.741917873207146e-05,
+      "loss": 2.0518,
+      "step": 9575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7416886982736303e-05,
+      "loss": 2.0662,
+      "step": 9576
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.741459518968897e-05,
+      "loss": 2.0791,
+      "step": 9577
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7412303352989553e-05,
+      "loss": 2.1021,
+      "step": 9578
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7410011472698162e-05,
+      "loss": 2.1038,
+      "step": 9579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7407719548874895e-05,
+      "loss": 2.1038,
+      "step": 9580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.740542758157986e-05,
+      "loss": 2.1255,
+      "step": 9581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7403135570873157e-05,
+      "loss": 2.1328,
+      "step": 9582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.740084351681488e-05,
+      "loss": 2.1021,
+      "step": 9583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7398551419465158e-05,
+      "loss": 2.0925,
+      "step": 9584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7396259278884086e-05,
+      "loss": 2.0376,
+      "step": 9585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7393967095131778e-05,
+      "loss": 2.064,
+      "step": 9586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.739167486826833e-05,
+      "loss": 2.0964,
+      "step": 9587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7389382598353873e-05,
+      "loss": 2.0916,
+      "step": 9588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7387090285448505e-05,
+      "loss": 2.0513,
+      "step": 9589
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7384797929612343e-05,
+      "loss": 2.0266,
+      "step": 9590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7382505530905505e-05,
+      "loss": 2.0732,
+      "step": 9591
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.73802130893881e-05,
+      "loss": 2.0793,
+      "step": 9592
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.737792060512025e-05,
+      "loss": 2.0776,
+      "step": 9593
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.737562807816207e-05,
+      "loss": 2.0713,
+      "step": 9594
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7373335508573677e-05,
+      "loss": 2.1072,
+      "step": 9595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7371042896415197e-05,
+      "loss": 2.1147,
+      "step": 9596
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7368750241746746e-05,
+      "loss": 2.1008,
+      "step": 9597
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.736645754462845e-05,
+      "loss": 2.1218,
+      "step": 9598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7364164805120422e-05,
+      "loss": 2.1079,
+      "step": 9599
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7361872023282802e-05,
+      "loss": 2.127,
+      "step": 9600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.735957919917571e-05,
+      "loss": 2.0398,
+      "step": 9601
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7357286332859265e-05,
+      "loss": 2.0403,
+      "step": 9602
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.73549934243936e-05,
+      "loss": 2.1111,
+      "step": 9603
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.735270047383885e-05,
+      "loss": 2.1279,
+      "step": 9604
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7350407481255137e-05,
+      "loss": 2.0642,
+      "step": 9605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7348114446702594e-05,
+      "loss": 2.0889,
+      "step": 9606
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7345821370241357e-05,
+      "loss": 2.0127,
+      "step": 9607
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7343528251931553e-05,
+      "loss": 2.0247,
+      "step": 9608
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7341235091833318e-05,
+      "loss": 2.0798,
+      "step": 9609
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7338941890006788e-05,
+      "loss": 2.1167,
+      "step": 9610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7336648646512104e-05,
+      "loss": 2.0754,
+      "step": 9611
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.73343553614094e-05,
+      "loss": 2.0635,
+      "step": 9612
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7332062034758813e-05,
+      "loss": 2.0413,
+      "step": 9613
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7329768666620483e-05,
+      "loss": 2.0554,
+      "step": 9614
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7327475257054555e-05,
+      "loss": 2.0923,
+      "step": 9615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7325181806121167e-05,
+      "loss": 2.0859,
+      "step": 9616
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7322888313880466e-05,
+      "loss": 2.0334,
+      "step": 9617
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7320594780392588e-05,
+      "loss": 2.0757,
+      "step": 9618
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.731830120571769e-05,
+      "loss": 2.0603,
+      "step": 9619
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7316007589915907e-05,
+      "loss": 2.0645,
+      "step": 9620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7313713933047394e-05,
+      "loss": 2.0745,
+      "step": 9621
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.73114202351723e-05,
+      "loss": 2.0825,
+      "step": 9622
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.730912649635077e-05,
+      "loss": 2.0359,
+      "step": 9623
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7306832716642958e-05,
+      "loss": 2.0127,
+      "step": 9624
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.730453889610901e-05,
+      "loss": 2.0698,
+      "step": 9625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7302245034809085e-05,
+      "loss": 2.0928,
+      "step": 9626
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7299951132803334e-05,
+      "loss": 2.0979,
+      "step": 9627
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7297657190151915e-05,
+      "loss": 2.0864,
+      "step": 9628
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.729536320691498e-05,
+      "loss": 2.0576,
+      "step": 9629
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7293069183152688e-05,
+      "loss": 2.0625,
+      "step": 9630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7290775118925193e-05,
+      "loss": 2.0562,
+      "step": 9631
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7288481014292665e-05,
+      "loss": 2.0471,
+      "step": 9632
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7286186869315252e-05,
+      "loss": 2.0442,
+      "step": 9633
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7283892684053122e-05,
+      "loss": 1.991,
+      "step": 9634
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7281598458566435e-05,
+      "loss": 2.0632,
+      "step": 9635
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.727930419291536e-05,
+      "loss": 2.0532,
+      "step": 9636
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.727700988716005e-05,
+      "loss": 1.9968,
+      "step": 9637
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7274715541360678e-05,
+      "loss": 2.0374,
+      "step": 9638
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7272421155577413e-05,
+      "loss": 2.0544,
+      "step": 9639
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.727012672987042e-05,
+      "loss": 2.0366,
+      "step": 9640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7267832264299865e-05,
+      "loss": 2.0574,
+      "step": 9641
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.726553775892592e-05,
+      "loss": 2.0652,
+      "step": 9642
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7263243213808756e-05,
+      "loss": 2.0854,
+      "step": 9643
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7260948629008544e-05,
+      "loss": 2.0693,
+      "step": 9644
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7258654004585456e-05,
+      "loss": 2.0505,
+      "step": 9645
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7256359340599667e-05,
+      "loss": 2.0776,
+      "step": 9646
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7254064637111355e-05,
+      "loss": 2.0298,
+      "step": 9647
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.725176989418069e-05,
+      "loss": 2.0371,
+      "step": 9648
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7249475111867855e-05,
+      "loss": 2.0435,
+      "step": 9649
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.724718029023302e-05,
+      "loss": 2.0435,
+      "step": 9650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.724488542933637e-05,
+      "loss": 2.1282,
+      "step": 9651
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7242590529238087e-05,
+      "loss": 2.1609,
+      "step": 9652
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.724029558999834e-05,
+      "loss": 2.0054,
+      "step": 9653
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7238000611677328e-05,
+      "loss": 2.0129,
+      "step": 9654
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7235705594335225e-05,
+      "loss": 1.9857,
+      "step": 9655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.723341053803222e-05,
+      "loss": 2.03,
+      "step": 9656
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.723111544282849e-05,
+      "loss": 2.0369,
+      "step": 9657
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7228820308784226e-05,
+      "loss": 2.0327,
+      "step": 9658
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7226525135959617e-05,
+      "loss": 2.0847,
+      "step": 9659
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7224229924414844e-05,
+      "loss": 2.0735,
+      "step": 9660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7221934674210106e-05,
+      "loss": 1.9607,
+      "step": 9661
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.721963938540559e-05,
+      "loss": 1.9861,
+      "step": 9662
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7217344058061485e-05,
+      "loss": 2.0344,
+      "step": 9663
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7215048692237983e-05,
+      "loss": 2.0598,
+      "step": 9664
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.721275328799528e-05,
+      "loss": 2.0071,
+      "step": 9665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7210457845393565e-05,
+      "loss": 1.9978,
+      "step": 9666
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.720816236449304e-05,
+      "loss": 1.9812,
+      "step": 9667
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7205866845353902e-05,
+      "loss": 2.0222,
+      "step": 9668
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7203571288036342e-05,
+      "loss": 2.0293,
+      "step": 9669
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.720127569260056e-05,
+      "loss": 2.0508,
+      "step": 9670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7198980059106756e-05,
+      "loss": 2.0742,
+      "step": 9671
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7196684387615135e-05,
+      "loss": 2.0605,
+      "step": 9672
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.719438867818589e-05,
+      "loss": 2.0623,
+      "step": 9673
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.719209293087923e-05,
+      "loss": 2.064,
+      "step": 9674
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7189797145755354e-05,
+      "loss": 2.0488,
+      "step": 9675
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.718750132287447e-05,
+      "loss": 2.0625,
+      "step": 9676
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7185205462296782e-05,
+      "loss": 2.0432,
+      "step": 9677
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7182909564082498e-05,
+      "loss": 2.053,
+      "step": 9678
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.718061362829182e-05,
+      "loss": 2.1023,
+      "step": 9679
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7178317654984955e-05,
+      "loss": 2.0981,
+      "step": 9680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7176021644222123e-05,
+      "loss": 2.0237,
+      "step": 9681
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7173725596063525e-05,
+      "loss": 1.998,
+      "step": 9682
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.717142951056938e-05,
+      "loss": 2.0767,
+      "step": 9683
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7169133387799888e-05,
+      "loss": 2.0779,
+      "step": 9684
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7166837227815272e-05,
+      "loss": 2.1228,
+      "step": 9685
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7164541030675742e-05,
+      "loss": 2.1531,
+      "step": 9686
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7162244796441516e-05,
+      "loss": 2.0508,
+      "step": 9687
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7159948525172807e-05,
+      "loss": 2.0957,
+      "step": 9688
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7157652216929832e-05,
+      "loss": 2.0366,
+      "step": 9689
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7155355871772817e-05,
+      "loss": 2.0532,
+      "step": 9690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.715305948976197e-05,
+      "loss": 2.0879,
+      "step": 9691
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7150763070957516e-05,
+      "loss": 2.0693,
+      "step": 9692
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7148466615419676e-05,
+      "loss": 2.0894,
+      "step": 9693
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7146170123208674e-05,
+      "loss": 2.0374,
+      "step": 9694
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7143873594384727e-05,
+      "loss": 2.1157,
+      "step": 9695
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7141577029008062e-05,
+      "loss": 2.106,
+      "step": 9696
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7139280427138904e-05,
+      "loss": 2.0608,
+      "step": 9697
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7136983788837483e-05,
+      "loss": 2.0667,
+      "step": 9698
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7134687114164017e-05,
+      "loss": 2.0981,
+      "step": 9699
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7132390403178736e-05,
+      "loss": 2.0798,
+      "step": 9700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7130093655941875e-05,
+      "loss": 2.0225,
+      "step": 9701
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7127796872513655e-05,
+      "loss": 2.0344,
+      "step": 9702
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7125500052954315e-05,
+      "loss": 2.0303,
+      "step": 9703
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7123203197324078e-05,
+      "loss": 2.0637,
+      "step": 9704
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7120906305683183e-05,
+      "loss": 2.083,
+      "step": 9705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.711860937809186e-05,
+      "loss": 2.1021,
+      "step": 9706
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.711631241461034e-05,
+      "loss": 2.0159,
+      "step": 9707
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.711401541529886e-05,
+      "loss": 2.0327,
+      "step": 9708
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7111718380217664e-05,
+      "loss": 2.0938,
+      "step": 9709
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.710942130942698e-05,
+      "loss": 2.0964,
+      "step": 9710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7107124202987055e-05,
+      "loss": 2.0586,
+      "step": 9711
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7104827060958115e-05,
+      "loss": 2.0801,
+      "step": 9712
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7102529883400412e-05,
+      "loss": 2.0964,
+      "step": 9713
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.710023267037418e-05,
+      "loss": 2.0459,
+      "step": 9714
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7097935421939664e-05,
+      "loss": 2.0559,
+      "step": 9715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.70956381381571e-05,
+      "loss": 2.042,
+      "step": 9716
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7093340819086742e-05,
+      "loss": 2.0669,
+      "step": 9717
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7091043464788832e-05,
+      "loss": 2.0657,
+      "step": 9718
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.708874607532361e-05,
+      "loss": 2.0769,
+      "step": 9719
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7086448650751325e-05,
+      "loss": 2.0762,
+      "step": 9720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7084151191132223e-05,
+      "loss": 2.0085,
+      "step": 9721
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7081853696526554e-05,
+      "loss": 2.0166,
+      "step": 9722
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.707955616699457e-05,
+      "loss": 2.0553,
+      "step": 9723
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7077258602596517e-05,
+      "loss": 2.0781,
+      "step": 9724
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7074961003392643e-05,
+      "loss": 2.0229,
+      "step": 9725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7072663369443207e-05,
+      "loss": 2.0476,
+      "step": 9726
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7070365700808456e-05,
+      "loss": 2.0605,
+      "step": 9727
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.706806799754865e-05,
+      "loss": 2.0552,
+      "step": 9728
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.706577025972404e-05,
+      "loss": 2.0774,
+      "step": 9729
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7063472487394877e-05,
+      "loss": 2.0696,
+      "step": 9730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7061174680621423e-05,
+      "loss": 2.1062,
+      "step": 9731
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7058876839463937e-05,
+      "loss": 2.1084,
+      "step": 9732
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7056578963982673e-05,
+      "loss": 2.106,
+      "step": 9733
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7054281054237893e-05,
+      "loss": 2.1021,
+      "step": 9734
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7051983110289855e-05,
+      "loss": 2.0752,
+      "step": 9735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.704968513219882e-05,
+      "loss": 2.1006,
+      "step": 9736
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7047387120025046e-05,
+      "loss": 2.0161,
+      "step": 9737
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7045089073828805e-05,
+      "loss": 2.0283,
+      "step": 9738
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7042790993670354e-05,
+      "loss": 2.0645,
+      "step": 9739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7040492879609962e-05,
+      "loss": 2.0508,
+      "step": 9740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.703819473170789e-05,
+      "loss": 2.0378,
+      "step": 9741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.70358965500244e-05,
+      "loss": 2.0771,
+      "step": 9742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7033598334619766e-05,
+      "loss": 2.0542,
+      "step": 9743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.703130008555426e-05,
+      "loss": 2.0374,
+      "step": 9744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7029001802888143e-05,
+      "loss": 2.0696,
+      "step": 9745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7026703486681685e-05,
+      "loss": 2.1211,
+      "step": 9746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.702440513699516e-05,
+      "loss": 2.0498,
+      "step": 9747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7022106753888844e-05,
+      "loss": 2.0293,
+      "step": 9748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7019808337423002e-05,
+      "loss": 2.1079,
+      "step": 9749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7017509887657907e-05,
+      "loss": 2.0979,
+      "step": 9750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.701521140465384e-05,
+      "loss": 1.9286,
+      "step": 9751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7012912888471063e-05,
+      "loss": 1.9688,
+      "step": 9752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.701061433916987e-05,
+      "loss": 2.0676,
+      "step": 9753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7008315756810524e-05,
+      "loss": 2.0674,
+      "step": 9754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.700601714145331e-05,
+      "loss": 2.1155,
+      "step": 9755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.70037184931585e-05,
+      "loss": 2.0737,
+      "step": 9756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.7001419811986378e-05,
+      "loss": 2.0676,
+      "step": 9757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.699912109799722e-05,
+      "loss": 2.0745,
+      "step": 9758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6996822351251315e-05,
+      "loss": 2.0649,
+      "step": 9759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6994523571808938e-05,
+      "loss": 2.0986,
+      "step": 9760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6992224759730375e-05,
+      "loss": 2.0671,
+      "step": 9761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.698992591507591e-05,
+      "loss": 2.0417,
+      "step": 9762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6987627037905822e-05,
+      "loss": 2.0236,
+      "step": 9763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6985328128280407e-05,
+      "loss": 2.0476,
+      "step": 9764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6983029186259946e-05,
+      "loss": 2.0503,
+      "step": 9765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.698073021190472e-05,
+      "loss": 2.0481,
+      "step": 9766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6978431205275018e-05,
+      "loss": 2.0868,
+      "step": 9767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6976132166431138e-05,
+      "loss": 2.1279,
+      "step": 9768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.697383309543337e-05,
+      "loss": 2.0293,
+      "step": 9769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.697153399234199e-05,
+      "loss": 2.0657,
+      "step": 9770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6969234857217302e-05,
+      "loss": 2.0635,
+      "step": 9771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6966935690119593e-05,
+      "loss": 2.0962,
+      "step": 9772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6964636491109156e-05,
+      "loss": 2.0764,
+      "step": 9773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6962337260246292e-05,
+      "loss": 2.064,
+      "step": 9774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6960037997591283e-05,
+      "loss": 2.1106,
+      "step": 9775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6957738703204435e-05,
+      "loss": 2.0771,
+      "step": 9776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6955439377146038e-05,
+      "loss": 2.0518,
+      "step": 9777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6953140019476396e-05,
+      "loss": 2.0566,
+      "step": 9778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6950840630255796e-05,
+      "loss": 2.0474,
+      "step": 9779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6948541209544548e-05,
+      "loss": 2.0586,
+      "step": 9780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.694624175740295e-05,
+      "loss": 2.0242,
+      "step": 9781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6943942273891292e-05,
+      "loss": 2.0493,
+      "step": 9782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.694164275906989e-05,
+      "loss": 2.0295,
+      "step": 9783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.693934321299904e-05,
+      "loss": 2.0442,
+      "step": 9784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.693704363573904e-05,
+      "loss": 2.084,
+      "step": 9785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.69347440273502e-05,
+      "loss": 2.116,
+      "step": 9786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6932444387892817e-05,
+      "loss": 2.085,
+      "step": 9787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6930144717427206e-05,
+      "loss": 2.062,
+      "step": 9788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6927845016013667e-05,
+      "loss": 2.0388,
+      "step": 9789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6925545283712516e-05,
+      "loss": 2.0503,
+      "step": 9790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6923245520584044e-05,
+      "loss": 2.085,
+      "step": 9791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6920945726688575e-05,
+      "loss": 2.1196,
+      "step": 9792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6918645902086407e-05,
+      "loss": 2.0833,
+      "step": 9793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.691634604683786e-05,
+      "loss": 2.0759,
+      "step": 9794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6914046161003244e-05,
+      "loss": 2.0474,
+      "step": 9795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6911746244642863e-05,
+      "loss": 2.0457,
+      "step": 9796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6909446297817035e-05,
+      "loss": 2.051,
+      "step": 9797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6907146320586075e-05,
+      "loss": 2.0605,
+      "step": 9798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6904846313010292e-05,
+      "loss": 2.1047,
+      "step": 9799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6902546275150012e-05,
+      "loss": 2.1062,
+      "step": 9800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6900246207065536e-05,
+      "loss": 2.0542,
+      "step": 9801
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.689794610881719e-05,
+      "loss": 2.0315,
+      "step": 9802
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6895645980465286e-05,
+      "loss": 2.0476,
+      "step": 9803
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6893345822070147e-05,
+      "loss": 2.0515,
+      "step": 9804
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.68910456336921e-05,
+      "loss": 2.0618,
+      "step": 9805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6888745415391442e-05,
+      "loss": 2.0713,
+      "step": 9806
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.688644516722851e-05,
+      "loss": 2.1084,
+      "step": 9807
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6884144889263618e-05,
+      "loss": 2.1091,
+      "step": 9808
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.68818445815571e-05,
+      "loss": 2.0356,
+      "step": 9809
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.687954424416927e-05,
+      "loss": 2.0537,
+      "step": 9810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6877243877160448e-05,
+      "loss": 2.0991,
+      "step": 9811
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6874943480590964e-05,
+      "loss": 2.0913,
+      "step": 9812
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6872643054521142e-05,
+      "loss": 2.1045,
+      "step": 9813
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.687034259901131e-05,
+      "loss": 2.1023,
+      "step": 9814
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6868042114121793e-05,
+      "loss": 2.0798,
+      "step": 9815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6865741599912917e-05,
+      "loss": 2.061,
+      "step": 9816
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6863441056445013e-05,
+      "loss": 2.0592,
+      "step": 9817
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.686114048377841e-05,
+      "loss": 2.0886,
+      "step": 9818
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6858839881973434e-05,
+      "loss": 2.0769,
+      "step": 9819
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6856539251090422e-05,
+      "loss": 2.0823,
+      "step": 9820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.68542385911897e-05,
+      "loss": 2.0605,
+      "step": 9821
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6851937902331603e-05,
+      "loss": 2.0579,
+      "step": 9822
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.684963718457646e-05,
+      "loss": 2.0566,
+      "step": 9823
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6847336437984612e-05,
+      "loss": 2.0525,
+      "step": 9824
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6845035662616386e-05,
+      "loss": 2.0593,
+      "step": 9825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6842734858532124e-05,
+      "loss": 2.0857,
+      "step": 9826
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6840434025792155e-05,
+      "loss": 2.0535,
+      "step": 9827
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6838133164456814e-05,
+      "loss": 2.0498,
+      "step": 9828
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.683583227458645e-05,
+      "loss": 2.0459,
+      "step": 9829
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6833531356241392e-05,
+      "loss": 2.0295,
+      "step": 9830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.683123040948198e-05,
+      "loss": 2.051,
+      "step": 9831
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6828929434368555e-05,
+      "loss": 2.0366,
+      "step": 9832
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.682662843096146e-05,
+      "loss": 2.0537,
+      "step": 9833
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.682432739932103e-05,
+      "loss": 2.0659,
+      "step": 9834
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6822026339507614e-05,
+      "loss": 2.0518,
+      "step": 9835
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6819725251581548e-05,
+      "loss": 2.0627,
+      "step": 9836
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.681742413560318e-05,
+      "loss": 2.0364,
+      "step": 9837
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.681512299163285e-05,
+      "loss": 2.0815,
+      "step": 9838
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6812821819730904e-05,
+      "loss": 2.0378,
+      "step": 9839
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6810520619957694e-05,
+      "loss": 2.0581,
+      "step": 9840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.680821939237356e-05,
+      "loss": 2.1179,
+      "step": 9841
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.680591813703885e-05,
+      "loss": 2.1042,
+      "step": 9842
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.680361685401391e-05,
+      "loss": 2.0828,
+      "step": 9843
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6801315543359092e-05,
+      "loss": 2.1057,
+      "step": 9844
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6799014205134742e-05,
+      "loss": 2.1304,
+      "step": 9845
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6796712839401212e-05,
+      "loss": 2.1003,
+      "step": 9846
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.679441144621885e-05,
+      "loss": 2.0881,
+      "step": 9847
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.679211002564801e-05,
+      "loss": 2.0859,
+      "step": 9848
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6789808577749045e-05,
+      "loss": 2.0513,
+      "step": 9849
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6787507102582304e-05,
+      "loss": 2.0911,
+      "step": 9850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.678520560020815e-05,
+      "loss": 2.0854,
+      "step": 9851
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.678290407068692e-05,
+      "loss": 2.0969,
+      "step": 9852
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6780602514078985e-05,
+      "loss": 2.0823,
+      "step": 9853
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6778300930444695e-05,
+      "loss": 2.0991,
+      "step": 9854
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6775999319844402e-05,
+      "loss": 2.0947,
+      "step": 9855
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6773697682338473e-05,
+      "loss": 2.1174,
+      "step": 9856
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.677139601798726e-05,
+      "loss": 2.0847,
+      "step": 9857
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6769094326851114e-05,
+      "loss": 2.0679,
+      "step": 9858
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6766792608990406e-05,
+      "loss": 1.9761,
+      "step": 9859
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764490864465493e-05,
+      "loss": 1.9961,
+      "step": 9860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762189093336736e-05,
+      "loss": 2.0979,
+      "step": 9861
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675988729566449e-05,
+      "loss": 2.0774,
+      "step": 9862
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675758547150912e-05,
+      "loss": 2.1089,
+      "step": 9863
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6755283620930992e-05,
+      "loss": 2.1113,
+      "step": 9864
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752981743990466e-05,
+      "loss": 2.0264,
+      "step": 9865
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6750679840747906e-05,
+      "loss": 2.012,
+      "step": 9866
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674837791126368e-05,
+      "loss": 2.0703,
+      "step": 9867
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6746075955598153e-05,
+      "loss": 2.1057,
+      "step": 9868
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674377397381169e-05,
+      "loss": 2.0735,
+      "step": 9869
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741471965964657e-05,
+      "loss": 2.0669,
+      "step": 9870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673916993211742e-05,
+      "loss": 1.9929,
+      "step": 9871
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673686787233035e-05,
+      "loss": 1.9712,
+      "step": 9872
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6734565786663816e-05,
+      "loss": 2.0642,
+      "step": 9873
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6732263675178183e-05,
+      "loss": 2.0603,
+      "step": 9874
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729961537933827e-05,
+      "loss": 2.0876,
+      "step": 9875
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672765937499112e-05,
+      "loss": 2.0737,
+      "step": 9876
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6725357186410427e-05,
+      "loss": 2.0764,
+      "step": 9877
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672305497225212e-05,
+      "loss": 2.0757,
+      "step": 9878
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6720752732576576e-05,
+      "loss": 2.0718,
+      "step": 9879
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671845046744417e-05,
+      "loss": 2.103,
+      "step": 9880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6716148176915278e-05,
+      "loss": 2.0332,
+      "step": 9881
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6713845861050267e-05,
+      "loss": 2.0569,
+      "step": 9882
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6711543519909516e-05,
+      "loss": 2.0154,
+      "step": 9883
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67092411535534e-05,
+      "loss": 2.0479,
+      "step": 9884
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67069387620423e-05,
+      "loss": 2.0996,
+      "step": 9885
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670463634543659e-05,
+      "loss": 2.1104,
+      "step": 9886
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670233390379665e-05,
+      "loss": 2.0881,
+      "step": 9887
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670003143718286e-05,
+      "loss": 2.075,
+      "step": 9888
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6697728945655593e-05,
+      "loss": 1.9882,
+      "step": 9889
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6695426429275238e-05,
+      "loss": 1.9944,
+      "step": 9890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.669312388810217e-05,
+      "loss": 2.0696,
+      "step": 9891
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6690821322196775e-05,
+      "loss": 2.0793,
+      "step": 9892
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.668851873161943e-05,
+      "loss": 2.0112,
+      "step": 9893
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.668621611643052e-05,
+      "loss": 2.0293,
+      "step": 9894
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6683913476690427e-05,
+      "loss": 2.042,
+      "step": 9895
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.668161081245954e-05,
+      "loss": 2.0608,
+      "step": 9896
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.667930812379824e-05,
+      "loss": 2.0923,
+      "step": 9897
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.667700541076691e-05,
+      "loss": 2.0774,
+      "step": 9898
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.667470267342594e-05,
+      "loss": 2.1052,
+      "step": 9899
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.667239991183571e-05,
+      "loss": 2.1096,
+      "step": 9900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6670097126056615e-05,
+      "loss": 2.0297,
+      "step": 9901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667794316149044e-05,
+      "loss": 2.0725,
+      "step": 9902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666549148217338e-05,
+      "loss": 2.0896,
+      "step": 9903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6663188624190014e-05,
+      "loss": 2.0925,
+      "step": 9904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6660885742259336e-05,
+      "loss": 2.0659,
+      "step": 9905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658582836441733e-05,
+      "loss": 2.0491,
+      "step": 9906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6656279906797604e-05,
+      "loss": 2.0349,
+      "step": 9907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6653976953387337e-05,
+      "loss": 2.0305,
+      "step": 9908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6651673976271318e-05,
+      "loss": 2.0432,
+      "step": 9909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6649370975509947e-05,
+      "loss": 2.0649,
+      "step": 9910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6647067951163615e-05,
+      "loss": 2.0276,
+      "step": 9911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6644764903292718e-05,
+      "loss": 2.0103,
+      "step": 9912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664246183195765e-05,
+      "loss": 2.0216,
+      "step": 9913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6640158737218804e-05,
+      "loss": 2.0854,
+      "step": 9914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6637855619136574e-05,
+      "loss": 2.0774,
+      "step": 9915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6635552477771366e-05,
+      "loss": 2.0693,
+      "step": 9916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6633249313183574e-05,
+      "loss": 2.0532,
+      "step": 9917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663094612543359e-05,
+      "loss": 2.0698,
+      "step": 9918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6628642914581812e-05,
+      "loss": 2.0918,
+      "step": 9919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6626339680688647e-05,
+      "loss": 2.0903,
+      "step": 9920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.662403642381449e-05,
+      "loss": 2.0276,
+      "step": 9921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6621733144019742e-05,
+      "loss": 2.0483,
+      "step": 9922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6619429841364804e-05,
+      "loss": 2.032,
+      "step": 9923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6617126515910074e-05,
+      "loss": 2.052,
+      "step": 9924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.661482316771596e-05,
+      "loss": 2.0515,
+      "step": 9925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.661251979684286e-05,
+      "loss": 2.0696,
+      "step": 9926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.661021640335118e-05,
+      "loss": 2.0623,
+      "step": 9927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6607912987301322e-05,
+      "loss": 2.071,
+      "step": 9928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6605609548753694e-05,
+      "loss": 2.1008,
+      "step": 9929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6603306087768694e-05,
+      "loss": 2.0947,
+      "step": 9930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6601002604406737e-05,
+      "loss": 2.1038,
+      "step": 9931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.659869909872822e-05,
+      "loss": 2.1157,
+      "step": 9932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6596395570793557e-05,
+      "loss": 1.9835,
+      "step": 9933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.659409202066315e-05,
+      "loss": 2.0417,
+      "step": 9934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.659178844839741e-05,
+      "loss": 2.0369,
+      "step": 9935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6589484854056738e-05,
+      "loss": 2.0215,
+      "step": 9936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6587181237701556e-05,
+      "loss": 2.0754,
+      "step": 9937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6584877599392265e-05,
+      "loss": 2.0867,
+      "step": 9938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.658257393918928e-05,
+      "loss": 2.0933,
+      "step": 9939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6580270257153005e-05,
+      "loss": 2.0732,
+      "step": 9940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6577966553343863e-05,
+      "loss": 2.0388,
+      "step": 9941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6575662827822254e-05,
+      "loss": 2.0452,
+      "step": 9942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6573359080648596e-05,
+      "loss": 2.0776,
+      "step": 9943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.65710553118833e-05,
+      "loss": 2.0647,
+      "step": 9944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6568751521586784e-05,
+      "loss": 2.0564,
+      "step": 9945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6566447709819454e-05,
+      "loss": 2.0269,
+      "step": 9946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6564143876641737e-05,
+      "loss": 2.0332,
+      "step": 9947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6561840022114035e-05,
+      "loss": 2.0447,
+      "step": 9948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6559536146296774e-05,
+      "loss": 2.0293,
+      "step": 9949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6557232249250364e-05,
+      "loss": 2.0244,
+      "step": 9950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6554928331035227e-05,
+      "loss": 2.0359,
+      "step": 9951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6552624391711776e-05,
+      "loss": 2.0662,
+      "step": 9952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6550320431340433e-05,
+      "loss": 2.0498,
+      "step": 9953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6548016449981617e-05,
+      "loss": 2.0725,
+      "step": 9954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6545712447695743e-05,
+      "loss": 2.0989,
+      "step": 9955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6543408424543234e-05,
+      "loss": 2.0996,
+      "step": 9956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6541104380584508e-05,
+      "loss": 2.002,
+      "step": 9957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6538800315879987e-05,
+      "loss": 2.0571,
+      "step": 9958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6536496230490095e-05,
+      "loss": 2.0447,
+      "step": 9959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.653419212447525e-05,
+      "loss": 2.0574,
+      "step": 9960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6531887997895876e-05,
+      "loss": 2.0764,
+      "step": 9961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6529583850812403e-05,
+      "loss": 2.0762,
+      "step": 9962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.652727968328524e-05,
+      "loss": 2.0681,
+      "step": 9963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.652497549537482e-05,
+      "loss": 2.0552,
+      "step": 9964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.652267128714157e-05,
+      "loss": 2.0178,
+      "step": 9965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6520367058645907e-05,
+      "loss": 2.0452,
+      "step": 9966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6518062809948266e-05,
+      "loss": 2.0225,
+      "step": 9967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6515758541109072e-05,
+      "loss": 2.0149,
+      "step": 9968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.651345425218874e-05,
+      "loss": 2.0962,
+      "step": 9969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.651114994324771e-05,
+      "loss": 2.0989,
+      "step": 9970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6508845614346407e-05,
+      "loss": 2.0183,
+      "step": 9971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.650654126554526e-05,
+      "loss": 2.0503,
+      "step": 9972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.650423689690469e-05,
+      "loss": 2.1003,
+      "step": 9973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6501932508485133e-05,
+      "loss": 2.1011,
+      "step": 9974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.649962810034702e-05,
+      "loss": 2.0774,
+      "step": 9975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6497323672550782e-05,
+      "loss": 2.0942,
+      "step": 9976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6495019225156847e-05,
+      "loss": 2.1177,
+      "step": 9977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.649271475822565e-05,
+      "loss": 2.0676,
+      "step": 9978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6490410271817614e-05,
+      "loss": 2.0955,
+      "step": 9979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.648810576599318e-05,
+      "loss": 2.0955,
+      "step": 9980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6485801240812778e-05,
+      "loss": 2.0481,
+      "step": 9981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.648349669633684e-05,
+      "loss": 2.0442,
+      "step": 9982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.648119213262581e-05,
+      "loss": 2.0835,
+      "step": 9983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6478887549740107e-05,
+      "loss": 2.0999,
+      "step": 9984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.647658294774018e-05,
+      "loss": 2.0525,
+      "step": 9985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.647427832668645e-05,
+      "loss": 2.0686,
+      "step": 9986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6471973686639365e-05,
+      "loss": 2.0774,
+      "step": 9987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.646966902765936e-05,
+      "loss": 2.1072,
+      "step": 9988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6467364349806866e-05,
+      "loss": 2.0735,
+      "step": 9989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6465059653142325e-05,
+      "loss": 2.1016,
+      "step": 9990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6462754937726176e-05,
+      "loss": 1.991,
+      "step": 9991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6460450203618853e-05,
+      "loss": 2.0054,
+      "step": 9992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.64581454508808e-05,
+      "loss": 2.103,
+      "step": 9993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.645584067957245e-05,
+      "loss": 2.1233,
+      "step": 9994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6453535889754253e-05,
+      "loss": 2.0249,
+      "step": 9995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6451231081486637e-05,
+      "loss": 2.0339,
+      "step": 9996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6448926254830054e-05,
+      "loss": 2.0962,
+      "step": 9997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6446621409844942e-05,
+      "loss": 2.0874,
+      "step": 9998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.644431654659174e-05,
+      "loss": 2.0737,
+      "step": 9999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.644201166513089e-05,
+      "loss": 2.0933,
+      "step": 10000
+    },
+    {
+      "Bleu_1": 0.7734011273006944,
+      "Bleu_2": 0.6134496749545776,
+      "Bleu_3": 0.4716705488676403,
+      "Bleu_4": 0.3583087417981248,
+      "CIDEr": 1.2056254715073709,
+      "METEOR": 0.27863516708627706,
+      "ROUGE_L": 0.5600760336463185,
+      "epoch": 0.5,
+      "step": 10000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.643970676552284e-05,
+      "loss": 2.0542,
+      "step": 10001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6437401847828026e-05,
+      "loss": 2.0569,
+      "step": 10002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.64350969121069e-05,
+      "loss": 2.0623,
+      "step": 10003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.64327919584199e-05,
+      "loss": 2.0371,
+      "step": 10004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6430486986827475e-05,
+      "loss": 2.0527,
+      "step": 10005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6428181997390065e-05,
+      "loss": 2.0769,
+      "step": 10006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.642587699016812e-05,
+      "loss": 2.0559,
+      "step": 10007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6423571965222087e-05,
+      "loss": 1.999,
+      "step": 10008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.642126692261241e-05,
+      "loss": 2.0601,
+      "step": 10009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6418961862399538e-05,
+      "loss": 2.0583,
+      "step": 10010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6416656784643913e-05,
+      "loss": 2.0784,
+      "step": 10011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.641435168940599e-05,
+      "loss": 2.0674,
+      "step": 10012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6412046576746215e-05,
+      "loss": 2.084,
+      "step": 10013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6409741446725036e-05,
+      "loss": 2.0859,
+      "step": 10014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.64074362994029e-05,
+      "loss": 2.0891,
+      "step": 10015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6405131134840264e-05,
+      "loss": 2.0979,
+      "step": 10016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6402825953097573e-05,
+      "loss": 2.0642,
+      "step": 10017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6400520754235278e-05,
+      "loss": 2.0754,
+      "step": 10018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6398215538313832e-05,
+      "loss": 2.0109,
+      "step": 10019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6395910305393688e-05,
+      "loss": 2.0425,
+      "step": 10020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.639360505553529e-05,
+      "loss": 2.0549,
+      "step": 10021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6391299788799097e-05,
+      "loss": 2.0359,
+      "step": 10022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6388994505245567e-05,
+      "loss": 2.0508,
+      "step": 10023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6386689204935143e-05,
+      "loss": 2.0508,
+      "step": 10024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.638438388792828e-05,
+      "loss": 2.0828,
+      "step": 10025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6382078554285442e-05,
+      "loss": 2.1094,
+      "step": 10026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6379773204067067e-05,
+      "loss": 2.0725,
+      "step": 10027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6377467837333627e-05,
+      "loss": 2.0818,
+      "step": 10028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6375162454145576e-05,
+      "loss": 2.0576,
+      "step": 10029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6372857054563358e-05,
+      "loss": 2.085,
+      "step": 10030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6370551638647438e-05,
+      "loss": 2.1135,
+      "step": 10031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.636824620645827e-05,
+      "loss": 2.071,
+      "step": 10032
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6365940758056315e-05,
+      "loss": 1.9882,
+      "step": 10033
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.636363529350203e-05,
+      "loss": 2.0447,
+      "step": 10034
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6361329812855867e-05,
+      "loss": 2.0688,
+      "step": 10035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6359024316178293e-05,
+      "loss": 2.0862,
+      "step": 10036
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6356718803529757e-05,
+      "loss": 2.0884,
+      "step": 10037
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6354413274970726e-05,
+      "loss": 2.1228,
+      "step": 10038
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6352107730561663e-05,
+      "loss": 2.094,
+      "step": 10039
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.634980217036302e-05,
+      "loss": 2.1106,
+      "step": 10040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.634749659443526e-05,
+      "loss": 2.082,
+      "step": 10041
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6345191002838844e-05,
+      "loss": 2.1111,
+      "step": 10042
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6342885395634234e-05,
+      "loss": 2.1099,
+      "step": 10043
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6340579772881897e-05,
+      "loss": 2.1145,
+      "step": 10044
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6338274134642285e-05,
+      "loss": 2.0916,
+      "step": 10045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6335968480975872e-05,
+      "loss": 2.0981,
+      "step": 10046
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6333662811943112e-05,
+      "loss": 2.1125,
+      "step": 10047
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6331357127604473e-05,
+      "loss": 2.0935,
+      "step": 10048
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.632905142802042e-05,
+      "loss": 2.0537,
+      "step": 10049
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.632674571325141e-05,
+      "loss": 2.0544,
+      "step": 10050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6324439983357915e-05,
+      "loss": 2.093,
+      "step": 10051
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6322134238400398e-05,
+      "loss": 2.0894,
+      "step": 10052
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6319828478439322e-05,
+      "loss": 2.0427,
+      "step": 10053
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.631752270353516e-05,
+      "loss": 2.0134,
+      "step": 10054
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6315216913748372e-05,
+      "loss": 2.0776,
+      "step": 10055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6312911109139424e-05,
+      "loss": 2.0557,
+      "step": 10056
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6310605289768785e-05,
+      "loss": 2.0308,
+      "step": 10057
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6308299455696924e-05,
+      "loss": 2.0391,
+      "step": 10058
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6305993606984312e-05,
+      "loss": 2.0613,
+      "step": 10059
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6303687743691406e-05,
+      "loss": 2.0481,
+      "step": 10060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6301381865878682e-05,
+      "loss": 2.0459,
+      "step": 10061
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6299075973606608e-05,
+      "loss": 2.0647,
+      "step": 10062
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6296770066935655e-05,
+      "loss": 2.0977,
+      "step": 10063
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6294464145926293e-05,
+      "loss": 2.0793,
+      "step": 10064
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6292158210638993e-05,
+      "loss": 2.0193,
+      "step": 10065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6289852261134213e-05,
+      "loss": 2.0398,
+      "step": 10066
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.628754629747244e-05,
+      "loss": 2.0789,
+      "step": 10067
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6285240319714142e-05,
+      "loss": 2.0879,
+      "step": 10068
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6282934327919786e-05,
+      "loss": 2.1169,
+      "step": 10069
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6280628322149843e-05,
+      "loss": 2.072,
+      "step": 10070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6278322302464786e-05,
+      "loss": 2.0327,
+      "step": 10071
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6276016268925092e-05,
+      "loss": 2.0654,
+      "step": 10072
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6273710221591233e-05,
+      "loss": 2.0674,
+      "step": 10073
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6271404160523683e-05,
+      "loss": 2.0642,
+      "step": 10074
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6269098085782907e-05,
+      "loss": 2.0354,
+      "step": 10075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.626679199742939e-05,
+      "loss": 2.0435,
+      "step": 10076
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.62644858955236e-05,
+      "loss": 2.0251,
+      "step": 10077
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6262179780126015e-05,
+      "loss": 2.032,
+      "step": 10078
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.625987365129711e-05,
+      "loss": 2.0872,
+      "step": 10079
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.625756750909736e-05,
+      "loss": 2.0784,
+      "step": 10080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.625526135358724e-05,
+      "loss": 2.0742,
+      "step": 10081
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.625295518482723e-05,
+      "loss": 2.0725,
+      "step": 10082
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6250649002877802e-05,
+      "loss": 2.0898,
+      "step": 10083
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6248342807799438e-05,
+      "loss": 2.0842,
+      "step": 10084
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.624603659965261e-05,
+      "loss": 2.0137,
+      "step": 10085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.624373037849779e-05,
+      "loss": 1.98,
+      "step": 10086
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6241424144395474e-05,
+      "loss": 2.0029,
+      "step": 10087
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6239117897406124e-05,
+      "loss": 2.0344,
+      "step": 10088
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6236811637590228e-05,
+      "loss": 1.9951,
+      "step": 10089
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.623450536500826e-05,
+      "loss": 2.0029,
+      "step": 10090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6232199079720702e-05,
+      "loss": 2.0437,
+      "step": 10091
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6229892781788032e-05,
+      "loss": 2.0515,
+      "step": 10092
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.622758647127073e-05,
+      "loss": 2.009,
+      "step": 10093
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6225280148229276e-05,
+      "loss": 2.0205,
+      "step": 10094
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6222973812724155e-05,
+      "loss": 2.0891,
+      "step": 10095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.622066746481584e-05,
+      "loss": 2.0913,
+      "step": 10096
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6218361104564815e-05,
+      "loss": 2.0535,
+      "step": 10097
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.621605473203157e-05,
+      "loss": 2.0273,
+      "step": 10098
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6213748347276578e-05,
+      "loss": 2.0789,
+      "step": 10099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6211441950360326e-05,
+      "loss": 2.0737,
+      "step": 10100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.620913554134329e-05,
+      "loss": 2.0786,
+      "step": 10101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6206829120285954e-05,
+      "loss": 2.0789,
+      "step": 10102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6204522687248812e-05,
+      "loss": 2.0999,
+      "step": 10103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6202216242292334e-05,
+      "loss": 2.0979,
+      "step": 10104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6199909785477015e-05,
+      "loss": 1.9996,
+      "step": 10105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.619760331686333e-05,
+      "loss": 2.0234,
+      "step": 10106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6195296836511768e-05,
+      "loss": 2.0559,
+      "step": 10107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6192990344482813e-05,
+      "loss": 2.0527,
+      "step": 10108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.619068384083695e-05,
+      "loss": 2.0725,
+      "step": 10109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6188377325634667e-05,
+      "loss": 2.082,
+      "step": 10110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6186070798936447e-05,
+      "loss": 2.0408,
+      "step": 10111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6183764260802773e-05,
+      "loss": 2.0559,
+      "step": 10112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6181457711294135e-05,
+      "loss": 2.0842,
+      "step": 10113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.617915115047102e-05,
+      "loss": 2.0667,
+      "step": 10114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.617684457839391e-05,
+      "loss": 2.1177,
+      "step": 10115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6174537995123302e-05,
+      "loss": 2.0972,
+      "step": 10116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6172231400719675e-05,
+      "loss": 2.1179,
+      "step": 10117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6169924795243518e-05,
+      "loss": 2.0986,
+      "step": 10118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6167618178755323e-05,
+      "loss": 2.0066,
+      "step": 10119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6165311551315577e-05,
+      "loss": 2.073,
+      "step": 10120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6163004912984764e-05,
+      "loss": 2.0698,
+      "step": 10121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6160698263823377e-05,
+      "loss": 2.0789,
+      "step": 10122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6158391603891906e-05,
+      "loss": 2.0979,
+      "step": 10123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6156084933250837e-05,
+      "loss": 2.1094,
+      "step": 10124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6153778251960664e-05,
+      "loss": 2.0308,
+      "step": 10125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6151471560081875e-05,
+      "loss": 2.0684,
+      "step": 10126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.614916485767496e-05,
+      "loss": 2.0144,
+      "step": 10127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6146858144800406e-05,
+      "loss": 1.9963,
+      "step": 10128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.614455142151871e-05,
+      "loss": 2.0588,
+      "step": 10129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6142244687890364e-05,
+      "loss": 2.0652,
+      "step": 10130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.613993794397585e-05,
+      "loss": 2.1079,
+      "step": 10131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6137631189835666e-05,
+      "loss": 2.1011,
+      "step": 10132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6135324425530308e-05,
+      "loss": 2.1252,
+      "step": 10133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.613301765112026e-05,
+      "loss": 2.0938,
+      "step": 10134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6130710866666024e-05,
+      "loss": 2.0581,
+      "step": 10135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6128404072228082e-05,
+      "loss": 2.0957,
+      "step": 10136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6126097267866936e-05,
+      "loss": 2.0293,
+      "step": 10137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6123790453643073e-05,
+      "loss": 2.0601,
+      "step": 10138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6121483629616986e-05,
+      "loss": 2.1008,
+      "step": 10139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.611917679584918e-05,
+      "loss": 2.1172,
+      "step": 10140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.611686995240014e-05,
+      "loss": 2.0632,
+      "step": 10141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6114563099330356e-05,
+      "loss": 2.0503,
+      "step": 10142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6112256236700326e-05,
+      "loss": 2.0994,
+      "step": 10143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.610994936457055e-05,
+      "loss": 2.0896,
+      "step": 10144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.610764248300152e-05,
+      "loss": 2.092,
+      "step": 10145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6105335592053728e-05,
+      "loss": 2.1226,
+      "step": 10146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6103028691787676e-05,
+      "loss": 2.0532,
+      "step": 10147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6100721782263852e-05,
+      "loss": 2.0437,
+      "step": 10148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6098414863542758e-05,
+      "loss": 2.1211,
+      "step": 10149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.609610793568489e-05,
+      "loss": 2.1328,
+      "step": 10150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.609380099875074e-05,
+      "loss": 2.0767,
+      "step": 10151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6091494052800805e-05,
+      "loss": 2.075,
+      "step": 10152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6089187097895595e-05,
+      "loss": 2.0054,
+      "step": 10153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6086880134095587e-05,
+      "loss": 2.0166,
+      "step": 10154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.608457316146129e-05,
+      "loss": 2.0691,
+      "step": 10155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6082266180053208e-05,
+      "loss": 2.0869,
+      "step": 10156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6079959189931824e-05,
+      "loss": 2.1011,
+      "step": 10157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6077652191157642e-05,
+      "loss": 2.1074,
+      "step": 10158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6075345183791162e-05,
+      "loss": 2.0977,
+      "step": 10159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6073038167892885e-05,
+      "loss": 2.0747,
+      "step": 10160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.607073114352331e-05,
+      "loss": 2.0635,
+      "step": 10161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6068424110742923e-05,
+      "loss": 2.0833,
+      "step": 10162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6066117069612236e-05,
+      "loss": 2.0708,
+      "step": 10163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6063810020191752e-05,
+      "loss": 2.1165,
+      "step": 10164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6061502962541957e-05,
+      "loss": 2.0928,
+      "step": 10165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.605919589672336e-05,
+      "loss": 2.0598,
+      "step": 10166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.605688882279646e-05,
+      "loss": 2.0486,
+      "step": 10167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.605458174082176e-05,
+      "loss": 2.0789,
+      "step": 10168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6052274650859752e-05,
+      "loss": 2.0974,
+      "step": 10169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6049967552970945e-05,
+      "loss": 2.1123,
+      "step": 10170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6047660447215838e-05,
+      "loss": 2.0935,
+      "step": 10171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6045353333654926e-05,
+      "loss": 2.0906,
+      "step": 10172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6043046212348718e-05,
+      "loss": 2.0735,
+      "step": 10173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6040739083357716e-05,
+      "loss": 2.094,
+      "step": 10174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6038431946742417e-05,
+      "loss": 1.998,
+      "step": 10175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6036124802563327e-05,
+      "loss": 2.0066,
+      "step": 10176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6033817650880942e-05,
+      "loss": 2.0691,
+      "step": 10177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6031510491755767e-05,
+      "loss": 2.0339,
+      "step": 10178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.602920332524831e-05,
+      "loss": 1.9985,
+      "step": 10179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.602689615141907e-05,
+      "loss": 2.0012,
+      "step": 10180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.602458897032855e-05,
+      "loss": 2.0994,
+      "step": 10181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.602228178203725e-05,
+      "loss": 2.0938,
+      "step": 10182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6019974586605676e-05,
+      "loss": 2.0989,
+      "step": 10183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6017667384094334e-05,
+      "loss": 2.0854,
+      "step": 10184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6015360174563722e-05,
+      "loss": 2.0334,
+      "step": 10185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6013052958074348e-05,
+      "loss": 2.043,
+      "step": 10186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6010745734686714e-05,
+      "loss": 2.0684,
+      "step": 10187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6008438504461332e-05,
+      "loss": 2.0769,
+      "step": 10188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6006131267458694e-05,
+      "loss": 2.0637,
+      "step": 10189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6003824023739314e-05,
+      "loss": 2.0806,
+      "step": 10190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6001516773363696e-05,
+      "loss": 2.0544,
+      "step": 10191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5999209516392334e-05,
+      "loss": 2.0623,
+      "step": 10192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5996902252885747e-05,
+      "loss": 2.0615,
+      "step": 10193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5994594982904436e-05,
+      "loss": 2.0747,
+      "step": 10194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.59922877065089e-05,
+      "loss": 2.0037,
+      "step": 10195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.598998042375966e-05,
+      "loss": 2.0254,
+      "step": 10196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.59876731347172e-05,
+      "loss": 2.0422,
+      "step": 10197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5985365839442047e-05,
+      "loss": 2.0769,
+      "step": 10198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.598305853799469e-05,
+      "loss": 2.0322,
+      "step": 10199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5980751230435646e-05,
+      "loss": 2.064,
+      "step": 10200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.597844391682542e-05,
+      "loss": 2.0513,
+      "step": 10201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5976136597224516e-05,
+      "loss": 2.0327,
+      "step": 10202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5973829271693443e-05,
+      "loss": 2.0889,
+      "step": 10203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.597152194029271e-05,
+      "loss": 2.105,
+      "step": 10204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5969214603082818e-05,
+      "loss": 2.0569,
+      "step": 10205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.596690726012428e-05,
+      "loss": 2.0884,
+      "step": 10206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.59645999114776e-05,
+      "loss": 2.0952,
+      "step": 10207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.596229255720328e-05,
+      "loss": 2.1106,
+      "step": 10208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.595998519736184e-05,
+      "loss": 2.0652,
+      "step": 10209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.595767783201378e-05,
+      "loss": 2.071,
+      "step": 10210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.595537046121961e-05,
+      "loss": 2.0701,
+      "step": 10211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.595306308503984e-05,
+      "loss": 2.0405,
+      "step": 10212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5950755703534975e-05,
+      "loss": 2.0649,
+      "step": 10213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5948448316765524e-05,
+      "loss": 2.0649,
+      "step": 10214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5946140924791994e-05,
+      "loss": 2.0317,
+      "step": 10215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5943833527674897e-05,
+      "loss": 2.0391,
+      "step": 10216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5941526125474745e-05,
+      "loss": 2.1016,
+      "step": 10217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.593921871825204e-05,
+      "loss": 2.1096,
+      "step": 10218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.593691130606729e-05,
+      "loss": 2.1301,
+      "step": 10219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.593460388898101e-05,
+      "loss": 2.1069,
+      "step": 10220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.593229646705371e-05,
+      "loss": 2.0232,
+      "step": 10221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5929989040345895e-05,
+      "loss": 2.0596,
+      "step": 10222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5927681608918072e-05,
+      "loss": 2.0876,
+      "step": 10223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5925374172830758e-05,
+      "loss": 2.0466,
+      "step": 10224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5923066732144458e-05,
+      "loss": 2.0815,
+      "step": 10225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5920759286919688e-05,
+      "loss": 2.0859,
+      "step": 10226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5918451837216953e-05,
+      "loss": 2.021,
+      "step": 10227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5916144383096765e-05,
+      "loss": 2.0503,
+      "step": 10228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.591383692461963e-05,
+      "loss": 2.0559,
+      "step": 10229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5911529461846064e-05,
+      "loss": 2.0559,
+      "step": 10230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.590922199483657e-05,
+      "loss": 2.1067,
+      "step": 10231
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5906914523651672e-05,
+      "loss": 2.0857,
+      "step": 10232
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5904607048351865e-05,
+      "loss": 2.0337,
+      "step": 10233
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.590229956899767e-05,
+      "loss": 2.0164,
+      "step": 10234
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5899992085649598e-05,
+      "loss": 2.0459,
+      "step": 10235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5897684598368155e-05,
+      "loss": 2.0603,
+      "step": 10236
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5895377107213857e-05,
+      "loss": 2.0664,
+      "step": 10237
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5893069612247212e-05,
+      "loss": 2.0588,
+      "step": 10238
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.589076211352873e-05,
+      "loss": 2.0413,
+      "step": 10239
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.588845461111893e-05,
+      "loss": 2.0498,
+      "step": 10240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5886147105078315e-05,
+      "loss": 2.0713,
+      "step": 10241
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5883839595467403e-05,
+      "loss": 2.0623,
+      "step": 10242
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5881532082346698e-05,
+      "loss": 2.0972,
+      "step": 10243
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5879224565776722e-05,
+      "loss": 2.0676,
+      "step": 10244
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5876917045817977e-05,
+      "loss": 2.04,
+      "step": 10245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5874609522530983e-05,
+      "loss": 2.0593,
+      "step": 10246
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5872301995976247e-05,
+      "loss": 2.0615,
+      "step": 10247
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5869994466214285e-05,
+      "loss": 2.0615,
+      "step": 10248
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5867686933305607e-05,
+      "loss": 2.0471,
+      "step": 10249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5865379397310722e-05,
+      "loss": 2.0388,
+      "step": 10250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.586307185829015e-05,
+      "loss": 2.0452,
+      "step": 10251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.58607643163044e-05,
+      "loss": 2.0378,
+      "step": 10252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5858456771413983e-05,
+      "loss": 2.0356,
+      "step": 10253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5856149223679412e-05,
+      "loss": 2.0715,
+      "step": 10254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.58538416731612e-05,
+      "loss": 2.0847,
+      "step": 10255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5851534119919864e-05,
+      "loss": 2.0745,
+      "step": 10256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5849226564015907e-05,
+      "loss": 2.0562,
+      "step": 10257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5846919005509855e-05,
+      "loss": 2.04,
+      "step": 10258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5844611444462213e-05,
+      "loss": 2.0581,
+      "step": 10259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5842303880933493e-05,
+      "loss": 2.0718,
+      "step": 10260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.583999631498421e-05,
+      "loss": 2.0129,
+      "step": 10261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5837688746674888e-05,
+      "loss": 2.0391,
+      "step": 10262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5835381176066022e-05,
+      "loss": 2.0481,
+      "step": 10263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5833073603218134e-05,
+      "loss": 2.0322,
+      "step": 10264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5830766028191738e-05,
+      "loss": 2.0266,
+      "step": 10265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5828458451047346e-05,
+      "loss": 2.043,
+      "step": 10266
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5826150871845477e-05,
+      "loss": 2.0737,
+      "step": 10267
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5823843290646637e-05,
+      "loss": 2.0618,
+      "step": 10268
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5821535707511337e-05,
+      "loss": 2.042,
+      "step": 10269
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5819228122500103e-05,
+      "loss": 2.0715,
+      "step": 10270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5816920535673437e-05,
+      "loss": 2.0857,
+      "step": 10271
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.581461294709186e-05,
+      "loss": 2.0784,
+      "step": 10272
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5812305356815883e-05,
+      "loss": 2.1016,
+      "step": 10273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5809997764906016e-05,
+      "loss": 2.092,
+      "step": 10274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5807690171422782e-05,
+      "loss": 2.103,
+      "step": 10275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.580538257642669e-05,
+      "loss": 2.0789,
+      "step": 10276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5803074979978255e-05,
+      "loss": 2.0806,
+      "step": 10277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5800767382137992e-05,
+      "loss": 2.0999,
+      "step": 10278
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.579845978296641e-05,
+      "loss": 2.0762,
+      "step": 10279
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5796152182524025e-05,
+      "loss": 2.0784,
+      "step": 10280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5793844580871354e-05,
+      "loss": 2.0557,
+      "step": 10281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.579153697806891e-05,
+      "loss": 2.0437,
+      "step": 10282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5789229374177208e-05,
+      "loss": 2.0396,
+      "step": 10283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5786921769256763e-05,
+      "loss": 2.073,
+      "step": 10284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5784614163368084e-05,
+      "loss": 2.0044,
+      "step": 10285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5782306556571688e-05,
+      "loss": 1.9993,
+      "step": 10286
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5779998948928093e-05,
+      "loss": 2.001,
+      "step": 10287
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.577769134049781e-05,
+      "loss": 2.0115,
+      "step": 10288
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.577538373134135e-05,
+      "loss": 1.9961,
+      "step": 10289
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5773076121519234e-05,
+      "loss": 2.0632,
+      "step": 10290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5770768511091973e-05,
+      "loss": 2.0793,
+      "step": 10291
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.576846090012008e-05,
+      "loss": 2.0791,
+      "step": 10292
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5766153288664075e-05,
+      "loss": 2.0449,
+      "step": 10293
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5763845676784468e-05,
+      "loss": 2.0598,
+      "step": 10294
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.576153806454177e-05,
+      "loss": 2.0789,
+      "step": 10295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.57592304519965e-05,
+      "loss": 2.0742,
+      "step": 10296
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5756922839209177e-05,
+      "loss": 2.0576,
+      "step": 10297
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5754615226240312e-05,
+      "loss": 2.0349,
+      "step": 10298
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5752307613150414e-05,
+      "loss": 2.0537,
+      "step": 10299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.575e-05,
+      "loss": 2.072,
+      "step": 10300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5747692386849593e-05,
+      "loss": 2.0427,
+      "step": 10301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.574538477375969e-05,
+      "loss": 2.0535,
+      "step": 10302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5743077160790827e-05,
+      "loss": 2.084,
+      "step": 10303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5740769548003495e-05,
+      "loss": 2.0884,
+      "step": 10304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5738461935458235e-05,
+      "loss": 2.0852,
+      "step": 10305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.573615432321554e-05,
+      "loss": 2.1021,
+      "step": 10306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5733846711335932e-05,
+      "loss": 2.0088,
+      "step": 10307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5731539099879923e-05,
+      "loss": 2.0244,
+      "step": 10308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.572923148890803e-05,
+      "loss": 2.0618,
+      "step": 10309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.572692387848077e-05,
+      "loss": 2.062,
+      "step": 10310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5724616268658653e-05,
+      "loss": 2.106,
+      "step": 10311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5722308659502196e-05,
+      "loss": 2.105,
+      "step": 10312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5720001051071914e-05,
+      "loss": 2.0562,
+      "step": 10313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5717693443428315e-05,
+      "loss": 2.063,
+      "step": 10314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.571538583663192e-05,
+      "loss": 2.1106,
+      "step": 10315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5713078230743244e-05,
+      "loss": 2.0864,
+      "step": 10316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5710770625822792e-05,
+      "loss": 2.0732,
+      "step": 10317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.570846302193109e-05,
+      "loss": 2.0725,
+      "step": 10318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5706155419128645e-05,
+      "loss": 2.105,
+      "step": 10319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5703847817475975e-05,
+      "loss": 2.0969,
+      "step": 10320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5701540217033597e-05,
+      "loss": 2.0603,
+      "step": 10321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.569923261786201e-05,
+      "loss": 2.0518,
+      "step": 10322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5696925020021752e-05,
+      "loss": 2.1094,
+      "step": 10323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.569461742357331e-05,
+      "loss": 2.0781,
+      "step": 10324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5692309828577218e-05,
+      "loss": 2.0825,
+      "step": 10325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5690002235093984e-05,
+      "loss": 2.1008,
+      "step": 10326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.568769464318412e-05,
+      "loss": 2.0132,
+      "step": 10327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5685387052908146e-05,
+      "loss": 2.0042,
+      "step": 10328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5683079464326567e-05,
+      "loss": 2.0745,
+      "step": 10329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.56807718774999e-05,
+      "loss": 2.0852,
+      "step": 10330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5678464292488666e-05,
+      "loss": 2.0146,
+      "step": 10331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.567615670935337e-05,
+      "loss": 2.0291,
+      "step": 10332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.567384912815453e-05,
+      "loss": 2.0562,
+      "step": 10333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5671541548952654e-05,
+      "loss": 2.0996,
+      "step": 10334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5669233971808265e-05,
+      "loss": 2.1013,
+      "step": 10335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5666926396781873e-05,
+      "loss": 2.1094,
+      "step": 10336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5664618823933984e-05,
+      "loss": 2.0825,
+      "step": 10337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5662311253325123e-05,
+      "loss": 2.0881,
+      "step": 10338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5660003685015786e-05,
+      "loss": 2.052,
+      "step": 10339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.565769611906651e-05,
+      "loss": 2.0957,
+      "step": 10340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5655388555537793e-05,
+      "loss": 2.0896,
+      "step": 10341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.565308099449015e-05,
+      "loss": 2.103,
+      "step": 10342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5650773435984093e-05,
+      "loss": 2.0696,
+      "step": 10343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5648465880080143e-05,
+      "loss": 2.0901,
+      "step": 10344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.56461583268388e-05,
+      "loss": 2.0432,
+      "step": 10345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5643850776320595e-05,
+      "loss": 2.063,
+      "step": 10346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.564154322858602e-05,
+      "loss": 2.0171,
+      "step": 10347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5639235683695606e-05,
+      "loss": 2.0652,
+      "step": 10348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.563692814170985e-05,
+      "loss": 1.9958,
+      "step": 10349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.563462060268928e-05,
+      "loss": 2.0425,
+      "step": 10350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5632313066694404e-05,
+      "loss": 2.0857,
+      "step": 10351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.563000553378572e-05,
+      "loss": 2.0762,
+      "step": 10352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.562769800402376e-05,
+      "loss": 2.0674,
+      "step": 10353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.562539047746902e-05,
+      "loss": 2.0715,
+      "step": 10354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.562308295418203e-05,
+      "loss": 2.0908,
+      "step": 10355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5620775434223288e-05,
+      "loss": 2.0747,
+      "step": 10356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5618467917653302e-05,
+      "loss": 2.0486,
+      "step": 10357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5616160404532607e-05,
+      "loss": 2.0798,
+      "step": 10358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5613852894921688e-05,
+      "loss": 2.0317,
+      "step": 10359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5611545388881077e-05,
+      "loss": 2.0452,
+      "step": 10360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.560923788647127e-05,
+      "loss": 2.084,
+      "step": 10361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.560693038775279e-05,
+      "loss": 2.0815,
+      "step": 10362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5604622892786147e-05,
+      "loss": 2.0857,
+      "step": 10363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5602315401631845e-05,
+      "loss": 2.1204,
+      "step": 10364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5600007914350405e-05,
+      "loss": 2.0635,
+      "step": 10365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5597700431002325e-05,
+      "loss": 2.0771,
+      "step": 10366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5595392951648135e-05,
+      "loss": 2.0386,
+      "step": 10367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5593085476348335e-05,
+      "loss": 2.0791,
+      "step": 10368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5590778005163428e-05,
+      "loss": 2.0786,
+      "step": 10369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5588470538153943e-05,
+      "loss": 2.0901,
+      "step": 10370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.558616307538037e-05,
+      "loss": 2.0332,
+      "step": 10371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5583855616903242e-05,
+      "loss": 2.0466,
+      "step": 10372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.558154816278305e-05,
+      "loss": 2.0984,
+      "step": 10373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5579240713080312e-05,
+      "loss": 2.0933,
+      "step": 10374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5576933267855542e-05,
+      "loss": 2.0813,
+      "step": 10375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5574625827169242e-05,
+      "loss": 2.1084,
+      "step": 10376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.557231839108193e-05,
+      "loss": 2.0745,
+      "step": 10377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.557001095965411e-05,
+      "loss": 2.082,
+      "step": 10378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5567703532946295e-05,
+      "loss": 2.05,
+      "step": 10379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5565396111018993e-05,
+      "loss": 2.0408,
+      "step": 10380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.556308869393271e-05,
+      "loss": 2.0317,
+      "step": 10381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5560781281747966e-05,
+      "loss": 2.0447,
+      "step": 10382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.555847387452526e-05,
+      "loss": 2.0466,
+      "step": 10383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5556166472325103e-05,
+      "loss": 2.0354,
+      "step": 10384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.555385907520801e-05,
+      "loss": 2.0796,
+      "step": 10385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.555155168323448e-05,
+      "loss": 2.1025,
+      "step": 10386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5549244296465032e-05,
+      "loss": 2.125,
+      "step": 10387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5546936914960165e-05,
+      "loss": 2.1001,
+      "step": 10388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5544629538780388e-05,
+      "loss": 2.1006,
+      "step": 10389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5542322167986222e-05,
+      "loss": 2.0842,
+      "step": 10390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.554001480263816e-05,
+      "loss": 2.0579,
+      "step": 10391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.553770744279672e-05,
+      "loss": 2.0691,
+      "step": 10392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.553540008852241e-05,
+      "loss": 2.0598,
+      "step": 10393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5533092739875724e-05,
+      "loss": 2.1001,
+      "step": 10394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.553078539691719e-05,
+      "loss": 2.1318,
+      "step": 10395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5528478059707293e-05,
+      "loss": 2.1248,
+      "step": 10396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5526170728306557e-05,
+      "loss": 2.0137,
+      "step": 10397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5523863402775484e-05,
+      "loss": 2.0374,
+      "step": 10398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.552155608317458e-05,
+      "loss": 2.0847,
+      "step": 10399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5519248769564354e-05,
+      "loss": 2.0764,
+      "step": 10400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5516941462005308e-05,
+      "loss": 2.0815,
+      "step": 10401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.551463416055796e-05,
+      "loss": 2.1086,
+      "step": 10402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5512326865282806e-05,
+      "loss": 2.094,
+      "step": 10403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5510019576240347e-05,
+      "loss": 2.1121,
+      "step": 10404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5507712293491102e-05,
+      "loss": 2.0283,
+      "step": 10405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5505405017095567e-05,
+      "loss": 2.0715,
+      "step": 10406
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5503097747114257e-05,
+      "loss": 2.0759,
+      "step": 10407
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.550079048360767e-05,
+      "loss": 2.0962,
+      "step": 10408
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.549848322663631e-05,
+      "loss": 2.0505,
+      "step": 10409
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5496175976260692e-05,
+      "loss": 2.0544,
+      "step": 10410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5493868732541306e-05,
+      "loss": 2.0305,
+      "step": 10411
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5491561495538675e-05,
+      "loss": 2.0359,
+      "step": 10412
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5489254265313286e-05,
+      "loss": 2.0801,
+      "step": 10413
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5486947041925655e-05,
+      "loss": 2.0867,
+      "step": 10414
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5484639825436285e-05,
+      "loss": 2.0647,
+      "step": 10415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5482332615905672e-05,
+      "loss": 2.0852,
+      "step": 10416
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.548002541339433e-05,
+      "loss": 2.0452,
+      "step": 10417
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.547771821796276e-05,
+      "loss": 2.0669,
+      "step": 10418
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5475411029671457e-05,
+      "loss": 2.0598,
+      "step": 10419
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5473103848580935e-05,
+      "loss": 2.0391,
+      "step": 10420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.547079667475169e-05,
+      "loss": 2.0603,
+      "step": 10421
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5468489508244233e-05,
+      "loss": 2.0559,
+      "step": 10422
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5466182349119068e-05,
+      "loss": 2.0759,
+      "step": 10423
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.546387519743668e-05,
+      "loss": 2.0881,
+      "step": 10424
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.546156805325759e-05,
+      "loss": 2.0762,
+      "step": 10425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5459260916642284e-05,
+      "loss": 2.1221,
+      "step": 10426
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.545695378765128e-05,
+      "loss": 2.0862,
+      "step": 10427
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.545464666634508e-05,
+      "loss": 2.0674,
+      "step": 10428
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.545233955278417e-05,
+      "loss": 2.01,
+      "step": 10429
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.545003244702906e-05,
+      "loss": 2.0298,
+      "step": 10430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5447725349140248e-05,
+      "loss": 2.0625,
+      "step": 10431
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5445418259178247e-05,
+      "loss": 2.0696,
+      "step": 10432
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5443111177203543e-05,
+      "loss": 2.0718,
+      "step": 10433
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.544080410327664e-05,
+      "loss": 2.0886,
+      "step": 10434
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5438497037458046e-05,
+      "loss": 2.0066,
+      "step": 10435
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.543618997980825e-05,
+      "loss": 2.0251,
+      "step": 10436
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5433882930387764e-05,
+      "loss": 1.9993,
+      "step": 10437
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.543157588925708e-05,
+      "loss": 2.0017,
+      "step": 10438
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5429268856476698e-05,
+      "loss": 2.0999,
+      "step": 10439
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5426961832107122e-05,
+      "loss": 2.0793,
+      "step": 10440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5424654816208838e-05,
+      "loss": 2.073,
+      "step": 10441
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.542234780884236e-05,
+      "loss": 2.0444,
+      "step": 10442
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5420040810068186e-05,
+      "loss": 2.0269,
+      "step": 10443
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.54177338199468e-05,
+      "loss": 2.0161,
+      "step": 10444
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.541542683853871e-05,
+      "loss": 2.1218,
+      "step": 10445
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5413119865904413e-05,
+      "loss": 2.1152,
+      "step": 10446
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.541081290210441e-05,
+      "loss": 2.0835,
+      "step": 10447
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5408505947199195e-05,
+      "loss": 2.1008,
+      "step": 10448
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5406199001249264e-05,
+      "loss": 2.0486,
+      "step": 10449
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5403892064315115e-05,
+      "loss": 2.0454,
+      "step": 10450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5401585136457246e-05,
+      "loss": 2.0632,
+      "step": 10451
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.539927821773615e-05,
+      "loss": 2.0603,
+      "step": 10452
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.539697130821233e-05,
+      "loss": 2.0522,
+      "step": 10453
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5394664407946275e-05,
+      "loss": 2.0642,
+      "step": 10454
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5392357516998487e-05,
+      "loss": 2.0823,
+      "step": 10455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5390050635429453e-05,
+      "loss": 2.0664,
+      "step": 10456
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5387743763299677e-05,
+      "loss": 2.0891,
+      "step": 10457
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5385436900669654e-05,
+      "loss": 2.0967,
+      "step": 10458
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5383130047599867e-05,
+      "loss": 2.0879,
+      "step": 10459
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5380823204150824e-05,
+      "loss": 2.0935,
+      "step": 10460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.537851637038301e-05,
+      "loss": 2.0461,
+      "step": 10461
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.537620954635693e-05,
+      "loss": 2.0605,
+      "step": 10462
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.537390273213307e-05,
+      "loss": 2.0264,
+      "step": 10463
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5371595927771918e-05,
+      "loss": 2.0569,
+      "step": 10464
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5369289133333982e-05,
+      "loss": 2.1069,
+      "step": 10465
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5366982348879736e-05,
+      "loss": 2.0977,
+      "step": 10466
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5364675574469695e-05,
+      "loss": 2.0237,
+      "step": 10467
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5362368810164337e-05,
+      "loss": 2.0137,
+      "step": 10468
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5360062056024153e-05,
+      "loss": 2.0742,
+      "step": 10469
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5357755312109643e-05,
+      "loss": 2.0923,
+      "step": 10470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5355448578481287e-05,
+      "loss": 2.0791,
+      "step": 10471
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5353141855199594e-05,
+      "loss": 2.083,
+      "step": 10472
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5350835142325047e-05,
+      "loss": 2.0286,
+      "step": 10473
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.534852843991813e-05,
+      "loss": 2.0735,
+      "step": 10474
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.534622174803934e-05,
+      "loss": 2.0708,
+      "step": 10475
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5343915066749163e-05,
+      "loss": 2.1052,
+      "step": 10476
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5341608396108094e-05,
+      "loss": 2.0466,
+      "step": 10477
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5339301736176626e-05,
+      "loss": 2.063,
+      "step": 10478
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5336995087015236e-05,
+      "loss": 2.0828,
+      "step": 10479
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5334688448684426e-05,
+      "loss": 2.0864,
+      "step": 10480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5332381821244677e-05,
+      "loss": 2.0176,
+      "step": 10481
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5330075204756482e-05,
+      "loss": 2.0254,
+      "step": 10482
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5327768599280325e-05,
+      "loss": 2.0674,
+      "step": 10483
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.53254620048767e-05,
+      "loss": 2.0964,
+      "step": 10484
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.532315542160609e-05,
+      "loss": 2.0691,
+      "step": 10485
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5320848849528986e-05,
+      "loss": 2.0813,
+      "step": 10486
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5318542288705868e-05,
+      "loss": 2.0483,
+      "step": 10487
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5316235739197233e-05,
+      "loss": 2.0542,
+      "step": 10488
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.531392920106356e-05,
+      "loss": 2.0427,
+      "step": 10489
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.531162267436534e-05,
+      "loss": 2.043,
+      "step": 10490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.530931615916305e-05,
+      "loss": 2.0115,
+      "step": 10491
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5307009655517187e-05,
+      "loss": 2.0544,
+      "step": 10492
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5304703163488232e-05,
+      "loss": 2.0564,
+      "step": 10493
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5302396683136673e-05,
+      "loss": 2.1003,
+      "step": 10494
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5300090214522992e-05,
+      "loss": 2.0479,
+      "step": 10495
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5297783757707666e-05,
+      "loss": 2.041,
+      "step": 10496
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5295477312751195e-05,
+      "loss": 2.0471,
+      "step": 10497
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5293170879714042e-05,
+      "loss": 2.0781,
+      "step": 10498
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5290864458656717e-05,
+      "loss": 2.1033,
+      "step": 10499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5288558049639685e-05,
+      "loss": 2.104,
+      "step": 10500
+    },
+    {
+      "Bleu_1": 0.7728853407796714,
+      "Bleu_2": 0.6130135528824922,
+      "Bleu_3": 0.4711986317899507,
+      "Bleu_4": 0.35837992730486906,
+      "CIDEr": 1.2013524463914542,
+      "METEOR": 0.2784851639538413,
+      "ROUGE_L": 0.5588288250165528,
+      "epoch": 0.53,
+      "step": 10500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5286251652723425e-05,
+      "loss": 2.0447,
+      "step": 10501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5283945267968435e-05,
+      "loss": 2.0554,
+      "step": 10502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.528163889543518e-05,
+      "loss": 2.0259,
+      "step": 10503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5279332535184167e-05,
+      "loss": 2.0386,
+      "step": 10504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.527702618727585e-05,
+      "loss": 2.0476,
+      "step": 10505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5274719851770724e-05,
+      "loss": 2.0708,
+      "step": 10506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5272413528729274e-05,
+      "loss": 2.1072,
+      "step": 10507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.527010721821197e-05,
+      "loss": 2.0796,
+      "step": 10508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.52678009202793e-05,
+      "loss": 2.0522,
+      "step": 10509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5265494634991742e-05,
+      "loss": 2.074,
+      "step": 10510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5263188362409776e-05,
+      "loss": 2.0547,
+      "step": 10511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.526088210259388e-05,
+      "loss": 2.0674,
+      "step": 10512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5258575855604526e-05,
+      "loss": 2.0952,
+      "step": 10513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.525626962150221e-05,
+      "loss": 2.1125,
+      "step": 10514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.52539634003474e-05,
+      "loss": 2.0615,
+      "step": 10515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5251657192200565e-05,
+      "loss": 2.0483,
+      "step": 10516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5249350997122203e-05,
+      "loss": 2.0208,
+      "step": 10517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5247044815172772e-05,
+      "loss": 2.0503,
+      "step": 10518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.524473864641276e-05,
+      "loss": 2.074,
+      "step": 10519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5242432490902645e-05,
+      "loss": 2.0469,
+      "step": 10520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5240126348702894e-05,
+      "loss": 1.9922,
+      "step": 10521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5237820219873988e-05,
+      "loss": 2.0386,
+      "step": 10522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5235514104476405e-05,
+      "loss": 2.0898,
+      "step": 10523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5233208002570615e-05,
+      "loss": 2.1143,
+      "step": 10524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.52309019142171e-05,
+      "loss": 2.064,
+      "step": 10525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5228595839476324e-05,
+      "loss": 2.0432,
+      "step": 10526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5226289778408774e-05,
+      "loss": 2.0205,
+      "step": 10527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5223983731074908e-05,
+      "loss": 2.0288,
+      "step": 10528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5221677697535219e-05,
+      "loss": 2.0566,
+      "step": 10529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5219371677850167e-05,
+      "loss": 2.0327,
+      "step": 10530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5217065672080221e-05,
+      "loss": 2.0588,
+      "step": 10531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5214759680285865e-05,
+      "loss": 2.0098,
+      "step": 10532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5212453702527561e-05,
+      "loss": 1.9905,
+      "step": 10533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5210147738865788e-05,
+      "loss": 1.9934,
+      "step": 10534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5207841789361017e-05,
+      "loss": 2.0808,
+      "step": 10535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5205535854073712e-05,
+      "loss": 2.0649,
+      "step": 10536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.520322993306435e-05,
+      "loss": 2.0923,
+      "step": 10537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5200924026393392e-05,
+      "loss": 2.0696,
+      "step": 10538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5198618134121324e-05,
+      "loss": 2.0991,
+      "step": 10539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5196312256308599e-05,
+      "loss": 2.1155,
+      "step": 10540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5194006393015693e-05,
+      "loss": 2.0732,
+      "step": 10541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.519170054430308e-05,
+      "loss": 2.0881,
+      "step": 10542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5189394710231213e-05,
+      "loss": 2.0437,
+      "step": 10543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.518708889086058e-05,
+      "loss": 2.0408,
+      "step": 10544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5184783086251633e-05,
+      "loss": 2.04,
+      "step": 10545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5182477296464841e-05,
+      "loss": 2.0417,
+      "step": 10546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5180171521560681e-05,
+      "loss": 2.0854,
+      "step": 10547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5177865761599602e-05,
+      "loss": 2.1174,
+      "step": 10548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.517556001664209e-05,
+      "loss": 2.0688,
+      "step": 10549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5173254286748596e-05,
+      "loss": 2.0669,
+      "step": 10550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5170948571979586e-05,
+      "loss": 2.1211,
+      "step": 10551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5168642872395533e-05,
+      "loss": 2.1157,
+      "step": 10552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5166337188056888e-05,
+      "loss": 2.0842,
+      "step": 10553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5164031519024133e-05,
+      "loss": 2.1023,
+      "step": 10554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5161725865357717e-05,
+      "loss": 2.0474,
+      "step": 10555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5159420227118106e-05,
+      "loss": 2.0488,
+      "step": 10556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5157114604365768e-05,
+      "loss": 2.043,
+      "step": 10557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5154808997161158e-05,
+      "loss": 2.0876,
+      "step": 10558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5152503405564745e-05,
+      "loss": 2.0762,
+      "step": 10559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5150197829636986e-05,
+      "loss": 2.0461,
+      "step": 10560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5147892269438339e-05,
+      "loss": 2.0298,
+      "step": 10561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5145586725029275e-05,
+      "loss": 2.0415,
+      "step": 10562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5143281196470243e-05,
+      "loss": 2.083,
+      "step": 10563
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5140975683821714e-05,
+      "loss": 2.071,
+      "step": 10564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.513867018714414e-05,
+      "loss": 2.0625,
+      "step": 10565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5136364706497973e-05,
+      "loss": 2.0605,
+      "step": 10566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.513405924194369e-05,
+      "loss": 2.0715,
+      "step": 10567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5131753793541729e-05,
+      "loss": 2.0786,
+      "step": 10568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5129448361352566e-05,
+      "loss": 2.0618,
+      "step": 10569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5127142945436647e-05,
+      "loss": 2.0659,
+      "step": 10570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512483754585443e-05,
+      "loss": 2.0767,
+      "step": 10571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5122532162666378e-05,
+      "loss": 2.0779,
+      "step": 10572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5120226795932931e-05,
+      "loss": 2.0447,
+      "step": 10573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5117921445714565e-05,
+      "loss": 2.0427,
+      "step": 10574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5115616112071726e-05,
+      "loss": 2.0894,
+      "step": 10575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5113310795064862e-05,
+      "loss": 2.0833,
+      "step": 10576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.511100549475444e-05,
+      "loss": 2.0762,
+      "step": 10577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5108700211200903e-05,
+      "loss": 2.084,
+      "step": 10578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5106394944464715e-05,
+      "loss": 2.0686,
+      "step": 10579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.510408969460632e-05,
+      "loss": 2.0947,
+      "step": 10580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5101784461686168e-05,
+      "loss": 2.0779,
+      "step": 10581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5099479245764725e-05,
+      "loss": 2.071,
+      "step": 10582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5097174046902427e-05,
+      "loss": 2.0381,
+      "step": 10583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.509486886515974e-05,
+      "loss": 2.0276,
+      "step": 10584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5092563700597104e-05,
+      "loss": 1.9998,
+      "step": 10585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5090258553274967e-05,
+      "loss": 1.9724,
+      "step": 10586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.508795342325379e-05,
+      "loss": 2.0715,
+      "step": 10587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5085648310594013e-05,
+      "loss": 2.0386,
+      "step": 10588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.508334321535609e-05,
+      "loss": 2.0793,
+      "step": 10589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5081038137600469e-05,
+      "loss": 2.0679,
+      "step": 10590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5078733077387595e-05,
+      "loss": 2.051,
+      "step": 10591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5076428034777918e-05,
+      "loss": 2.0457,
+      "step": 10592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5074123009831882e-05,
+      "loss": 2.084,
+      "step": 10593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5071818002609937e-05,
+      "loss": 2.0613,
+      "step": 10594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5069513013172532e-05,
+      "loss": 2.0198,
+      "step": 10595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5067208041580103e-05,
+      "loss": 1.9907,
+      "step": 10596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5064903087893108e-05,
+      "loss": 2.0938,
+      "step": 10597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5062598152171976e-05,
+      "loss": 2.0757,
+      "step": 10598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5060293234477168e-05,
+      "loss": 2.0886,
+      "step": 10599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.505798833486912e-05,
+      "loss": 2.0911,
+      "step": 10600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5055683453408265e-05,
+      "loss": 2.0259,
+      "step": 10601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5053378590155067e-05,
+      "loss": 2.0552,
+      "step": 10602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.505107374516995e-05,
+      "loss": 2.0166,
+      "step": 10603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5048768918513366e-05,
+      "loss": 2.0071,
+      "step": 10604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.504646411024575e-05,
+      "loss": 2.0862,
+      "step": 10605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5044159320427553e-05,
+      "loss": 2.0967,
+      "step": 10606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.504185454911921e-05,
+      "loss": 2.0559,
+      "step": 10607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.503954979638115e-05,
+      "loss": 2.0854,
+      "step": 10608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.503724506227383e-05,
+      "loss": 2.0842,
+      "step": 10609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5034940346857678e-05,
+      "loss": 2.0586,
+      "step": 10610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5032635650193139e-05,
+      "loss": 2.02,
+      "step": 10611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5030330972340648e-05,
+      "loss": 2.0234,
+      "step": 10612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5028026313360637e-05,
+      "loss": 2.0566,
+      "step": 10613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5025721673313552e-05,
+      "loss": 2.0681,
+      "step": 10614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5023417052259824e-05,
+      "loss": 2.1121,
+      "step": 10615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5021112450259894e-05,
+      "loss": 2.1096,
+      "step": 10616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.50188078673742e-05,
+      "loss": 2.1123,
+      "step": 10617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5016503303663159e-05,
+      "loss": 2.1113,
+      "step": 10618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5014198759187227e-05,
+      "loss": 2.1316,
+      "step": 10619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5011894234006822e-05,
+      "loss": 2.1509,
+      "step": 10620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.500958972818239e-05,
+      "loss": 2.0505,
+      "step": 10621
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.500728524177436e-05,
+      "loss": 1.9924,
+      "step": 10622
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5004980774843153e-05,
+      "loss": 2.0771,
+      "step": 10623
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5002676327449221e-05,
+      "loss": 2.082,
+      "step": 10624
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5000371899652977e-05,
+      "loss": 1.9973,
+      "step": 10625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4998067491514868e-05,
+      "loss": 1.981,
+      "step": 10626
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4995763103095314e-05,
+      "loss": 2.0491,
+      "step": 10627
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4993458734454747e-05,
+      "loss": 2.0833,
+      "step": 10628
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4991154385653597e-05,
+      "loss": 2.0618,
+      "step": 10629
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4988850056752291e-05,
+      "loss": 2.0784,
+      "step": 10630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4986545747811262e-05,
+      "loss": 2.053,
+      "step": 10631
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4984241458890938e-05,
+      "loss": 2.053,
+      "step": 10632
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4981937190051734e-05,
+      "loss": 2.0481,
+      "step": 10633
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4979632941354095e-05,
+      "loss": 2.0701,
+      "step": 10634
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4977328712858432e-05,
+      "loss": 2.0825,
+      "step": 10635
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4975024504625183e-05,
+      "loss": 2.0444,
+      "step": 10636
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4972720316714766e-05,
+      "loss": 2.0618,
+      "step": 10637
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4970416149187603e-05,
+      "loss": 2.0488,
+      "step": 10638
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4968112002104127e-05,
+      "loss": 2.0312,
+      "step": 10639
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4965807875524751e-05,
+      "loss": 2.0615,
+      "step": 10640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.496350376950991e-05,
+      "loss": 2.1248,
+      "step": 10641
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4961199684120021e-05,
+      "loss": 2.1255,
+      "step": 10642
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4958895619415497e-05,
+      "loss": 2.0146,
+      "step": 10643
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4956591575456772e-05,
+      "loss": 2.0317,
+      "step": 10644
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4954287552304262e-05,
+      "loss": 2.0879,
+      "step": 10645
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4951983550018387e-05,
+      "loss": 2.0879,
+      "step": 10646
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4949679568659573e-05,
+      "loss": 2.0881,
+      "step": 10647
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4947375608288226e-05,
+      "loss": 2.1011,
+      "step": 10648
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.494507166896478e-05,
+      "loss": 2.0837,
+      "step": 10649
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4942767750749636e-05,
+      "loss": 2.0986,
+      "step": 10650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4940463853703231e-05,
+      "loss": 2.0828,
+      "step": 10651
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4938159977885971e-05,
+      "loss": 2.0603,
+      "step": 10652
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4935856123358268e-05,
+      "loss": 2.021,
+      "step": 10653
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.493355229018055e-05,
+      "loss": 2.0286,
+      "step": 10654
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.493124847841322e-05,
+      "loss": 2.0659,
+      "step": 10655
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4928944688116704e-05,
+      "loss": 2.0708,
+      "step": 10656
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4926640919351412e-05,
+      "loss": 2.0996,
+      "step": 10657
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.492433717217775e-05,
+      "loss": 2.1091,
+      "step": 10658
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4922033446656144e-05,
+      "loss": 2.0779,
+      "step": 10659
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4919729742846995e-05,
+      "loss": 2.0881,
+      "step": 10660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4917426060810725e-05,
+      "loss": 2.0215,
+      "step": 10661
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4915122400607738e-05,
+      "loss": 2.0208,
+      "step": 10662
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4912818762298448e-05,
+      "loss": 2.0398,
+      "step": 10663
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4910515145943264e-05,
+      "loss": 2.0518,
+      "step": 10664
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4908211551602596e-05,
+      "loss": 2.114,
+      "step": 10665
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4905907979336854e-05,
+      "loss": 2.1011,
+      "step": 10666
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.490360442920645e-05,
+      "loss": 2.061,
+      "step": 10667
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4901300901271781e-05,
+      "loss": 2.0825,
+      "step": 10668
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4898997395593268e-05,
+      "loss": 2.042,
+      "step": 10669
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4896693912231304e-05,
+      "loss": 2.0652,
+      "step": 10670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4894390451246308e-05,
+      "loss": 2.0945,
+      "step": 10671
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4892087012698681e-05,
+      "loss": 2.0391,
+      "step": 10672
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.488978359664882e-05,
+      "loss": 2.0591,
+      "step": 10673
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4887480203157144e-05,
+      "loss": 2.0293,
+      "step": 10674
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4885176832284042e-05,
+      "loss": 2.0764,
+      "step": 10675
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.488287348408993e-05,
+      "loss": 2.0854,
+      "step": 10676
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4880570158635204e-05,
+      "loss": 2.0557,
+      "step": 10677
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4878266855980263e-05,
+      "loss": 2.0725,
+      "step": 10678
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4875963576185512e-05,
+      "loss": 2.0815,
+      "step": 10679
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4873660319311355e-05,
+      "loss": 2.061,
+      "step": 10680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4871357085418191e-05,
+      "loss": 2.0574,
+      "step": 10681
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4869053874566418e-05,
+      "loss": 2.0564,
+      "step": 10682
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4866750686816431e-05,
+      "loss": 2.0134,
+      "step": 10683
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4864447522228638e-05,
+      "loss": 2.0332,
+      "step": 10684
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4862144380863423e-05,
+      "loss": 2.0444,
+      "step": 10685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4859841262781201e-05,
+      "loss": 1.9951,
+      "step": 10686
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4857538168042358e-05,
+      "loss": 2.0376,
+      "step": 10687
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4855235096707285e-05,
+      "loss": 2.0244,
+      "step": 10688
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.485293204883639e-05,
+      "loss": 2.0535,
+      "step": 10689
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4850629024490056e-05,
+      "loss": 2.0601,
+      "step": 10690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4848326023728685e-05,
+      "loss": 2.0618,
+      "step": 10691
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4846023046612673e-05,
+      "loss": 2.0911,
+      "step": 10692
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4843720093202397e-05,
+      "loss": 2.0615,
+      "step": 10693
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.484141716355827e-05,
+      "loss": 2.104,
+      "step": 10694
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4839114257740667e-05,
+      "loss": 2.054,
+      "step": 10695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.483681137580999e-05,
+      "loss": 2.0664,
+      "step": 10696
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4834508517826625e-05,
+      "loss": 2.051,
+      "step": 10697
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.483220568385096e-05,
+      "loss": 2.0515,
+      "step": 10698
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.4829902873943386e-05,
+      "loss": 2.0479,
+      "step": 10699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4827600088164292e-05,
+      "loss": 2.0693,
+      "step": 10700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4825297326574066e-05,
+      "loss": 2.0608,
+      "step": 10701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4822994589233098e-05,
+      "loss": 2.0549,
+      "step": 10702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4820691876201766e-05,
+      "loss": 2.0071,
+      "step": 10703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4818389187540467e-05,
+      "loss": 2.0403,
+      "step": 10704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4816086523309574e-05,
+      "loss": 2.0737,
+      "step": 10705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4813783883569485e-05,
+      "loss": 2.0283,
+      "step": 10706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4811481268380577e-05,
+      "loss": 2.0559,
+      "step": 10707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4809178677803229e-05,
+      "loss": 2.0308,
+      "step": 10708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4806876111897835e-05,
+      "loss": 2.0867,
+      "step": 10709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4804573570724764e-05,
+      "loss": 2.0588,
+      "step": 10710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.480227105434441e-05,
+      "loss": 2.0752,
+      "step": 10711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4799968562817149e-05,
+      "loss": 2.0391,
+      "step": 10712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4797666096203354e-05,
+      "loss": 2.0476,
+      "step": 10713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4795363654563412e-05,
+      "loss": 2.0591,
+      "step": 10714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4793061237957703e-05,
+      "loss": 2.0566,
+      "step": 10715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4790758846446601e-05,
+      "loss": 2.0989,
+      "step": 10716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.478845648009049e-05,
+      "loss": 2.0649,
+      "step": 10717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4786154138949736e-05,
+      "loss": 2.061,
+      "step": 10718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4783851823084727e-05,
+      "loss": 2.0203,
+      "step": 10719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4781549532555828e-05,
+      "loss": 2.0198,
+      "step": 10720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4779247267423425e-05,
+      "loss": 1.9998,
+      "step": 10721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4776945027747887e-05,
+      "loss": 2.0303,
+      "step": 10722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4774642813589578e-05,
+      "loss": 2.0725,
+      "step": 10723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4772340625008888e-05,
+      "loss": 2.0759,
+      "step": 10724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4770038462066174e-05,
+      "loss": 1.9895,
+      "step": 10725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.476773632482182e-05,
+      "loss": 2.0112,
+      "step": 10726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4765434213336187e-05,
+      "loss": 2.0222,
+      "step": 10727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4763132127669653e-05,
+      "loss": 2.0242,
+      "step": 10728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4760830067882586e-05,
+      "loss": 2.0437,
+      "step": 10729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4758528034035347e-05,
+      "loss": 2.0283,
+      "step": 10730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4756226026188316e-05,
+      "loss": 2.0618,
+      "step": 10731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4753924044401848e-05,
+      "loss": 2.051,
+      "step": 10732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4751622088736322e-05,
+      "loss": 2.0408,
+      "step": 10733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4749320159252095e-05,
+      "loss": 2.0591,
+      "step": 10734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4747018256009539e-05,
+      "loss": 2.0833,
+      "step": 10735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4744716379069013e-05,
+      "loss": 2.073,
+      "step": 10736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4742414528490884e-05,
+      "loss": 2.0291,
+      "step": 10737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4740112704335515e-05,
+      "loss": 2.0449,
+      "step": 10738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4737810906663273e-05,
+      "loss": 2.0527,
+      "step": 10739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4735509135534509e-05,
+      "loss": 2.0818,
+      "step": 10740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4733207391009596e-05,
+      "loss": 2.0105,
+      "step": 10741
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4730905673148884e-05,
+      "loss": 2.0261,
+      "step": 10742
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4728603982012746e-05,
+      "loss": 2.0447,
+      "step": 10743
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4726302317661532e-05,
+      "loss": 2.0654,
+      "step": 10744
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4724000680155596e-05,
+      "loss": 2.0786,
+      "step": 10745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4721699069555309e-05,
+      "loss": 2.0935,
+      "step": 10746
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4719397485921016e-05,
+      "loss": 2.0356,
+      "step": 10747
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4717095929313082e-05,
+      "loss": 1.9993,
+      "step": 10748
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4714794399791857e-05,
+      "loss": 2.1143,
+      "step": 10749
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4712492897417696e-05,
+      "loss": 2.1067,
+      "step": 10750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4710191422250958e-05,
+      "loss": 2.0527,
+      "step": 10751
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4707889974351991e-05,
+      "loss": 2.0994,
+      "step": 10752
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4705588553781155e-05,
+      "loss": 2.0891,
+      "step": 10753
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4703287160598797e-05,
+      "loss": 2.0791,
+      "step": 10754
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4700985794865261e-05,
+      "loss": 2.0466,
+      "step": 10755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4698684456640915e-05,
+      "loss": 2.0686,
+      "step": 10756
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4696383145986092e-05,
+      "loss": 2.0476,
+      "step": 10757
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4694081862961156e-05,
+      "loss": 2.0613,
+      "step": 10758
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4691780607626446e-05,
+      "loss": 2.0876,
+      "step": 10759
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4689479380042307e-05,
+      "loss": 2.0615,
+      "step": 10760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4687178180269098e-05,
+      "loss": 2.0815,
+      "step": 10761
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.468487700836715e-05,
+      "loss": 2.0818,
+      "step": 10762
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4682575864396825e-05,
+      "loss": 2.0762,
+      "step": 10763
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4680274748418459e-05,
+      "loss": 2.083,
+      "step": 10764
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4677973660492391e-05,
+      "loss": 2.0483,
+      "step": 10765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4675672600678975e-05,
+      "loss": 2.0642,
+      "step": 10766
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4673371569038545e-05,
+      "loss": 2.1121,
+      "step": 10767
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.467107056563145e-05,
+      "loss": 2.1052,
+      "step": 10768
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4668769590518024e-05,
+      "loss": 2.041,
+      "step": 10769
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4666468643758615e-05,
+      "loss": 2.0708,
+      "step": 10770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4664167725413555e-05,
+      "loss": 2.0652,
+      "step": 10771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4661866835543187e-05,
+      "loss": 2.0842,
+      "step": 10772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4659565974207852e-05,
+      "loss": 2.0447,
+      "step": 10773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4657265141467884e-05,
+      "loss": 2.0845,
+      "step": 10774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4654964337383615e-05,
+      "loss": 2.0752,
+      "step": 10775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4652663562015393e-05,
+      "loss": 2.1018,
+      "step": 10776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4650362815423538e-05,
+      "loss": 2.0586,
+      "step": 10777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4648062097668402e-05,
+      "loss": 2.0615,
+      "step": 10778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4645761408810306e-05,
+      "loss": 2.0579,
+      "step": 10779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.464346074890958e-05,
+      "loss": 2.0728,
+      "step": 10780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4641160118026569e-05,
+      "loss": 1.9983,
+      "step": 10781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4638859516221593e-05,
+      "loss": 2.0215,
+      "step": 10782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.463655894355499e-05,
+      "loss": 2.0518,
+      "step": 10783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4634258400087087e-05,
+      "loss": 2.0674,
+      "step": 10784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.463195788587821e-05,
+      "loss": 2.0603,
+      "step": 10785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4629657400988694e-05,
+      "loss": 2.0347,
+      "step": 10786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4627356945478862e-05,
+      "loss": 2.0298,
+      "step": 10787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.462505651940904e-05,
+      "loss": 2.0552,
+      "step": 10788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4622756122839559e-05,
+      "loss": 2.1084,
+      "step": 10789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4620455755830735e-05,
+      "loss": 2.1294,
+      "step": 10790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4618155418442904e-05,
+      "loss": 2.0564,
+      "step": 10791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4615855110736378e-05,
+      "loss": 2.0591,
+      "step": 10792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4613554832771495e-05,
+      "loss": 2.0713,
+      "step": 10793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4611254584608567e-05,
+      "loss": 2.0769,
+      "step": 10794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4608954366307908e-05,
+      "loss": 2.0227,
+      "step": 10795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4606654177929855e-05,
+      "loss": 2.0212,
+      "step": 10796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4604354019534714e-05,
+      "loss": 1.9905,
+      "step": 10797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4602053891182813e-05,
+      "loss": 2.0037,
+      "step": 10798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4599753792934471e-05,
+      "loss": 2.0269,
+      "step": 10799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4597453724849993e-05,
+      "loss": 2.0261,
+      "step": 10800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.459515368698971e-05,
+      "loss": 2.0408,
+      "step": 10801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4592853679413926e-05,
+      "loss": 2.0471,
+      "step": 10802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4590553702182968e-05,
+      "loss": 2.092,
+      "step": 10803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.458825375535714e-05,
+      "loss": 2.1191,
+      "step": 10804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4585953838996762e-05,
+      "loss": 2.0234,
+      "step": 10805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4583653953162142e-05,
+      "loss": 2.0237,
+      "step": 10806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4581354097913593e-05,
+      "loss": 2.0667,
+      "step": 10807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.457905427331143e-05,
+      "loss": 2.0605,
+      "step": 10808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4576754479415962e-05,
+      "loss": 2.0847,
+      "step": 10809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4574454716287491e-05,
+      "loss": 2.0747,
+      "step": 10810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4572154983986337e-05,
+      "loss": 2.0806,
+      "step": 10811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4569855282572796e-05,
+      "loss": 2.0918,
+      "step": 10812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4567555612107187e-05,
+      "loss": 2.0898,
+      "step": 10813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.456525597264981e-05,
+      "loss": 2.1018,
+      "step": 10814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4562956364260965e-05,
+      "loss": 2.0938,
+      "step": 10815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4560656787000968e-05,
+      "loss": 2.0437,
+      "step": 10816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4558357240930113e-05,
+      "loss": 2.0674,
+      "step": 10817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.455605772610871e-05,
+      "loss": 2.0771,
+      "step": 10818
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4553758242597056e-05,
+      "loss": 2.0703,
+      "step": 10819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4551458790455454e-05,
+      "loss": 2.0894,
+      "step": 10820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4549159369744207e-05,
+      "loss": 1.9849,
+      "step": 10821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4546859980523611e-05,
+      "loss": 1.9834,
+      "step": 10822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4544560622853964e-05,
+      "loss": 2.0874,
+      "step": 10823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4542261296795572e-05,
+      "loss": 2.0845,
+      "step": 10824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4539962002408719e-05,
+      "loss": 2.0415,
+      "step": 10825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4537662739753716e-05,
+      "loss": 2.0652,
+      "step": 10826
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4535363508890846e-05,
+      "loss": 2.0459,
+      "step": 10827
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4533064309880414e-05,
+      "loss": 2.033,
+      "step": 10828
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4530765142782707e-05,
+      "loss": 2.0229,
+      "step": 10829
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4528466007658015e-05,
+      "loss": 2.0173,
+      "step": 10830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.452616690456664e-05,
+      "loss": 2.064,
+      "step": 10831
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4523867833568862e-05,
+      "loss": 2.0684,
+      "step": 10832
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4521568794724984e-05,
+      "loss": 2.0837,
+      "step": 10833
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.451926978809529e-05,
+      "loss": 2.0986,
+      "step": 10834
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4516970813740061e-05,
+      "loss": 2.0906,
+      "step": 10835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4514671871719598e-05,
+      "loss": 2.1157,
+      "step": 10836
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4512372962094177e-05,
+      "loss": 2.0691,
+      "step": 10837
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4510074084924095e-05,
+      "loss": 2.0593,
+      "step": 10838
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4507775240269629e-05,
+      "loss": 2.0671,
+      "step": 10839
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4505476428191065e-05,
+      "loss": 2.0576,
+      "step": 10840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.450317764874869e-05,
+      "loss": 2.0657,
+      "step": 10841
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4500878902002782e-05,
+      "loss": 2.0808,
+      "step": 10842
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4498580188013629e-05,
+      "loss": 2.071,
+      "step": 10843
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4496281506841509e-05,
+      "loss": 2.0806,
+      "step": 10844
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4493982858546696e-05,
+      "loss": 2.0818,
+      "step": 10845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4491684243189481e-05,
+      "loss": 2.0674,
+      "step": 10846
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4489385660830133e-05,
+      "loss": 2.0376,
+      "step": 10847
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4487087111528938e-05,
+      "loss": 2.0493,
+      "step": 10848
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4484788595346165e-05,
+      "loss": 2.0903,
+      "step": 10849
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4482490112342097e-05,
+      "loss": 2.1035,
+      "step": 10850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4480191662577006e-05,
+      "loss": 2.0605,
+      "step": 10851
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4477893246111159e-05,
+      "loss": 2.0786,
+      "step": 10852
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.447559486300484e-05,
+      "loss": 2.0723,
+      "step": 10853
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4473296513318315e-05,
+      "loss": 2.0964,
+      "step": 10854
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4470998197111862e-05,
+      "loss": 2.0237,
+      "step": 10855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4468699914445745e-05,
+      "loss": 2.0164,
+      "step": 10856
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4466401665380235e-05,
+      "loss": 2.0493,
+      "step": 10857
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4464103449975603e-05,
+      "loss": 2.0278,
+      "step": 10858
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4461805268292117e-05,
+      "loss": 2.0635,
+      "step": 10859
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4459507120390043e-05,
+      "loss": 2.0571,
+      "step": 10860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4457209006329651e-05,
+      "loss": 2.0583,
+      "step": 10861
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4454910926171197e-05,
+      "loss": 2.0444,
+      "step": 10862
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4452612879974957e-05,
+      "loss": 2.0774,
+      "step": 10863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4450314867801182e-05,
+      "loss": 2.0679,
+      "step": 10864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4448016889710148e-05,
+      "loss": 2.0112,
+      "step": 10865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4445718945762112e-05,
+      "loss": 1.9988,
+      "step": 10866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4443421036017327e-05,
+      "loss": 2.0652,
+      "step": 10867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4441123160536067e-05,
+      "loss": 2.053,
+      "step": 10868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4438825319378575e-05,
+      "loss": 2.031,
+      "step": 10869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4436527512605124e-05,
+      "loss": 2.0481,
+      "step": 10870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4434229740275968e-05,
+      "loss": 2.04,
+      "step": 10871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4431932002451353e-05,
+      "loss": 2.0608,
+      "step": 10872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4429634299191547e-05,
+      "loss": 2.0769,
+      "step": 10873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4427336630556795e-05,
+      "loss": 2.0874,
+      "step": 10874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4425038996607362e-05,
+      "loss": 2.0415,
+      "step": 10875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.442274139740349e-05,
+      "loss": 2.0308,
+      "step": 10876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4420443833005436e-05,
+      "loss": 2.0747,
+      "step": 10877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4418146303473449e-05,
+      "loss": 2.0945,
+      "step": 10878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4415848808867777e-05,
+      "loss": 2.0105,
+      "step": 10879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.441355134924868e-05,
+      "loss": 2.0232,
+      "step": 10880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4411253924676397e-05,
+      "loss": 2.0654,
+      "step": 10881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4408956535211171e-05,
+      "loss": 2.0857,
+      "step": 10882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.440665918091326e-05,
+      "loss": 2.0254,
+      "step": 10883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4404361861842899e-05,
+      "loss": 2.0352,
+      "step": 10884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.440206457806034e-05,
+      "loss": 2.0837,
+      "step": 10885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4399767329625826e-05,
+      "loss": 2.095,
+      "step": 10886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4397470116599591e-05,
+      "loss": 2.1008,
+      "step": 10887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4395172939041888e-05,
+      "loss": 2.085,
+      "step": 10888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4392875797012949e-05,
+      "loss": 2.0583,
+      "step": 10889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4390578690573022e-05,
+      "loss": 2.0801,
+      "step": 10890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.438828161978234e-05,
+      "loss": 2.0593,
+      "step": 10891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4385984584701142e-05,
+      "loss": 2.063,
+      "step": 10892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4383687585389666e-05,
+      "loss": 2.0872,
+      "step": 10893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4381390621908148e-05,
+      "loss": 2.0945,
+      "step": 10894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4379093694316824e-05,
+      "loss": 2.0256,
+      "step": 10895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4376796802675929e-05,
+      "loss": 2.0659,
+      "step": 10896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4374499947045688e-05,
+      "loss": 2.0066,
+      "step": 10897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4372203127486349e-05,
+      "loss": 1.9927,
+      "step": 10898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.4369906344058128e-05,
+      "loss": 2.0801,
+      "step": 10899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4367609596821267e-05,
+      "loss": 2.075,
+      "step": 10900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.436531288583599e-05,
+      "loss": 2.0537,
+      "step": 10901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4363016211162522e-05,
+      "loss": 2.0579,
+      "step": 10902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.43607195728611e-05,
+      "loss": 2.0679,
+      "step": 10903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.435842297099194e-05,
+      "loss": 2.0603,
+      "step": 10904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4356126405615277e-05,
+      "loss": 2.0811,
+      "step": 10905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4353829876791333e-05,
+      "loss": 2.1086,
+      "step": 10906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4351533384580326e-05,
+      "loss": 2.0667,
+      "step": 10907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4349236929042489e-05,
+      "loss": 2.0703,
+      "step": 10908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4346940510238032e-05,
+      "loss": 2.0222,
+      "step": 10909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4344644128227188e-05,
+      "loss": 2.0376,
+      "step": 10910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.434234778307017e-05,
+      "loss": 2.0476,
+      "step": 10911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4340051474827196e-05,
+      "loss": 2.0327,
+      "step": 10912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4337755203558489e-05,
+      "loss": 2.0791,
+      "step": 10913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4335458969324258e-05,
+      "loss": 2.0708,
+      "step": 10914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4333162772184733e-05,
+      "loss": 2.0723,
+      "step": 10915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4330866612200119e-05,
+      "loss": 2.0437,
+      "step": 10916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4328570489430628e-05,
+      "loss": 2.0964,
+      "step": 10917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.432627440393648e-05,
+      "loss": 2.0762,
+      "step": 10918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4323978355777879e-05,
+      "loss": 2.0779,
+      "step": 10919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4321682345015045e-05,
+      "loss": 2.0708,
+      "step": 10920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4319386371708188e-05,
+      "loss": 2.0891,
+      "step": 10921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4317090435917507e-05,
+      "loss": 2.1025,
+      "step": 10922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4314794537703223e-05,
+      "loss": 2.0757,
+      "step": 10923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.431249867712553e-05,
+      "loss": 2.0791,
+      "step": 10924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4310202854244648e-05,
+      "loss": 2.0896,
+      "step": 10925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4307907069120773e-05,
+      "loss": 2.093,
+      "step": 10926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4305611321814114e-05,
+      "loss": 2.0588,
+      "step": 10927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4303315612384868e-05,
+      "loss": 2.0266,
+      "step": 10928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4301019940893245e-05,
+      "loss": 2.0186,
+      "step": 10929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4298724307399444e-05,
+      "loss": 2.0627,
+      "step": 10930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4296428711963665e-05,
+      "loss": 2.0532,
+      "step": 10931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4294133154646103e-05,
+      "loss": 2.0405,
+      "step": 10932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4291837635506962e-05,
+      "loss": 2.0974,
+      "step": 10933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4289542154606436e-05,
+      "loss": 2.1067,
+      "step": 10934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4287246712004727e-05,
+      "loss": 2.116,
+      "step": 10935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4284951307762024e-05,
+      "loss": 2.0967,
+      "step": 10936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.428265594193852e-05,
+      "loss": 2.1145,
+      "step": 10937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4280360614594417e-05,
+      "loss": 2.1138,
+      "step": 10938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4278065325789896e-05,
+      "loss": 2.0847,
+      "step": 10939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4275770075585159e-05,
+      "loss": 2.064,
+      "step": 10940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4273474864040393e-05,
+      "loss": 2.0823,
+      "step": 10941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4271179691215779e-05,
+      "loss": 2.0664,
+      "step": 10942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4268884557171518e-05,
+      "loss": 2.0625,
+      "step": 10943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4266589461967785e-05,
+      "loss": 2.061,
+      "step": 10944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4264294405664778e-05,
+      "loss": 2.0398,
+      "step": 10945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4261999388322675e-05,
+      "loss": 2.0386,
+      "step": 10946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4259704410001662e-05,
+      "loss": 2.0518,
+      "step": 10947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4257409470761921e-05,
+      "loss": 2.075,
+      "step": 10948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4255114570663632e-05,
+      "loss": 2.012,
+      "step": 10949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4252819709766984e-05,
+      "loss": 2.0278,
+      "step": 10950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4250524888132156e-05,
+      "loss": 2.0186,
+      "step": 10951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4248230105819313e-05,
+      "loss": 2.0105,
+      "step": 10952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4245935362888651e-05,
+      "loss": 2.0571,
+      "step": 10953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4243640659400335e-05,
+      "loss": 2.0605,
+      "step": 10954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4241345995414548e-05,
+      "loss": 2.061,
+      "step": 10955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4239051370991464e-05,
+      "loss": 2.0571,
+      "step": 10956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4236756786191249e-05,
+      "loss": 2.0706,
+      "step": 10957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4234462241074086e-05,
+      "loss": 2.0684,
+      "step": 10958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4232167735700138e-05,
+      "loss": 2.0886,
+      "step": 10959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4229873270129586e-05,
+      "loss": 2.0698,
+      "step": 10960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4227578844422589e-05,
+      "loss": 2.0471,
+      "step": 10961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4225284458639324e-05,
+      "loss": 2.0269,
+      "step": 10962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4222990112839954e-05,
+      "loss": 2.0874,
+      "step": 10963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4220695807084646e-05,
+      "loss": 2.0581,
+      "step": 10964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4218401541433567e-05,
+      "loss": 2.0295,
+      "step": 10965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4216107315946884e-05,
+      "loss": 2.0449,
+      "step": 10966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4213813130684751e-05,
+      "loss": 2.0664,
+      "step": 10967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4211518985707342e-05,
+      "loss": 2.0771,
+      "step": 10968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4209224881074807e-05,
+      "loss": 2.03,
+      "step": 10969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4206930816847317e-05,
+      "loss": 2.0361,
+      "step": 10970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4204636793085022e-05,
+      "loss": 2.116,
+      "step": 10971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.420234280984809e-05,
+      "loss": 2.0962,
+      "step": 10972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.420004886719667e-05,
+      "loss": 2.0422,
+      "step": 10973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4197754965190917e-05,
+      "loss": 2.0405,
+      "step": 10974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4195461103890995e-05,
+      "loss": 2.0845,
+      "step": 10975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4193167283357045e-05,
+      "loss": 2.0566,
+      "step": 10976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4190873503649234e-05,
+      "loss": 2.0273,
+      "step": 10977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4188579764827706e-05,
+      "loss": 2.0544,
+      "step": 10978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4186286066952606e-05,
+      "loss": 2.075,
+      "step": 10979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4183992410084096e-05,
+      "loss": 2.0999,
+      "step": 10980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4181698794282313e-05,
+      "loss": 2.0908,
+      "step": 10981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4179405219607415e-05,
+      "loss": 2.0964,
+      "step": 10982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.417711168611954e-05,
+      "loss": 2.0947,
+      "step": 10983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4174818193878833e-05,
+      "loss": 2.0686,
+      "step": 10984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4172524742945448e-05,
+      "loss": 2.0881,
+      "step": 10985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4170231333379515e-05,
+      "loss": 2.1064,
+      "step": 10986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.416793796524119e-05,
+      "loss": 2.0645,
+      "step": 10987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4165644638590606e-05,
+      "loss": 2.0295,
+      "step": 10988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4163351353487896e-05,
+      "loss": 2.1179,
+      "step": 10989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4161058109993214e-05,
+      "loss": 2.1018,
+      "step": 10990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4158764908166682e-05,
+      "loss": 2.0781,
+      "step": 10991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4156471748068452e-05,
+      "loss": 2.0916,
+      "step": 10992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.415417862975865e-05,
+      "loss": 2.0994,
+      "step": 10993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4151885553297408e-05,
+      "loss": 2.0671,
+      "step": 10994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4149592518744868e-05,
+      "loss": 2.1079,
+      "step": 10995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.414729952616115e-05,
+      "loss": 2.125,
+      "step": 10996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.41450065756064e-05,
+      "loss": 2.0554,
+      "step": 10997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4142713667140739e-05,
+      "loss": 2.0732,
+      "step": 10998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4140420800824295e-05,
+      "loss": 2.0977,
+      "step": 10999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.41381279767172e-05,
+      "loss": 2.1042,
+      "step": 11000
+    },
+    {
+      "Bleu_1": 0.7745843426549229,
+      "Bleu_2": 0.6135439693922384,
+      "Bleu_3": 0.4711994549663151,
+      "Bleu_4": 0.35766783933354623,
+      "CIDEr": 1.2067796264743937,
+      "METEOR": 0.2788430901447944,
+      "ROUGE_L": 0.5604946848294885,
+      "epoch": 0.55,
+      "step": 11000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4135835194879579e-05,
+      "loss": 2.0264,
+      "step": 11001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4133542455371558e-05,
+      "loss": 2.0725,
+      "step": 11002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4131249758253263e-05,
+      "loss": 2.0845,
+      "step": 11003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4128957103584806e-05,
+      "loss": 2.0955,
+      "step": 11004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4126664491426328e-05,
+      "loss": 2.0859,
+      "step": 11005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4124371921837934e-05,
+      "loss": 2.0471,
+      "step": 11006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4122079394879756e-05,
+      "loss": 2.0779,
+      "step": 11007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4119786910611908e-05,
+      "loss": 2.0735,
+      "step": 11008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4117494469094498e-05,
+      "loss": 2.0867,
+      "step": 11009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4115202070387662e-05,
+      "loss": 2.0862,
+      "step": 11010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4112909714551497e-05,
+      "loss": 2.1069,
+      "step": 11011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4110617401646132e-05,
+      "loss": 2.0896,
+      "step": 11012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4108325131731674e-05,
+      "loss": 2.0327,
+      "step": 11013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4106032904868227e-05,
+      "loss": 2.0684,
+      "step": 11014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4103740721115917e-05,
+      "loss": 2.0879,
+      "step": 11015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4101448580534842e-05,
+      "loss": 2.0852,
+      "step": 11016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.409915648318512e-05,
+      "loss": 2.072,
+      "step": 11017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4096864429126853e-05,
+      "loss": 2.1011,
+      "step": 11018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4094572418420143e-05,
+      "loss": 2.0073,
+      "step": 11019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4092280451125107e-05,
+      "loss": 2.0059,
+      "step": 11020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4089988527301838e-05,
+      "loss": 2.1028,
+      "step": 11021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.408769664701045e-05,
+      "loss": 2.0759,
+      "step": 11022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.408540481031104e-05,
+      "loss": 2.0886,
+      "step": 11023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4083113017263698e-05,
+      "loss": 2.0876,
+      "step": 11024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4080821267928544e-05,
+      "loss": 2.0706,
+      "step": 11025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4078529562365657e-05,
+      "loss": 2.0627,
+      "step": 11026
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4076237900635151e-05,
+      "loss": 2.063,
+      "step": 11027
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4073946282797113e-05,
+      "loss": 2.0532,
+      "step": 11028
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4071654708911634e-05,
+      "loss": 2.0815,
+      "step": 11029
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.406936317903882e-05,
+      "loss": 2.1057,
+      "step": 11030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4067071693238747e-05,
+      "loss": 2.0425,
+      "step": 11031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4064780251571526e-05,
+      "loss": 2.0608,
+      "step": 11032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4062488854097233e-05,
+      "loss": 2.0664,
+      "step": 11033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4060197500875964e-05,
+      "loss": 2.0637,
+      "step": 11034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4057906191967804e-05,
+      "loss": 2.0679,
+      "step": 11035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4055614927432842e-05,
+      "loss": 2.0754,
+      "step": 11036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4053323707331162e-05,
+      "loss": 2.0239,
+      "step": 11037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4051032531722853e-05,
+      "loss": 2.0454,
+      "step": 11038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4048741400667987e-05,
+      "loss": 2.0942,
+      "step": 11039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4046450314226664e-05,
+      "loss": 2.0815,
+      "step": 11040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4044159272458945e-05,
+      "loss": 2.1111,
+      "step": 11041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4041868275424929e-05,
+      "loss": 2.1299,
+      "step": 11042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4039577323184685e-05,
+      "loss": 2.0134,
+      "step": 11043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4037286415798284e-05,
+      "loss": 2.0,
+      "step": 11044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4034995553325818e-05,
+      "loss": 2.0461,
+      "step": 11045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4032704735827345e-05,
+      "loss": 2.0283,
+      "step": 11046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4030413963362957e-05,
+      "loss": 2.0767,
+      "step": 11047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4028123235992717e-05,
+      "loss": 2.0908,
+      "step": 11048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4025832553776691e-05,
+      "loss": 2.0923,
+      "step": 11049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4023541916774963e-05,
+      "loss": 2.0945,
+      "step": 11050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4021251325047587e-05,
+      "loss": 2.0166,
+      "step": 11051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4018960778654647e-05,
+      "loss": 2.0227,
+      "step": 11052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4016670277656203e-05,
+      "loss": 2.0417,
+      "step": 11053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4014379822112313e-05,
+      "loss": 2.0408,
+      "step": 11054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4012089412083054e-05,
+      "loss": 2.0771,
+      "step": 11055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.400979904762848e-05,
+      "loss": 2.0737,
+      "step": 11056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4007508728808662e-05,
+      "loss": 2.0603,
+      "step": 11057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4005218455683658e-05,
+      "loss": 2.0571,
+      "step": 11058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4002928228313517e-05,
+      "loss": 2.0752,
+      "step": 11059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4000638046758314e-05,
+      "loss": 2.0859,
+      "step": 11060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3998347911078094e-05,
+      "loss": 2.0571,
+      "step": 11061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3996057821332923e-05,
+      "loss": 2.0684,
+      "step": 11062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.399376777758285e-05,
+      "loss": 2.0754,
+      "step": 11063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3991477779887924e-05,
+      "loss": 2.0588,
+      "step": 11064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3989187828308211e-05,
+      "loss": 2.0112,
+      "step": 11065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3986897922903745e-05,
+      "loss": 2.0256,
+      "step": 11066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3984608063734593e-05,
+      "loss": 2.0349,
+      "step": 11067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3982318250860794e-05,
+      "loss": 2.0283,
+      "step": 11068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3980028484342397e-05,
+      "loss": 2.001,
+      "step": 11069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3977738764239453e-05,
+      "loss": 2.0171,
+      "step": 11070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3975449090611999e-05,
+      "loss": 2.1313,
+      "step": 11071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3973159463520088e-05,
+      "loss": 2.1404,
+      "step": 11072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.397086988302376e-05,
+      "loss": 2.0371,
+      "step": 11073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.396858034918305e-05,
+      "loss": 2.0588,
+      "step": 11074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3966290862058009e-05,
+      "loss": 2.1028,
+      "step": 11075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3964001421708663e-05,
+      "loss": 2.071,
+      "step": 11076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3961712028195065e-05,
+      "loss": 2.0571,
+      "step": 11077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3959422681577247e-05,
+      "loss": 2.0315,
+      "step": 11078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3957133381915234e-05,
+      "loss": 2.0977,
+      "step": 11079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3954844129269075e-05,
+      "loss": 2.1306,
+      "step": 11080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3952554923698789e-05,
+      "loss": 2.0459,
+      "step": 11081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.395026576526442e-05,
+      "loss": 2.0603,
+      "step": 11082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3947976654025996e-05,
+      "loss": 2.1206,
+      "step": 11083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3945687590043536e-05,
+      "loss": 2.0947,
+      "step": 11084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3943398573377082e-05,
+      "loss": 2.0505,
+      "step": 11085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3941109604086646e-05,
+      "loss": 2.0425,
+      "step": 11086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.393882068223227e-05,
+      "loss": 2.071,
+      "step": 11087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3936531807873971e-05,
+      "loss": 2.0596,
+      "step": 11088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3934242981071765e-05,
+      "loss": 2.0793,
+      "step": 11089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3931954201885687e-05,
+      "loss": 2.075,
+      "step": 11090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3929665470375745e-05,
+      "loss": 2.0234,
+      "step": 11091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.392737678660197e-05,
+      "loss": 2.0339,
+      "step": 11092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.392508815062437e-05,
+      "loss": 2.0315,
+      "step": 11093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.392279956250297e-05,
+      "loss": 2.0598,
+      "step": 11094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3920511022297782e-05,
+      "loss": 2.0706,
+      "step": 11095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3918222530068814e-05,
+      "loss": 2.073,
+      "step": 11096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3915934085876091e-05,
+      "loss": 2.0063,
+      "step": 11097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3913645689779613e-05,
+      "loss": 2.0242,
+      "step": 11098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.3911357341839403e-05,
+      "loss": 2.0396,
+      "step": 11099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3909069042115463e-05,
+      "loss": 2.0803,
+      "step": 11100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3906780790667796e-05,
+      "loss": 2.0972,
+      "step": 11101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.390449258755642e-05,
+      "loss": 2.1133,
+      "step": 11102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.390220443284133e-05,
+      "loss": 2.1238,
+      "step": 11103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.389991632658254e-05,
+      "loss": 2.1245,
+      "step": 11104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3897628268840047e-05,
+      "loss": 2.0845,
+      "step": 11105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3895340259673852e-05,
+      "loss": 2.0786,
+      "step": 11106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3893052299143956e-05,
+      "loss": 2.0684,
+      "step": 11107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3890764387310361e-05,
+      "loss": 2.0823,
+      "step": 11108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3888476524233062e-05,
+      "loss": 2.0886,
+      "step": 11109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.388618870997206e-05,
+      "loss": 2.1052,
+      "step": 11110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.388390094458734e-05,
+      "loss": 2.0662,
+      "step": 11111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3881613228138912e-05,
+      "loss": 2.0952,
+      "step": 11112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.387932556068675e-05,
+      "loss": 2.0444,
+      "step": 11113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3877037942290863e-05,
+      "loss": 2.0889,
+      "step": 11114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.387475037301123e-05,
+      "loss": 2.0657,
+      "step": 11115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.387246285290784e-05,
+      "loss": 2.0566,
+      "step": 11116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.387017538204069e-05,
+      "loss": 2.0634,
+      "step": 11117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3867887960469752e-05,
+      "loss": 2.073,
+      "step": 11118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3865600588255025e-05,
+      "loss": 2.0708,
+      "step": 11119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3863313265456487e-05,
+      "loss": 2.0864,
+      "step": 11120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3861025992134114e-05,
+      "loss": 2.1057,
+      "step": 11121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3858738768347897e-05,
+      "loss": 2.1001,
+      "step": 11122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3856451594157808e-05,
+      "loss": 2.041,
+      "step": 11123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3854164469623833e-05,
+      "loss": 2.0295,
+      "step": 11124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3851877394805947e-05,
+      "loss": 2.0448,
+      "step": 11125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.384959036976412e-05,
+      "loss": 2.062,
+      "step": 11126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3847303394558333e-05,
+      "loss": 1.9828,
+      "step": 11127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3845016469248554e-05,
+      "loss": 1.99,
+      "step": 11128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3842729593894762e-05,
+      "loss": 2.0496,
+      "step": 11129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3840442768556927e-05,
+      "loss": 2.061,
+      "step": 11130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3838155993295006e-05,
+      "loss": 2.093,
+      "step": 11131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3835869268168984e-05,
+      "loss": 2.0801,
+      "step": 11132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3833582593238811e-05,
+      "loss": 2.084,
+      "step": 11133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3831295968564471e-05,
+      "loss": 2.0696,
+      "step": 11134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3829009394205915e-05,
+      "loss": 2.019,
+      "step": 11135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3826722870223106e-05,
+      "loss": 2.0425,
+      "step": 11136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3824436396676013e-05,
+      "loss": 2.0898,
+      "step": 11137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3822149973624586e-05,
+      "loss": 2.1304,
+      "step": 11138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3819863601128796e-05,
+      "loss": 2.0823,
+      "step": 11139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3817577279248589e-05,
+      "loss": 2.073,
+      "step": 11140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3815291008043928e-05,
+      "loss": 2.0342,
+      "step": 11141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3813004787574764e-05,
+      "loss": 2.0466,
+      "step": 11142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3810718617901054e-05,
+      "loss": 2.0137,
+      "step": 11143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3808432499082748e-05,
+      "loss": 2.0183,
+      "step": 11144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.38061464311798e-05,
+      "loss": 2.0815,
+      "step": 11145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.380386041425215e-05,
+      "loss": 2.0999,
+      "step": 11146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3801574448359761e-05,
+      "loss": 2.0151,
+      "step": 11147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3799288533562563e-05,
+      "loss": 2.0271,
+      "step": 11148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3797002669920517e-05,
+      "loss": 2.024,
+      "step": 11149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3794716857493563e-05,
+      "loss": 2.0513,
+      "step": 11150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.379243109634163e-05,
+      "loss": 2.0293,
+      "step": 11151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3790145386524681e-05,
+      "loss": 2.0674,
+      "step": 11152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3787859728102639e-05,
+      "loss": 2.0779,
+      "step": 11153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3785574121135454e-05,
+      "loss": 2.126,
+      "step": 11154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.378328856568306e-05,
+      "loss": 2.0649,
+      "step": 11155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3781003061805384e-05,
+      "loss": 2.1008,
+      "step": 11156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3778717609562372e-05,
+      "loss": 2.0564,
+      "step": 11157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3776432209013953e-05,
+      "loss": 2.0208,
+      "step": 11158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3774146860220061e-05,
+      "loss": 2.0422,
+      "step": 11159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.377186156324063e-05,
+      "loss": 2.064,
+      "step": 11160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3769576318135575e-05,
+      "loss": 2.1064,
+      "step": 11161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.376729112496484e-05,
+      "loss": 2.0955,
+      "step": 11162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3765005983788341e-05,
+      "loss": 2.1177,
+      "step": 11163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3762720894666014e-05,
+      "loss": 2.1138,
+      "step": 11164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3760435857657775e-05,
+      "loss": 1.9951,
+      "step": 11165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.375815087282354e-05,
+      "loss": 2.0056,
+      "step": 11166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3755865940223245e-05,
+      "loss": 2.1052,
+      "step": 11167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3753581059916796e-05,
+      "loss": 2.1155,
+      "step": 11168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3751296231964124e-05,
+      "loss": 2.0049,
+      "step": 11169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.374901145642514e-05,
+      "loss": 2.0215,
+      "step": 11170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3746726733359753e-05,
+      "loss": 2.1013,
+      "step": 11171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.374444206282789e-05,
+      "loss": 2.1123,
+      "step": 11172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3742157444889449e-05,
+      "loss": 2.0688,
+      "step": 11173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3739872879604356e-05,
+      "loss": 2.0386,
+      "step": 11174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3737588367032513e-05,
+      "loss": 2.1086,
+      "step": 11175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3735303907233829e-05,
+      "loss": 2.1274,
+      "step": 11176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3733019500268213e-05,
+      "loss": 2.0813,
+      "step": 11177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3730735146195572e-05,
+      "loss": 2.062,
+      "step": 11178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3728450845075809e-05,
+      "loss": 2.092,
+      "step": 11179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.372616659696883e-05,
+      "loss": 2.1116,
+      "step": 11180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3723882401934528e-05,
+      "loss": 2.1167,
+      "step": 11181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3721598260032815e-05,
+      "loss": 2.113,
+      "step": 11182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.371931417132358e-05,
+      "loss": 2.0698,
+      "step": 11183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.371703013586673e-05,
+      "loss": 2.0356,
+      "step": 11184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3714746153722157e-05,
+      "loss": 2.0171,
+      "step": 11185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3712462224949747e-05,
+      "loss": 2.0251,
+      "step": 11186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3710178349609408e-05,
+      "loss": 2.0459,
+      "step": 11187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.370789452776102e-05,
+      "loss": 2.0286,
+      "step": 11188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3705610759464486e-05,
+      "loss": 2.0876,
+      "step": 11189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3703327044779688e-05,
+      "loss": 2.0837,
+      "step": 11190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3701043383766506e-05,
+      "loss": 2.0046,
+      "step": 11191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3698759776484838e-05,
+      "loss": 2.0234,
+      "step": 11192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3696476222994565e-05,
+      "loss": 2.0618,
+      "step": 11193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3694192723355573e-05,
+      "loss": 2.1021,
+      "step": 11194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3691909277627742e-05,
+      "loss": 2.064,
+      "step": 11195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3689625885870948e-05,
+      "loss": 2.0808,
+      "step": 11196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.368734254814508e-05,
+      "loss": 2.0212,
+      "step": 11197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3685059264510005e-05,
+      "loss": 2.032,
+      "step": 11198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.368277603502561e-05,
+      "loss": 2.0635,
+      "step": 11199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3680492859751767e-05,
+      "loss": 2.0591,
+      "step": 11200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3678209738748342e-05,
+      "loss": 2.072,
+      "step": 11201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3675926672075219e-05,
+      "loss": 2.0854,
+      "step": 11202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3673643659792255e-05,
+      "loss": 2.0032,
+      "step": 11203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3671360701959336e-05,
+      "loss": 2.0405,
+      "step": 11204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.366907779863632e-05,
+      "loss": 2.0347,
+      "step": 11205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3666794949883069e-05,
+      "loss": 2.051,
+      "step": 11206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.366451215575946e-05,
+      "loss": 2.0896,
+      "step": 11207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3662229416325342e-05,
+      "loss": 2.0969,
+      "step": 11208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3659946731640594e-05,
+      "loss": 2.063,
+      "step": 11209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.365766410176506e-05,
+      "loss": 2.0708,
+      "step": 11210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3655381526758615e-05,
+      "loss": 2.0505,
+      "step": 11211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3653099006681105e-05,
+      "loss": 2.0942,
+      "step": 11212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3650816541592394e-05,
+      "loss": 1.9724,
+      "step": 11213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3648534131552332e-05,
+      "loss": 1.9878,
+      "step": 11214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3646251776620775e-05,
+      "loss": 2.0259,
+      "step": 11215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3643969476857573e-05,
+      "loss": 2.0168,
+      "step": 11216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3641687232322582e-05,
+      "loss": 2.0065,
+      "step": 11217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3639405043075643e-05,
+      "loss": 1.9924,
+      "step": 11218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3637122909176613e-05,
+      "loss": 2.1162,
+      "step": 11219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3634840830685328e-05,
+      "loss": 2.1404,
+      "step": 11220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3632558807661644e-05,
+      "loss": 2.0198,
+      "step": 11221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3630276840165399e-05,
+      "loss": 2.0159,
+      "step": 11222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3627994928256427e-05,
+      "loss": 2.0728,
+      "step": 11223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3625713071994583e-05,
+      "loss": 2.062,
+      "step": 11224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3623431271439694e-05,
+      "loss": 2.0166,
+      "step": 11225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.362114952665161e-05,
+      "loss": 2.01,
+      "step": 11226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3618867837690154e-05,
+      "loss": 1.9482,
+      "step": 11227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3616586204615168e-05,
+      "loss": 2.0034,
+      "step": 11228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3614304627486485e-05,
+      "loss": 2.0552,
+      "step": 11229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3612023106363935e-05,
+      "loss": 2.0481,
+      "step": 11230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3609741641307349e-05,
+      "loss": 2.0867,
+      "step": 11231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3607460232376557e-05,
+      "loss": 2.1018,
+      "step": 11232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.360517887963138e-05,
+      "loss": 2.0569,
+      "step": 11233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3602897583131655e-05,
+      "loss": 2.0637,
+      "step": 11234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3600616342937195e-05,
+      "loss": 2.0789,
+      "step": 11235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.359833515910783e-05,
+      "loss": 2.0884,
+      "step": 11236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3596054031703384e-05,
+      "loss": 2.0957,
+      "step": 11237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3593772960783662e-05,
+      "loss": 2.1072,
+      "step": 11238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.35914919464085e-05,
+      "loss": 2.0273,
+      "step": 11239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3589210988637702e-05,
+      "loss": 2.0613,
+      "step": 11240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3586930087531095e-05,
+      "loss": 2.0857,
+      "step": 11241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3584649243148488e-05,
+      "loss": 2.0798,
+      "step": 11242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3582368455549684e-05,
+      "loss": 2.0564,
+      "step": 11243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3580087724794512e-05,
+      "loss": 2.0667,
+      "step": 11244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3577807050942762e-05,
+      "loss": 2.071,
+      "step": 11245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3575526434054262e-05,
+      "loss": 2.0466,
+      "step": 11246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3573245874188804e-05,
+      "loss": 2.123,
+      "step": 11247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3570965371406197e-05,
+      "loss": 2.1174,
+      "step": 11248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3568684925766244e-05,
+      "loss": 2.1484,
+      "step": 11249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3566404537328752e-05,
+      "loss": 2.115,
+      "step": 11250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3564124206153516e-05,
+      "loss": 2.0779,
+      "step": 11251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3561843932300341e-05,
+      "loss": 2.0891,
+      "step": 11252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3559563715829013e-05,
+      "loss": 2.0977,
+      "step": 11253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.355728355679934e-05,
+      "loss": 2.1323,
+      "step": 11254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3555003455271107e-05,
+      "loss": 2.0632,
+      "step": 11255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3552723411304118e-05,
+      "loss": 2.0723,
+      "step": 11256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3550443424958158e-05,
+      "loss": 2.0149,
+      "step": 11257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3548163496293015e-05,
+      "loss": 2.0222,
+      "step": 11258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3545883625368483e-05,
+      "loss": 2.1116,
+      "step": 11259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3543603812244339e-05,
+      "loss": 2.1067,
+      "step": 11260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3541324056980383e-05,
+      "loss": 2.0754,
+      "step": 11261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3539044359636387e-05,
+      "loss": 2.0764,
+      "step": 11262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3536764720272138e-05,
+      "loss": 1.9863,
+      "step": 11263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3534485138947416e-05,
+      "loss": 1.9897,
+      "step": 11264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3532205615722004e-05,
+      "loss": 2.1038,
+      "step": 11265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3529926150655674e-05,
+      "loss": 2.1011,
+      "step": 11266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.352764674380821e-05,
+      "loss": 1.9871,
+      "step": 11267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3525367395239374e-05,
+      "loss": 2.0237,
+      "step": 11268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3523088105008955e-05,
+      "loss": 2.0273,
+      "step": 11269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3520808873176709e-05,
+      "loss": 2.0435,
+      "step": 11270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3518529699802421e-05,
+      "loss": 2.0078,
+      "step": 11271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3516250584945854e-05,
+      "loss": 2.0122,
+      "step": 11272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3513971528666766e-05,
+      "loss": 2.0223,
+      "step": 11273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3511692531024939e-05,
+      "loss": 2.0615,
+      "step": 11274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3509413592080122e-05,
+      "loss": 2.0493,
+      "step": 11275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.350713471189209e-05,
+      "loss": 2.0618,
+      "step": 11276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3504855890520597e-05,
+      "loss": 2.0762,
+      "step": 11277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3502577128025398e-05,
+      "loss": 2.1108,
+      "step": 11278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3500298424466265e-05,
+      "loss": 2.0415,
+      "step": 11279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3498019779902937e-05,
+      "loss": 2.0483,
+      "step": 11280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3495741194395184e-05,
+      "loss": 2.0327,
+      "step": 11281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.349346266800275e-05,
+      "loss": 1.9951,
+      "step": 11282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3491184200785391e-05,
+      "loss": 2.0225,
+      "step": 11283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3488905792802855e-05,
+      "loss": 2.0757,
+      "step": 11284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3486627444114891e-05,
+      "loss": 2.113,
+      "step": 11285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3484349154781247e-05,
+      "loss": 2.114,
+      "step": 11286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3482070924861672e-05,
+      "loss": 2.0193,
+      "step": 11287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.34797927544159e-05,
+      "loss": 2.0613,
+      "step": 11288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3477514643503684e-05,
+      "loss": 2.063,
+      "step": 11289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3475236592184755e-05,
+      "loss": 2.0461,
+      "step": 11290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3472958600518863e-05,
+      "loss": 2.04,
+      "step": 11291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.347068066856574e-05,
+      "loss": 2.072,
+      "step": 11292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3468402796385115e-05,
+      "loss": 2.0588,
+      "step": 11293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3466124984036737e-05,
+      "loss": 2.075,
+      "step": 11294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3463847231580327e-05,
+      "loss": 2.0654,
+      "step": 11295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3461569539075625e-05,
+      "loss": 2.0583,
+      "step": 11296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3459291906582356e-05,
+      "loss": 2.0579,
+      "step": 11297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3457014334160247e-05,
+      "loss": 2.0955,
+      "step": 11298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3454736821869028e-05,
+      "loss": 2.0688,
+      "step": 11299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.3452459369768423e-05,
+      "loss": 2.0789,
+      "step": 11300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3450181977918157e-05,
+      "loss": 2.0859,
+      "step": 11301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3447904646377953e-05,
+      "loss": 2.0837,
+      "step": 11302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3445627375207521e-05,
+      "loss": 2.0471,
+      "step": 11303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3443350164466594e-05,
+      "loss": 2.0752,
+      "step": 11304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.344107301421488e-05,
+      "loss": 2.0413,
+      "step": 11305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.34387959245121e-05,
+      "loss": 2.0791,
+      "step": 11306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.343651889541797e-05,
+      "loss": 2.1282,
+      "step": 11307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.343424192699219e-05,
+      "loss": 2.106,
+      "step": 11308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3431965019294483e-05,
+      "loss": 2.0405,
+      "step": 11309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3429688172384549e-05,
+      "loss": 2.0559,
+      "step": 11310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3427411386322106e-05,
+      "loss": 2.0952,
+      "step": 11311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3425134661166859e-05,
+      "loss": 2.0662,
+      "step": 11312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3422857996978498e-05,
+      "loss": 2.0701,
+      "step": 11313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3420581393816744e-05,
+      "loss": 2.0813,
+      "step": 11314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3418304851741284e-05,
+      "loss": 2.0513,
+      "step": 11315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3416028370811832e-05,
+      "loss": 2.075,
+      "step": 11316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3413751951088073e-05,
+      "loss": 2.02,
+      "step": 11317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.341147559262971e-05,
+      "loss": 2.0376,
+      "step": 11318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3409199295496437e-05,
+      "loss": 2.126,
+      "step": 11319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3406923059747946e-05,
+      "loss": 2.137,
+      "step": 11320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.340464688544393e-05,
+      "loss": 2.073,
+      "step": 11321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.340237077264408e-05,
+      "loss": 2.0969,
+      "step": 11322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3400094721408079e-05,
+      "loss": 2.0713,
+      "step": 11323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3397818731795624e-05,
+      "loss": 2.0906,
+      "step": 11324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3395542803866385e-05,
+      "loss": 2.0569,
+      "step": 11325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3393266937680065e-05,
+      "loss": 2.0952,
+      "step": 11326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3390991133296331e-05,
+      "loss": 2.0588,
+      "step": 11327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3388715390774864e-05,
+      "loss": 2.0527,
+      "step": 11328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3386439710175352e-05,
+      "loss": 2.0642,
+      "step": 11329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3384164091557463e-05,
+      "loss": 2.0735,
+      "step": 11330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3381888534980878e-05,
+      "loss": 2.0588,
+      "step": 11331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3379613040505265e-05,
+      "loss": 2.0916,
+      "step": 11332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3377337608190306e-05,
+      "loss": 2.0249,
+      "step": 11333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3375062238095664e-05,
+      "loss": 2.0217,
+      "step": 11334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3372786930281007e-05,
+      "loss": 1.9854,
+      "step": 11335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3370511684806006e-05,
+      "loss": 2.0193,
+      "step": 11336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3368236501730326e-05,
+      "loss": 2.0144,
+      "step": 11337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3365961381113632e-05,
+      "loss": 2.0188,
+      "step": 11338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3363686323015585e-05,
+      "loss": 2.0828,
+      "step": 11339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3361411327495842e-05,
+      "loss": 2.0881,
+      "step": 11340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.335913639461407e-05,
+      "loss": 2.092,
+      "step": 11341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3356861524429918e-05,
+      "loss": 2.1067,
+      "step": 11342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3354586717003052e-05,
+      "loss": 1.9436,
+      "step": 11343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3352311972393121e-05,
+      "loss": 1.9717,
+      "step": 11344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.335003729065977e-05,
+      "loss": 2.0613,
+      "step": 11345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3347762671862663e-05,
+      "loss": 2.0852,
+      "step": 11346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3345488116061435e-05,
+      "loss": 2.1057,
+      "step": 11347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.334321362331575e-05,
+      "loss": 2.082,
+      "step": 11348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3340939193685247e-05,
+      "loss": 2.0532,
+      "step": 11349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.333866482722956e-05,
+      "loss": 2.0618,
+      "step": 11350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3336390524008348e-05,
+      "loss": 2.0122,
+      "step": 11351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3334116284081237e-05,
+      "loss": 2.0532,
+      "step": 11352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3331842107507881e-05,
+      "loss": 2.0396,
+      "step": 11353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3329567994347906e-05,
+      "loss": 2.064,
+      "step": 11354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3327293944660955e-05,
+      "loss": 2.0402,
+      "step": 11355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.332501995850666e-05,
+      "loss": 2.0325,
+      "step": 11356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3322746035944646e-05,
+      "loss": 2.0701,
+      "step": 11357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.332047217703456e-05,
+      "loss": 2.0688,
+      "step": 11358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3318198381836022e-05,
+      "loss": 2.0603,
+      "step": 11359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3315924650408656e-05,
+      "loss": 2.0552,
+      "step": 11360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3313650982812096e-05,
+      "loss": 2.0947,
+      "step": 11361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3311377379105958e-05,
+      "loss": 2.1174,
+      "step": 11362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3309103839349873e-05,
+      "loss": 2.0532,
+      "step": 11363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.330683036360346e-05,
+      "loss": 2.0776,
+      "step": 11364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.330455695192633e-05,
+      "loss": 2.0732,
+      "step": 11365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3302283604378115e-05,
+      "loss": 2.0879,
+      "step": 11366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3300010321018415e-05,
+      "loss": 2.0415,
+      "step": 11367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3297737101906858e-05,
+      "loss": 2.0427,
+      "step": 11368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3295463947103046e-05,
+      "loss": 2.074,
+      "step": 11369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3293190856666596e-05,
+      "loss": 2.083,
+      "step": 11370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3290917830657115e-05,
+      "loss": 2.0242,
+      "step": 11371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3288644869134213e-05,
+      "loss": 2.0288,
+      "step": 11372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3286371972157492e-05,
+      "loss": 2.0474,
+      "step": 11373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3284099139786563e-05,
+      "loss": 2.0327,
+      "step": 11374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3281826372081014e-05,
+      "loss": 2.0784,
+      "step": 11375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3279553669100463e-05,
+      "loss": 2.0867,
+      "step": 11376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3277281030904494e-05,
+      "loss": 2.0195,
+      "step": 11377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3275008457552718e-05,
+      "loss": 2.033,
+      "step": 11378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3272735949104725e-05,
+      "loss": 2.0208,
+      "step": 11379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.32704635056201e-05,
+      "loss": 2.019,
+      "step": 11380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3268191127158451e-05,
+      "loss": 2.0234,
+      "step": 11381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3265918813779354e-05,
+      "loss": 2.0295,
+      "step": 11382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.326364656554241e-05,
+      "loss": 2.0969,
+      "step": 11383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3261374382507201e-05,
+      "loss": 2.073,
+      "step": 11384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3259102264733304e-05,
+      "loss": 2.0652,
+      "step": 11385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3256830212280318e-05,
+      "loss": 2.0938,
+      "step": 11386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3254558225207809e-05,
+      "loss": 2.0317,
+      "step": 11387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3252286303575374e-05,
+      "loss": 2.0435,
+      "step": 11388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.325001444744258e-05,
+      "loss": 2.0442,
+      "step": 11389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3247742656869006e-05,
+      "loss": 2.053,
+      "step": 11390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3245470931914227e-05,
+      "loss": 2.052,
+      "step": 11391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3243199272637815e-05,
+      "loss": 2.072,
+      "step": 11392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.324092767909935e-05,
+      "loss": 2.1147,
+      "step": 11393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3238656151358396e-05,
+      "loss": 2.0925,
+      "step": 11394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3236384689474514e-05,
+      "loss": 2.1001,
+      "step": 11395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3234113293507285e-05,
+      "loss": 2.0667,
+      "step": 11396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3231841963516259e-05,
+      "loss": 2.0444,
+      "step": 11397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3229570699561012e-05,
+      "loss": 2.0676,
+      "step": 11398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3227299501701103e-05,
+      "loss": 2.1003,
+      "step": 11399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3225028369996078e-05,
+      "loss": 2.115,
+      "step": 11400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3222757304505511e-05,
+      "loss": 2.0614,
+      "step": 11401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3220486305288946e-05,
+      "loss": 2.1099,
+      "step": 11402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.321821537240595e-05,
+      "loss": 2.0149,
+      "step": 11403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3215944505916066e-05,
+      "loss": 2.0571,
+      "step": 11404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3213673705878847e-05,
+      "loss": 2.0498,
+      "step": 11405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3211402972353843e-05,
+      "loss": 2.0425,
+      "step": 11406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.32091323054006e-05,
+      "loss": 2.0942,
+      "step": 11407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3206861705078666e-05,
+      "loss": 2.1025,
+      "step": 11408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3204591171447584e-05,
+      "loss": 2.0735,
+      "step": 11409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3202320704566892e-05,
+      "loss": 2.0833,
+      "step": 11410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.320005030449614e-05,
+      "loss": 2.073,
+      "step": 11411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3197779971294853e-05,
+      "loss": 2.0764,
+      "step": 11412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3195509705022582e-05,
+      "loss": 2.0918,
+      "step": 11413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3193239505738854e-05,
+      "loss": 2.0964,
+      "step": 11414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.31909693735032e-05,
+      "loss": 2.0386,
+      "step": 11415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.318869930837516e-05,
+      "loss": 2.0613,
+      "step": 11416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3186429310414255e-05,
+      "loss": 2.0396,
+      "step": 11417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3184159379680023e-05,
+      "loss": 2.0459,
+      "step": 11418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3181889516231983e-05,
+      "loss": 2.0276,
+      "step": 11419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3179619720129657e-05,
+      "loss": 2.0691,
+      "step": 11420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3177349991432577e-05,
+      "loss": 2.0764,
+      "step": 11421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3175080330200252e-05,
+      "loss": 2.0703,
+      "step": 11422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3172810736492215e-05,
+      "loss": 2.1089,
+      "step": 11423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3170541210367974e-05,
+      "loss": 2.0925,
+      "step": 11424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3168271751887045e-05,
+      "loss": 2.0896,
+      "step": 11425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3166002361108947e-05,
+      "loss": 2.1096,
+      "step": 11426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3163733038093185e-05,
+      "loss": 2.0662,
+      "step": 11427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3161463782899282e-05,
+      "loss": 2.0547,
+      "step": 11428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3159194595586734e-05,
+      "loss": 1.9802,
+      "step": 11429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3156925476215048e-05,
+      "loss": 2.0017,
+      "step": 11430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3154656424843741e-05,
+      "loss": 2.0325,
+      "step": 11431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3152387441532301e-05,
+      "loss": 2.0972,
+      "step": 11432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3150118526340243e-05,
+      "loss": 2.0271,
+      "step": 11433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.314784967932706e-05,
+      "loss": 2.0088,
+      "step": 11434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3145580900552245e-05,
+      "loss": 2.0515,
+      "step": 11435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3143312190075304e-05,
+      "loss": 2.0227,
+      "step": 11436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3141043547955723e-05,
+      "loss": 2.0178,
+      "step": 11437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3138774974253002e-05,
+      "loss": 2.0337,
+      "step": 11438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3136506469026627e-05,
+      "loss": 2.0962,
+      "step": 11439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3134238032336089e-05,
+      "loss": 2.0911,
+      "step": 11440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.313196966424087e-05,
+      "loss": 2.1133,
+      "step": 11441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3129701364800462e-05,
+      "loss": 2.0947,
+      "step": 11442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3127433134074346e-05,
+      "loss": 2.0598,
+      "step": 11443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3125164972122005e-05,
+      "loss": 2.071,
+      "step": 11444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3122896879002912e-05,
+      "loss": 2.0757,
+      "step": 11445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3120628854776556e-05,
+      "loss": 2.0916,
+      "step": 11446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3118360899502403e-05,
+      "loss": 2.0354,
+      "step": 11447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.311609301323994e-05,
+      "loss": 2.0645,
+      "step": 11448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.311382519604863e-05,
+      "loss": 2.0662,
+      "step": 11449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.311155744798794e-05,
+      "loss": 2.0393,
+      "step": 11450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3109289769117351e-05,
+      "loss": 2.0371,
+      "step": 11451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3107022159496318e-05,
+      "loss": 2.0193,
+      "step": 11452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3104754619184321e-05,
+      "loss": 2.0315,
+      "step": 11453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3102487148240808e-05,
+      "loss": 2.0688,
+      "step": 11454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3100219746725253e-05,
+      "loss": 2.0898,
+      "step": 11455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3097952414697112e-05,
+      "loss": 2.1184,
+      "step": 11456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3095685152215837e-05,
+      "loss": 2.0891,
+      "step": 11457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3093417959340894e-05,
+      "loss": 2.0642,
+      "step": 11458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3091150836131727e-05,
+      "loss": 2.0984,
+      "step": 11459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3088883782647802e-05,
+      "loss": 2.0754,
+      "step": 11460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3086616798948559e-05,
+      "loss": 2.0774,
+      "step": 11461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3084349885093448e-05,
+      "loss": 2.1057,
+      "step": 11462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3082083041141925e-05,
+      "loss": 2.074,
+      "step": 11463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3079816267153421e-05,
+      "loss": 2.0554,
+      "step": 11464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3077549563187396e-05,
+      "loss": 2.0923,
+      "step": 11465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3075282929303283e-05,
+      "loss": 2.0881,
+      "step": 11466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3073016365560516e-05,
+      "loss": 2.0842,
+      "step": 11467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3070749872018546e-05,
+      "loss": 2.1016,
+      "step": 11468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3068483448736797e-05,
+      "loss": 2.1091,
+      "step": 11469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3066217095774712e-05,
+      "loss": 2.1201,
+      "step": 11470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.306395081319172e-05,
+      "loss": 2.1189,
+      "step": 11471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.306168460104725e-05,
+      "loss": 2.1028,
+      "step": 11472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3059418459400735e-05,
+      "loss": 2.0381,
+      "step": 11473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3057152388311594e-05,
+      "loss": 2.0261,
+      "step": 11474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3054886387839264e-05,
+      "loss": 1.9852,
+      "step": 11475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3052620458043159e-05,
+      "loss": 2.0403,
+      "step": 11476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3050354598982702e-05,
+      "loss": 2.03,
+      "step": 11477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3048088810717316e-05,
+      "loss": 2.0557,
+      "step": 11478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3045823093306414e-05,
+      "loss": 2.0261,
+      "step": 11479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3043557446809416e-05,
+      "loss": 2.0334,
+      "step": 11480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3041291871285737e-05,
+      "loss": 2.0359,
+      "step": 11481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.303902636679478e-05,
+      "loss": 2.0471,
+      "step": 11482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3036760933395967e-05,
+      "loss": 2.0057,
+      "step": 11483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3034495571148695e-05,
+      "loss": 2.0227,
+      "step": 11484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3032230280112382e-05,
+      "loss": 2.0801,
+      "step": 11485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3029965060346427e-05,
+      "loss": 2.0706,
+      "step": 11486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3027699911910228e-05,
+      "loss": 2.0762,
+      "step": 11487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3025434834863197e-05,
+      "loss": 2.0674,
+      "step": 11488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3023169829264718e-05,
+      "loss": 2.0388,
+      "step": 11489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3020904895174204e-05,
+      "loss": 2.0171,
+      "step": 11490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3018640032651044e-05,
+      "loss": 2.0464,
+      "step": 11491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3016375241754624e-05,
+      "loss": 2.0818,
+      "step": 11492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3014110522544348e-05,
+      "loss": 2.0945,
+      "step": 11493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3011845875079593e-05,
+      "loss": 2.0803,
+      "step": 11494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.300958129941976e-05,
+      "loss": 2.1208,
+      "step": 11495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3007316795624224e-05,
+      "loss": 2.1028,
+      "step": 11496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.300505236375237e-05,
+      "loss": 2.0342,
+      "step": 11497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3002788003863592e-05,
+      "loss": 2.0403,
+      "step": 11498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.3000523716017254e-05,
+      "loss": 2.1128,
+      "step": 11499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.2998259500272748e-05,
+      "loss": 2.0959,
+      "step": 11500
+    },
+    {
+      "Bleu_1": 0.7730332411767514,
+      "Bleu_2": 0.6119389743114464,
+      "Bleu_3": 0.46958079962703664,
+      "Bleu_4": 0.3563338879881635,
+      "CIDEr": 1.201691518216842,
+      "METEOR": 0.27768164897049125,
+      "ROUGE_L": 0.5587687041497557,
+      "epoch": 0.57,
+      "step": 11500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2995995356689445e-05,
+      "loss": 2.0847,
+      "step": 11501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2993731285326711e-05,
+      "loss": 2.095,
+      "step": 11502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2991467286243935e-05,
+      "loss": 1.9966,
+      "step": 11503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.298920335950047e-05,
+      "loss": 2.0217,
+      "step": 11504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2986939505155702e-05,
+      "loss": 2.0735,
+      "step": 11505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.298467572326899e-05,
+      "loss": 2.0952,
+      "step": 11506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2982412013899695e-05,
+      "loss": 2.0889,
+      "step": 11507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2980148377107185e-05,
+      "loss": 2.084,
+      "step": 11508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2977884812950818e-05,
+      "loss": 2.0408,
+      "step": 11509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2975621321489962e-05,
+      "loss": 2.0432,
+      "step": 11510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2973357902783964e-05,
+      "loss": 2.0435,
+      "step": 11511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2971094556892183e-05,
+      "loss": 2.0525,
+      "step": 11512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2968831283873977e-05,
+      "loss": 2.0742,
+      "step": 11513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2966568083788692e-05,
+      "loss": 2.0781,
+      "step": 11514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2964304956695679e-05,
+      "loss": 1.9783,
+      "step": 11515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.296204190265429e-05,
+      "loss": 1.9993,
+      "step": 11516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2959778921723863e-05,
+      "loss": 2.075,
+      "step": 11517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2957516013963751e-05,
+      "loss": 2.062,
+      "step": 11518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2955253179433287e-05,
+      "loss": 2.0181,
+      "step": 11519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2952990418191822e-05,
+      "loss": 2.0334,
+      "step": 11520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2950727730298688e-05,
+      "loss": 2.0386,
+      "step": 11521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2948465115813216e-05,
+      "loss": 2.0583,
+      "step": 11522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2946202574794754e-05,
+      "loss": 2.0747,
+      "step": 11523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.294394010730262e-05,
+      "loss": 2.062,
+      "step": 11524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2941677713396157e-05,
+      "loss": 2.0728,
+      "step": 11525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2939415393134687e-05,
+      "loss": 2.1094,
+      "step": 11526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2937153146577532e-05,
+      "loss": 2.0989,
+      "step": 11527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.293489097378403e-05,
+      "loss": 2.0972,
+      "step": 11528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2932628874813489e-05,
+      "loss": 2.1257,
+      "step": 11529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.293036684972524e-05,
+      "loss": 2.1082,
+      "step": 11530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2928104898578602e-05,
+      "loss": 2.0681,
+      "step": 11531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2925843021432883e-05,
+      "loss": 2.0415,
+      "step": 11532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2923581218347414e-05,
+      "loss": 2.0986,
+      "step": 11533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.292131948938149e-05,
+      "loss": 2.1033,
+      "step": 11534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2919057834594437e-05,
+      "loss": 2.0342,
+      "step": 11535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2916796254045558e-05,
+      "loss": 2.0098,
+      "step": 11536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2914534747794153e-05,
+      "loss": 2.1267,
+      "step": 11537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2912273315899543e-05,
+      "loss": 2.1123,
+      "step": 11538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2910011958421016e-05,
+      "loss": 2.0964,
+      "step": 11539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2907750675417887e-05,
+      "loss": 2.103,
+      "step": 11540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2905489466949449e-05,
+      "loss": 2.0327,
+      "step": 11541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2903228333074995e-05,
+      "loss": 2.0601,
+      "step": 11542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.290096727385383e-05,
+      "loss": 2.1055,
+      "step": 11543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2898706289345237e-05,
+      "loss": 2.0957,
+      "step": 11544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.289644537960852e-05,
+      "loss": 2.0063,
+      "step": 11545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2894184544702962e-05,
+      "loss": 2.0,
+      "step": 11546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2891923784687848e-05,
+      "loss": 2.0364,
+      "step": 11547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2889663099622467e-05,
+      "loss": 2.0369,
+      "step": 11548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2887402489566104e-05,
+      "loss": 2.1008,
+      "step": 11549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2885141954578039e-05,
+      "loss": 2.105,
+      "step": 11550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2882881494717556e-05,
+      "loss": 2.0671,
+      "step": 11551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2880621110043924e-05,
+      "loss": 2.0507,
+      "step": 11552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.287836080061643e-05,
+      "loss": 1.9927,
+      "step": 11553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2876100566494338e-05,
+      "loss": 2.0166,
+      "step": 11554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.287384040773693e-05,
+      "loss": 2.0122,
+      "step": 11555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.287158032440347e-05,
+      "loss": 2.0273,
+      "step": 11556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2869320316553221e-05,
+      "loss": 2.0192,
+      "step": 11557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2867060384245463e-05,
+      "loss": 2.0725,
+      "step": 11558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2864800527539442e-05,
+      "loss": 2.1096,
+      "step": 11559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2862540746494438e-05,
+      "loss": 2.1086,
+      "step": 11560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2860281041169702e-05,
+      "loss": 2.0901,
+      "step": 11561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2858021411624489e-05,
+      "loss": 2.1003,
+      "step": 11562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2855761857918064e-05,
+      "loss": 2.0317,
+      "step": 11563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2853502380109671e-05,
+      "loss": 2.0786,
+      "step": 11564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2851242978258573e-05,
+      "loss": 2.0833,
+      "step": 11565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2848983652424014e-05,
+      "loss": 2.0981,
+      "step": 11566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.284672440266524e-05,
+      "loss": 2.0117,
+      "step": 11567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2844465229041505e-05,
+      "loss": 2.0349,
+      "step": 11568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2842206131612044e-05,
+      "loss": 2.052,
+      "step": 11569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.283994711043611e-05,
+      "loss": 2.0347,
+      "step": 11570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2837688165572935e-05,
+      "loss": 2.0388,
+      "step": 11571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2835429297081756e-05,
+      "loss": 2.0371,
+      "step": 11572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2833170505021816e-05,
+      "loss": 2.1025,
+      "step": 11573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.283091178945234e-05,
+      "loss": 2.1208,
+      "step": 11574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2828653150432571e-05,
+      "loss": 2.0986,
+      "step": 11575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.282639458802173e-05,
+      "loss": 2.0718,
+      "step": 11576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2824136102279055e-05,
+      "loss": 2.0457,
+      "step": 11577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2821877693263767e-05,
+      "loss": 2.0278,
+      "step": 11578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2819619361035082e-05,
+      "loss": 2.0381,
+      "step": 11579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2817361105652236e-05,
+      "loss": 2.0554,
+      "step": 11580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2815102927174437e-05,
+      "loss": 2.1016,
+      "step": 11581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2812844825660914e-05,
+      "loss": 2.0928,
+      "step": 11582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2810586801170876e-05,
+      "loss": 2.0862,
+      "step": 11583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.280832885376354e-05,
+      "loss": 2.0718,
+      "step": 11584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2806070983498114e-05,
+      "loss": 2.0793,
+      "step": 11585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2803813190433815e-05,
+      "loss": 2.0552,
+      "step": 11586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2801555474629847e-05,
+      "loss": 2.0391,
+      "step": 11587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2799297836145416e-05,
+      "loss": 2.0173,
+      "step": 11588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2797040275039722e-05,
+      "loss": 2.0669,
+      "step": 11589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2794782791371977e-05,
+      "loss": 2.0771,
+      "step": 11590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2792525385201369e-05,
+      "loss": 2.0232,
+      "step": 11591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2790268056587106e-05,
+      "loss": 2.0449,
+      "step": 11592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2788010805588382e-05,
+      "loss": 2.0916,
+      "step": 11593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2785753632264381e-05,
+      "loss": 2.0605,
+      "step": 11594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2783496536674308e-05,
+      "loss": 2.0222,
+      "step": 11595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2781239518877342e-05,
+      "loss": 2.0408,
+      "step": 11596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2778982578932679e-05,
+      "loss": 2.0825,
+      "step": 11597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2776725716899505e-05,
+      "loss": 2.1172,
+      "step": 11598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2774468932836992e-05,
+      "loss": 2.0601,
+      "step": 11599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2772212226804331e-05,
+      "loss": 2.074,
+      "step": 11600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2769955598860701e-05,
+      "loss": 2.0725,
+      "step": 11601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2767699049065278e-05,
+      "loss": 2.051,
+      "step": 11602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.276544257747724e-05,
+      "loss": 2.068,
+      "step": 11603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2763186184155752e-05,
+      "loss": 2.0911,
+      "step": 11604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2760929869159999e-05,
+      "loss": 2.0635,
+      "step": 11605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2758673632549136e-05,
+      "loss": 2.0718,
+      "step": 11606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2756417474382342e-05,
+      "loss": 2.1177,
+      "step": 11607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2754161394718777e-05,
+      "loss": 2.0815,
+      "step": 11608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2751905393617599e-05,
+      "loss": 2.0872,
+      "step": 11609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2749649471137977e-05,
+      "loss": 2.0706,
+      "step": 11610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2747393627339062e-05,
+      "loss": 2.1409,
+      "step": 11611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2745137862280024e-05,
+      "loss": 2.1379,
+      "step": 11612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2742882176020008e-05,
+      "loss": 2.0735,
+      "step": 11613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2740626568618162e-05,
+      "loss": 2.0693,
+      "step": 11614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.273837104013365e-05,
+      "loss": 2.0605,
+      "step": 11615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2736115590625606e-05,
+      "loss": 2.0645,
+      "step": 11616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2733860220153191e-05,
+      "loss": 2.0969,
+      "step": 11617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2731604928775541e-05,
+      "loss": 2.0603,
+      "step": 11618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2729349716551799e-05,
+      "loss": 2.061,
+      "step": 11619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2727094583541107e-05,
+      "loss": 2.0767,
+      "step": 11620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.27248395298026e-05,
+      "loss": 2.0972,
+      "step": 11621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.272258455539542e-05,
+      "loss": 2.0828,
+      "step": 11622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.27203296603787e-05,
+      "loss": 2.0603,
+      "step": 11623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2718074844811564e-05,
+      "loss": 2.0393,
+      "step": 11624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2715820108753151e-05,
+      "loss": 2.1213,
+      "step": 11625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2713565452262582e-05,
+      "loss": 2.1162,
+      "step": 11626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2711310875398992e-05,
+      "loss": 2.0698,
+      "step": 11627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2709056378221502e-05,
+      "loss": 2.0591,
+      "step": 11628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.270680196078922e-05,
+      "loss": 2.0593,
+      "step": 11629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2704547623161284e-05,
+      "loss": 2.0625,
+      "step": 11630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.27022933653968e-05,
+      "loss": 2.0078,
+      "step": 11631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2700039187554889e-05,
+      "loss": 2.0549,
+      "step": 11632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2697785089694665e-05,
+      "loss": 2.1006,
+      "step": 11633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2695531071875227e-05,
+      "loss": 2.0718,
+      "step": 11634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2693277134155699e-05,
+      "loss": 2.0669,
+      "step": 11635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2691023276595179e-05,
+      "loss": 2.0742,
+      "step": 11636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2688769499252777e-05,
+      "loss": 2.0667,
+      "step": 11637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2686515802187597e-05,
+      "loss": 2.0386,
+      "step": 11638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2684262185458729e-05,
+      "loss": 2.0449,
+      "step": 11639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2682008649125285e-05,
+      "loss": 2.0974,
+      "step": 11640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2679755193246347e-05,
+      "loss": 2.0264,
+      "step": 11641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2677501817881026e-05,
+      "loss": 2.082,
+      "step": 11642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2675248523088405e-05,
+      "loss": 2.041,
+      "step": 11643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2672995308927568e-05,
+      "loss": 2.0549,
+      "step": 11644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2670742175457615e-05,
+      "loss": 2.0632,
+      "step": 11645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.266848912273762e-05,
+      "loss": 2.0481,
+      "step": 11646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2666236150826679e-05,
+      "loss": 2.0562,
+      "step": 11647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2663983259783868e-05,
+      "loss": 2.0762,
+      "step": 11648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.266173044966826e-05,
+      "loss": 2.0266,
+      "step": 11649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2659477720538943e-05,
+      "loss": 2.0503,
+      "step": 11650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2657225072454981e-05,
+      "loss": 2.1091,
+      "step": 11651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2654972505475461e-05,
+      "loss": 2.1172,
+      "step": 11652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2652720019659445e-05,
+      "loss": 2.0911,
+      "step": 11653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2650467615066e-05,
+      "loss": 2.1099,
+      "step": 11654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2648215291754196e-05,
+      "loss": 2.0852,
+      "step": 11655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.26459630497831e-05,
+      "loss": 2.1101,
+      "step": 11656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.264371088921177e-05,
+      "loss": 2.1213,
+      "step": 11657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2641458810099273e-05,
+      "loss": 2.1262,
+      "step": 11658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2639206812504656e-05,
+      "loss": 2.1074,
+      "step": 11659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2636954896486987e-05,
+      "loss": 2.1123,
+      "step": 11660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2634703062105309e-05,
+      "loss": 1.9946,
+      "step": 11661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2632451309418683e-05,
+      "loss": 2.0549,
+      "step": 11662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2630199638486155e-05,
+      "loss": 2.0613,
+      "step": 11663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2627948049366769e-05,
+      "loss": 2.0505,
+      "step": 11664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2625696542119578e-05,
+      "loss": 2.0837,
+      "step": 11665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2623445116803612e-05,
+      "loss": 2.0781,
+      "step": 11666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.262119377347793e-05,
+      "loss": 2.0796,
+      "step": 11667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.261894251220156e-05,
+      "loss": 2.0574,
+      "step": 11668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2616691333033537e-05,
+      "loss": 2.0444,
+      "step": 11669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2614440236032898e-05,
+      "loss": 2.0564,
+      "step": 11670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.261218922125868e-05,
+      "loss": 1.9939,
+      "step": 11671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2609938288769907e-05,
+      "loss": 2.0364,
+      "step": 11672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2607687438625611e-05,
+      "loss": 2.0942,
+      "step": 11673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2605436670884812e-05,
+      "loss": 2.1162,
+      "step": 11674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2603185985606545e-05,
+      "loss": 2.1052,
+      "step": 11675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.260093538284982e-05,
+      "loss": 2.1084,
+      "step": 11676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2598684862673664e-05,
+      "loss": 2.105,
+      "step": 11677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2596434425137093e-05,
+      "loss": 2.1204,
+      "step": 11678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2594184070299115e-05,
+      "loss": 2.104,
+      "step": 11679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2591933798218756e-05,
+      "loss": 2.1013,
+      "step": 11680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2589683608955012e-05,
+      "loss": 2.0842,
+      "step": 11681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2587433502566905e-05,
+      "loss": 2.0356,
+      "step": 11682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2585183479113436e-05,
+      "loss": 2.0674,
+      "step": 11683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2582933538653603e-05,
+      "loss": 2.0374,
+      "step": 11684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2580683681246419e-05,
+      "loss": 2.1201,
+      "step": 11685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2578433906950872e-05,
+      "loss": 2.0488,
+      "step": 11686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2576184215825973e-05,
+      "loss": 2.0908,
+      "step": 11687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2573934607930708e-05,
+      "loss": 2.0957,
+      "step": 11688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2571685083324074e-05,
+      "loss": 2.0415,
+      "step": 11689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.256943564206506e-05,
+      "loss": 2.0671,
+      "step": 11690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2567186284212656e-05,
+      "loss": 2.0344,
+      "step": 11691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2564937009825852e-05,
+      "loss": 2.0342,
+      "step": 11692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2562687818963633e-05,
+      "loss": 2.0093,
+      "step": 11693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.256043871168497e-05,
+      "loss": 2.0247,
+      "step": 11694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2558189688048857e-05,
+      "loss": 2.0027,
+      "step": 11695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2555940748114263e-05,
+      "loss": 2.0234,
+      "step": 11696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2553691891940173e-05,
+      "loss": 2.0317,
+      "step": 11697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2551443119585549e-05,
+      "loss": 2.0574,
+      "step": 11698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2549194431109374e-05,
+      "loss": 2.0396,
+      "step": 11699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2546945826570611e-05,
+      "loss": 2.0568,
+      "step": 11700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2544697306028222e-05,
+      "loss": 2.0381,
+      "step": 11701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2542448869541184e-05,
+      "loss": 2.0261,
+      "step": 11702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2540200517168447e-05,
+      "loss": 2.0779,
+      "step": 11703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2537952248968983e-05,
+      "loss": 2.0627,
+      "step": 11704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2535704065001746e-05,
+      "loss": 2.0996,
+      "step": 11705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2533455965325687e-05,
+      "loss": 2.1096,
+      "step": 11706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2531207949999763e-05,
+      "loss": 2.0122,
+      "step": 11707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.252896001908293e-05,
+      "loss": 1.9385,
+      "step": 11708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2526712172634132e-05,
+      "loss": 2.0652,
+      "step": 11709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.252446441071232e-05,
+      "loss": 2.0669,
+      "step": 11710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.252221673337643e-05,
+      "loss": 2.0786,
+      "step": 11711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.251996914068542e-05,
+      "loss": 2.095,
+      "step": 11712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2517721632698215e-05,
+      "loss": 1.9902,
+      "step": 11713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2515474209473766e-05,
+      "loss": 2.0117,
+      "step": 11714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2513226871071004e-05,
+      "loss": 2.0139,
+      "step": 11715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2510979617548857e-05,
+      "loss": 2.0454,
+      "step": 11716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2508732448966267e-05,
+      "loss": 2.0649,
+      "step": 11717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2506485365382153e-05,
+      "loss": 2.1013,
+      "step": 11718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2504238366855453e-05,
+      "loss": 2.0066,
+      "step": 11719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2501991453445085e-05,
+      "loss": 2.0359,
+      "step": 11720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2499744625209969e-05,
+      "loss": 2.0659,
+      "step": 11721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2497497882209035e-05,
+      "loss": 2.0483,
+      "step": 11722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.249525122450119e-05,
+      "loss": 2.0417,
+      "step": 11723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.249300465214536e-05,
+      "loss": 2.0569,
+      "step": 11724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2490758165200455e-05,
+      "loss": 2.0986,
+      "step": 11725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2488511763725382e-05,
+      "loss": 2.105,
+      "step": 11726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2486265447779057e-05,
+      "loss": 1.99,
+      "step": 11727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2484019217420382e-05,
+      "loss": 2.0002,
+      "step": 11728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2481773072708267e-05,
+      "loss": 2.083,
+      "step": 11729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2479527013701614e-05,
+      "loss": 2.1123,
+      "step": 11730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2477281040459313e-05,
+      "loss": 2.092,
+      "step": 11731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2475035153040277e-05,
+      "loss": 2.095,
+      "step": 11732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.247278935150339e-05,
+      "loss": 2.0747,
+      "step": 11733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2470543635907557e-05,
+      "loss": 2.0754,
+      "step": 11734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2468298006311661e-05,
+      "loss": 2.0642,
+      "step": 11735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2466052462774585e-05,
+      "loss": 2.0723,
+      "step": 11736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.246380700535523e-05,
+      "loss": 2.0596,
+      "step": 11737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.246156163411247e-05,
+      "loss": 2.1006,
+      "step": 11738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2459316349105195e-05,
+      "loss": 2.0977,
+      "step": 11739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.245707115039228e-05,
+      "loss": 2.0703,
+      "step": 11740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2454826038032604e-05,
+      "loss": 2.094,
+      "step": 11741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.245258101208504e-05,
+      "loss": 2.0906,
+      "step": 11742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2450336072608467e-05,
+      "loss": 1.9741,
+      "step": 11743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.244809121966175e-05,
+      "loss": 1.9602,
+      "step": 11744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2445846453303764e-05,
+      "loss": 2.0559,
+      "step": 11745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2443601773593365e-05,
+      "loss": 2.061,
+      "step": 11746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2441357180589432e-05,
+      "loss": 2.0439,
+      "step": 11747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2439112674350809e-05,
+      "loss": 2.0288,
+      "step": 11748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2436868254936373e-05,
+      "loss": 1.9878,
+      "step": 11749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2434623922404972e-05,
+      "loss": 1.9688,
+      "step": 11750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2432379676815457e-05,
+      "loss": 2.0662,
+      "step": 11751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2430135518226694e-05,
+      "loss": 2.0554,
+      "step": 11752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2427891446697518e-05,
+      "loss": 2.0605,
+      "step": 11753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2425647462286791e-05,
+      "loss": 2.032,
+      "step": 11754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2423403565053353e-05,
+      "loss": 2.0981,
+      "step": 11755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2421159755056042e-05,
+      "loss": 2.0911,
+      "step": 11756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2418916032353709e-05,
+      "loss": 2.0537,
+      "step": 11757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2416672397005184e-05,
+      "loss": 2.0762,
+      "step": 11758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2414428849069314e-05,
+      "loss": 2.0608,
+      "step": 11759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2412185388604923e-05,
+      "loss": 2.0244,
+      "step": 11760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.240994201567085e-05,
+      "loss": 2.0525,
+      "step": 11761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2407698730325921e-05,
+      "loss": 2.0383,
+      "step": 11762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2405455532628968e-05,
+      "loss": 2.1343,
+      "step": 11763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.240321242263881e-05,
+      "loss": 2.1531,
+      "step": 11764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2400969400414279e-05,
+      "loss": 2.042,
+      "step": 11765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2398726466014185e-05,
+      "loss": 2.0164,
+      "step": 11766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2396483619497356e-05,
+      "loss": 2.0615,
+      "step": 11767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2394240860922597e-05,
+      "loss": 2.084,
+      "step": 11768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2391998190348736e-05,
+      "loss": 2.0217,
+      "step": 11769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2389755607834576e-05,
+      "loss": 2.0581,
+      "step": 11770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2387513113438921e-05,
+      "loss": 2.0618,
+      "step": 11771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2385270707220592e-05,
+      "loss": 2.074,
+      "step": 11772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2383028389238375e-05,
+      "loss": 2.0969,
+      "step": 11773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.238078615955109e-05,
+      "loss": 2.1079,
+      "step": 11774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2378544018217526e-05,
+      "loss": 2.0918,
+      "step": 11775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2376301965296487e-05,
+      "loss": 2.0925,
+      "step": 11776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2374060000846763e-05,
+      "loss": 2.0554,
+      "step": 11777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2371818124927148e-05,
+      "loss": 2.0347,
+      "step": 11778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2369576337596435e-05,
+      "loss": 2.0005,
+      "step": 11779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2367334638913414e-05,
+      "loss": 2.0298,
+      "step": 11780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.236509302893686e-05,
+      "loss": 2.0593,
+      "step": 11781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2362851507725571e-05,
+      "loss": 2.0129,
+      "step": 11782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2360610075338316e-05,
+      "loss": 2.0488,
+      "step": 11783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2358368731833886e-05,
+      "loss": 2.05,
+      "step": 11784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.235612747727105e-05,
+      "loss": 2.0105,
+      "step": 11785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.235388631170858e-05,
+      "loss": 2.0098,
+      "step": 11786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2351645235205256e-05,
+      "loss": 2.0793,
+      "step": 11787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2349404247819838e-05,
+      "loss": 2.0828,
+      "step": 11788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2347163349611104e-05,
+      "loss": 2.0432,
+      "step": 11789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2344922540637816e-05,
+      "loss": 2.093,
+      "step": 11790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2342681820958725e-05,
+      "loss": 2.1211,
+      "step": 11791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2340441190632607e-05,
+      "loss": 2.1196,
+      "step": 11792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2338200649718206e-05,
+      "loss": 2.0981,
+      "step": 11793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2335960198274291e-05,
+      "loss": 2.083,
+      "step": 11794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2333719836359611e-05,
+      "loss": 1.9954,
+      "step": 11795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2331479564032908e-05,
+      "loss": 1.9829,
+      "step": 11796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2329239381352943e-05,
+      "loss": 2.0977,
+      "step": 11797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2326999288378455e-05,
+      "loss": 2.0945,
+      "step": 11798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2324759285168187e-05,
+      "loss": 2.0764,
+      "step": 11799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.232251937178089e-05,
+      "loss": 2.0557,
+      "step": 11800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.232027954827529e-05,
+      "loss": 2.0837,
+      "step": 11801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2318039814710132e-05,
+      "loss": 2.0735,
+      "step": 11802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2315800171144147e-05,
+      "loss": 2.0774,
+      "step": 11803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2313560617636072e-05,
+      "loss": 2.0647,
+      "step": 11804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2311321154244635e-05,
+      "loss": 2.0725,
+      "step": 11805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2309081781028552e-05,
+      "loss": 2.0647,
+      "step": 11806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2306842498046566e-05,
+      "loss": 1.9937,
+      "step": 11807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2304603305357384e-05,
+      "loss": 2.0059,
+      "step": 11808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2302364203019741e-05,
+      "loss": 2.0476,
+      "step": 11809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2300125191092342e-05,
+      "loss": 2.0073,
+      "step": 11810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.229788626963391e-05,
+      "loss": 2.0833,
+      "step": 11811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2295647438703153e-05,
+      "loss": 2.0933,
+      "step": 11812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2293408698358787e-05,
+      "loss": 2.1104,
+      "step": 11813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2291170048659515e-05,
+      "loss": 2.1111,
+      "step": 11814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2288931489664051e-05,
+      "loss": 2.0298,
+      "step": 11815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2286693021431089e-05,
+      "loss": 2.0574,
+      "step": 11816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2284454644019336e-05,
+      "loss": 2.0239,
+      "step": 11817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2282216357487487e-05,
+      "loss": 2.0393,
+      "step": 11818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2279978161894247e-05,
+      "loss": 2.0767,
+      "step": 11819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2277740057298296e-05,
+      "loss": 2.1084,
+      "step": 11820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2275502043758342e-05,
+      "loss": 2.0601,
+      "step": 11821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2273264121333063e-05,
+      "loss": 2.0603,
+      "step": 11822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2271026290081147e-05,
+      "loss": 1.9834,
+      "step": 11823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2268788550061284e-05,
+      "loss": 1.9949,
+      "step": 11824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2266550901332147e-05,
+      "loss": 2.0833,
+      "step": 11825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2264313343952429e-05,
+      "loss": 2.1079,
+      "step": 11826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2262075877980798e-05,
+      "loss": 2.0129,
+      "step": 11827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2259838503475924e-05,
+      "loss": 1.9978,
+      "step": 11828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2257601220496493e-05,
+      "loss": 2.0759,
+      "step": 11829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2255364029101165e-05,
+      "loss": 2.0806,
+      "step": 11830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2253126929348614e-05,
+      "loss": 2.1208,
+      "step": 11831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2250889921297501e-05,
+      "loss": 2.1328,
+      "step": 11832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.224865300500649e-05,
+      "loss": 2.073,
+      "step": 11833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2246416180534241e-05,
+      "loss": 2.0698,
+      "step": 11834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2244179447939411e-05,
+      "loss": 2.0903,
+      "step": 11835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2241942807280663e-05,
+      "loss": 2.0808,
+      "step": 11836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2239706258616647e-05,
+      "loss": 1.9973,
+      "step": 11837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2237469802006007e-05,
+      "loss": 1.9963,
+      "step": 11838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.22352334375074e-05,
+      "loss": 2.0757,
+      "step": 11839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2232997165179463e-05,
+      "loss": 2.022,
+      "step": 11840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2230760985080851e-05,
+      "loss": 2.0723,
+      "step": 11841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.22285248972702e-05,
+      "loss": 2.0454,
+      "step": 11842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.222628890180614e-05,
+      "loss": 2.084,
+      "step": 11843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2224052998747325e-05,
+      "loss": 2.0632,
+      "step": 11844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.222181718815237e-05,
+      "loss": 2.022,
+      "step": 11845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2219581470079925e-05,
+      "loss": 2.0747,
+      "step": 11846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2217345844588607e-05,
+      "loss": 2.1003,
+      "step": 11847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2215110311737047e-05,
+      "loss": 2.1123,
+      "step": 11848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2212874871583867e-05,
+      "loss": 2.0486,
+      "step": 11849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.221063952418769e-05,
+      "loss": 2.041,
+      "step": 11850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2208404269607135e-05,
+      "loss": 2.0706,
+      "step": 11851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2206169107900822e-05,
+      "loss": 2.0698,
+      "step": 11852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2203934039127359e-05,
+      "loss": 2.0696,
+      "step": 11853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2201699063345367e-05,
+      "loss": 2.0613,
+      "step": 11854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2199464180613442e-05,
+      "loss": 2.0422,
+      "step": 11855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2197229390990209e-05,
+      "loss": 2.0225,
+      "step": 11856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2194994694534263e-05,
+      "loss": 2.105,
+      "step": 11857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.21927600913042e-05,
+      "loss": 2.1121,
+      "step": 11858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2190525581358634e-05,
+      "loss": 1.9983,
+      "step": 11859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2188291164756147e-05,
+      "loss": 1.9297,
+      "step": 11860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2186056841555351e-05,
+      "loss": 1.9854,
+      "step": 11861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2183822611814829e-05,
+      "loss": 1.9922,
+      "step": 11862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2181588475593165e-05,
+      "loss": 2.0522,
+      "step": 11863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.217935443294896e-05,
+      "loss": 2.0715,
+      "step": 11864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2177120483940786e-05,
+      "loss": 2.0552,
+      "step": 11865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.217488662862724e-05,
+      "loss": 2.0273,
+      "step": 11866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2172652867066889e-05,
+      "loss": 2.0803,
+      "step": 11867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2170419199318318e-05,
+      "loss": 2.0864,
+      "step": 11868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2168185625440101e-05,
+      "loss": 2.009,
+      "step": 11869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2165952145490808e-05,
+      "loss": 2.0266,
+      "step": 11870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2163718759529015e-05,
+      "loss": 2.041,
+      "step": 11871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2161485467613291e-05,
+      "loss": 2.0425,
+      "step": 11872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.215925226980219e-05,
+      "loss": 2.075,
+      "step": 11873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2157019166154289e-05,
+      "loss": 2.0984,
+      "step": 11874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2154786156728133e-05,
+      "loss": 2.1084,
+      "step": 11875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.21525532415823e-05,
+      "loss": 2.1096,
+      "step": 11876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2150320420775332e-05,
+      "loss": 2.012,
+      "step": 11877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.214808769436578e-05,
+      "loss": 2.0281,
+      "step": 11878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2145855062412205e-05,
+      "loss": 2.0237,
+      "step": 11879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2143622524973143e-05,
+      "loss": 2.0261,
+      "step": 11880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2141390082107154e-05,
+      "loss": 2.0371,
+      "step": 11881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.213915773387277e-05,
+      "loss": 2.0557,
+      "step": 11882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2136925480328536e-05,
+      "loss": 2.0942,
+      "step": 11883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2134693321532989e-05,
+      "loss": 2.0938,
+      "step": 11884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2132461257544667e-05,
+      "loss": 2.0212,
+      "step": 11885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2130229288422098e-05,
+      "loss": 2.0354,
+      "step": 11886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2127997414223824e-05,
+      "loss": 2.0613,
+      "step": 11887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2125765635008357e-05,
+      "loss": 2.0515,
+      "step": 11888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2123533950834238e-05,
+      "loss": 2.052,
+      "step": 11889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.212130236175998e-05,
+      "loss": 2.0518,
+      "step": 11890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2119070867844112e-05,
+      "loss": 2.0969,
+      "step": 11891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.211683946914515e-05,
+      "loss": 2.0823,
+      "step": 11892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2114608165721599e-05,
+      "loss": 2.0803,
+      "step": 11893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.211237695763199e-05,
+      "loss": 2.0901,
+      "step": 11894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2110145844934817e-05,
+      "loss": 2.0056,
+      "step": 11895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2107914827688603e-05,
+      "loss": 1.9785,
+      "step": 11896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.210568390595185e-05,
+      "loss": 2.1086,
+      "step": 11897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2103453079783051e-05,
+      "loss": 2.1187,
+      "step": 11898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2101222349240719e-05,
+      "loss": 2.0471,
+      "step": 11899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2098991714383343e-05,
+      "loss": 2.0344,
+      "step": 11900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2096761175269432e-05,
+      "loss": 2.0754,
+      "step": 11901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2094530731957466e-05,
+      "loss": 2.0632,
+      "step": 11902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2092300384505942e-05,
+      "loss": 2.0325,
+      "step": 11903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2090070132973345e-05,
+      "loss": 2.0686,
+      "step": 11904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.208783997741816e-05,
+      "loss": 2.0449,
+      "step": 11905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2085609917898881e-05,
+      "loss": 2.071,
+      "step": 11906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.208337995447398e-05,
+      "loss": 2.0361,
+      "step": 11907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2081150087201929e-05,
+      "loss": 2.04,
+      "step": 11908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2078920316141217e-05,
+      "loss": 2.1289,
+      "step": 11909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2076690641350307e-05,
+      "loss": 2.1057,
+      "step": 11910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2074461062887678e-05,
+      "loss": 2.0957,
+      "step": 11911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2072231580811792e-05,
+      "loss": 2.0964,
+      "step": 11912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2070002195181111e-05,
+      "loss": 2.0991,
+      "step": 11913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.206777290605411e-05,
+      "loss": 2.1272,
+      "step": 11914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2065543713489237e-05,
+      "loss": 2.0771,
+      "step": 11915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.206331461754496e-05,
+      "loss": 2.0542,
+      "step": 11916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.206108561827973e-05,
+      "loss": 2.073,
+      "step": 11917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2058856715751997e-05,
+      "loss": 2.0251,
+      "step": 11918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2056627910020218e-05,
+      "loss": 2.0715,
+      "step": 11919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2054399201142834e-05,
+      "loss": 2.1062,
+      "step": 11920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2052170589178296e-05,
+      "loss": 2.1204,
+      "step": 11921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2049942074185046e-05,
+      "loss": 2.1025,
+      "step": 11922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2047713656221515e-05,
+      "loss": 2.0667,
+      "step": 11923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2045485335346159e-05,
+      "loss": 2.052,
+      "step": 11924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2043257111617393e-05,
+      "loss": 2.0781,
+      "step": 11925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2041028985093665e-05,
+      "loss": 2.1013,
+      "step": 11926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.20388009558334e-05,
+      "loss": 2.0913,
+      "step": 11927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2036573023895018e-05,
+      "loss": 2.0891,
+      "step": 11928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2034345189336957e-05,
+      "loss": 2.1167,
+      "step": 11929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2032117452217625e-05,
+      "loss": 2.1213,
+      "step": 11930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2029889812595457e-05,
+      "loss": 2.1206,
+      "step": 11931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2027662270528865e-05,
+      "loss": 2.1282,
+      "step": 11932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2025434826076252e-05,
+      "loss": 2.0535,
+      "step": 11933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2023207479296049e-05,
+      "loss": 2.0408,
+      "step": 11934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2020980230246648e-05,
+      "loss": 2.0364,
+      "step": 11935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.201875307898647e-05,
+      "loss": 2.031,
+      "step": 11936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.201652602557391e-05,
+      "loss": 2.063,
+      "step": 11937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2014299070067377e-05,
+      "loss": 2.0752,
+      "step": 11938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2012072212525267e-05,
+      "loss": 2.0969,
+      "step": 11939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2009845453005973e-05,
+      "loss": 2.0784,
+      "step": 11940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.20076187915679e-05,
+      "loss": 2.0654,
+      "step": 11941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2005392228269423e-05,
+      "loss": 2.0823,
+      "step": 11942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2003165763168948e-05,
+      "loss": 2.033,
+      "step": 11943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2000939396324857e-05,
+      "loss": 2.0037,
+      "step": 11944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1998713127795523e-05,
+      "loss": 2.0488,
+      "step": 11945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.199648695763934e-05,
+      "loss": 2.0669,
+      "step": 11946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1994260885914677e-05,
+      "loss": 2.0391,
+      "step": 11947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1992034912679923e-05,
+      "loss": 2.0173,
+      "step": 11948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.198980903799344e-05,
+      "loss": 2.0601,
+      "step": 11949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1987583261913599e-05,
+      "loss": 2.0801,
+      "step": 11950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1985357584498778e-05,
+      "loss": 2.0815,
+      "step": 11951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.198313200580733e-05,
+      "loss": 2.0779,
+      "step": 11952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1980906525897634e-05,
+      "loss": 2.075,
+      "step": 11953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1978681144828037e-05,
+      "loss": 2.0764,
+      "step": 11954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1976455862656902e-05,
+      "loss": 2.0762,
+      "step": 11955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1974230679442581e-05,
+      "loss": 2.0354,
+      "step": 11956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1972005595243436e-05,
+      "loss": 2.042,
+      "step": 11957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1969780610117807e-05,
+      "loss": 2.0544,
+      "step": 11958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1967555724124049e-05,
+      "loss": 2.0679,
+      "step": 11959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1965330937320502e-05,
+      "loss": 2.0588,
+      "step": 11960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1963106249765514e-05,
+      "loss": 2.1113,
+      "step": 11961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1960881661517415e-05,
+      "loss": 2.115,
+      "step": 11962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1958657172634556e-05,
+      "loss": 2.0762,
+      "step": 11963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1956432783175263e-05,
+      "loss": 2.0728,
+      "step": 11964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1954208493197863e-05,
+      "loss": 2.0239,
+      "step": 11965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1951984302760698e-05,
+      "loss": 2.0564,
+      "step": 11966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1949760211922083e-05,
+      "loss": 2.0225,
+      "step": 11967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1947536220740354e-05,
+      "loss": 2.042,
+      "step": 11968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1945312329273825e-05,
+      "loss": 2.0708,
+      "step": 11969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.194308853758081e-05,
+      "loss": 2.0654,
+      "step": 11970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.194086484571964e-05,
+      "loss": 2.0823,
+      "step": 11971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1938641253748613e-05,
+      "loss": 2.0818,
+      "step": 11972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1936417761726055e-05,
+      "loss": 2.0562,
+      "step": 11973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1934194369710261e-05,
+      "loss": 2.0645,
+      "step": 11974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1931971077759543e-05,
+      "loss": 2.0737,
+      "step": 11975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1929747885932206e-05,
+      "loss": 2.0891,
+      "step": 11976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1927524794286547e-05,
+      "loss": 2.0479,
+      "step": 11977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.192530180288087e-05,
+      "loss": 2.0474,
+      "step": 11978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1923078911773467e-05,
+      "loss": 2.0745,
+      "step": 11979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1920856121022623e-05,
+      "loss": 2.0859,
+      "step": 11980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1918633430686641e-05,
+      "loss": 2.0361,
+      "step": 11981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1916410840823798e-05,
+      "loss": 2.0676,
+      "step": 11982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1914188351492387e-05,
+      "loss": 2.1042,
+      "step": 11983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1911965962750688e-05,
+      "loss": 2.1311,
+      "step": 11984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1909743674656973e-05,
+      "loss": 2.0718,
+      "step": 11985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.190752148726953e-05,
+      "loss": 2.0815,
+      "step": 11986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1905299400646623e-05,
+      "loss": 2.0454,
+      "step": 11987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1903077414846534e-05,
+      "loss": 2.0574,
+      "step": 11988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1900855529927525e-05,
+      "loss": 2.0818,
+      "step": 11989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1898633745947864e-05,
+      "loss": 2.0828,
+      "step": 11990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1896412062965815e-05,
+      "loss": 2.0427,
+      "step": 11991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.189419048103964e-05,
+      "loss": 2.0391,
+      "step": 11992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1891969000227594e-05,
+      "loss": 2.0669,
+      "step": 11993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1889747620587941e-05,
+      "loss": 2.0786,
+      "step": 11994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.188752634217892e-05,
+      "loss": 2.0679,
+      "step": 11995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1885305165058796e-05,
+      "loss": 2.0684,
+      "step": 11996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1883084089285805e-05,
+      "loss": 2.0137,
+      "step": 11997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1880863114918204e-05,
+      "loss": 2.0149,
+      "step": 11998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1878642242014232e-05,
+      "loss": 2.0151,
+      "step": 11999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1876421470632117e-05,
+      "loss": 1.983,
+      "step": 12000
+    },
+    {
+      "Bleu_1": 0.7730527471064705,
+      "Bleu_2": 0.6117181153474853,
+      "Bleu_3": 0.4695419354077647,
+      "Bleu_4": 0.3562399591551594,
+      "CIDEr": 1.2036558759141944,
+      "METEOR": 0.2776531456615971,
+      "ROUGE_L": 0.5590537713820831,
+      "epoch": 0.6,
+      "step": 12000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.187420080083011e-05,
+      "loss": 2.0852,
+      "step": 12001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1871980232666435e-05,
+      "loss": 2.0918,
+      "step": 12002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1869759766199337e-05,
+      "loss": 2.0957,
+      "step": 12003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1867539401487037e-05,
+      "loss": 2.1064,
+      "step": 12004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1865319138587756e-05,
+      "loss": 2.0603,
+      "step": 12005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1863098977559731e-05,
+      "loss": 2.0649,
+      "step": 12006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1860878918461167e-05,
+      "loss": 2.0532,
+      "step": 12007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1858658961350298e-05,
+      "loss": 2.0725,
+      "step": 12008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1856439106285336e-05,
+      "loss": 2.1116,
+      "step": 12009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1854219353324484e-05,
+      "loss": 2.1116,
+      "step": 12010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1851999702525965e-05,
+      "loss": 2.0752,
+      "step": 12011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1849780153947977e-05,
+      "loss": 2.0774,
+      "step": 12012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1847560707648736e-05,
+      "loss": 2.0613,
+      "step": 12013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1845341363686439e-05,
+      "loss": 2.0537,
+      "step": 12014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1843122122119278e-05,
+      "loss": 2.1108,
+      "step": 12015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1840902983005462e-05,
+      "loss": 2.0874,
+      "step": 12016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1838683946403176e-05,
+      "loss": 2.052,
+      "step": 12017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1836465012370622e-05,
+      "loss": 2.0613,
+      "step": 12018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1834246180965982e-05,
+      "loss": 2.0886,
+      "step": 12019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1832027452247435e-05,
+      "loss": 2.0962,
+      "step": 12020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1829808826273181e-05,
+      "loss": 2.0513,
+      "step": 12021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1827590303101387e-05,
+      "loss": 2.0603,
+      "step": 12022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1825371882790243e-05,
+      "loss": 2.0642,
+      "step": 12023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1823153565397915e-05,
+      "loss": 2.0725,
+      "step": 12024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1820935350982577e-05,
+      "loss": 2.0518,
+      "step": 12025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1818717239602404e-05,
+      "loss": 2.0532,
+      "step": 12026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.181649923131556e-05,
+      "loss": 2.0798,
+      "step": 12027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.181428132618021e-05,
+      "loss": 2.1104,
+      "step": 12028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1812063524254518e-05,
+      "loss": 2.0823,
+      "step": 12029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1809845825596639e-05,
+      "loss": 2.0745,
+      "step": 12030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1807628230264735e-05,
+      "loss": 2.0315,
+      "step": 12031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1805410738316952e-05,
+      "loss": 2.0537,
+      "step": 12032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1803193349811454e-05,
+      "loss": 2.0532,
+      "step": 12033
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.180097606480638e-05,
+      "loss": 2.0664,
+      "step": 12034
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1798758883359873e-05,
+      "loss": 2.0591,
+      "step": 12035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1796541805530085e-05,
+      "loss": 2.0933,
+      "step": 12036
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1794324831375146e-05,
+      "loss": 2.0972,
+      "step": 12037
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1792107960953205e-05,
+      "loss": 2.0691,
+      "step": 12038
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.178989119432239e-05,
+      "loss": 2.0291,
+      "step": 12039
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.178767453154083e-05,
+      "loss": 2.0618,
+      "step": 12040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1785457972666663e-05,
+      "loss": 2.103,
+      "step": 12041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1783241517758005e-05,
+      "loss": 2.0901,
+      "step": 12042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1781025166872992e-05,
+      "loss": 2.074,
+      "step": 12043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.177880892006974e-05,
+      "loss": 2.0642,
+      "step": 12044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1776592777406363e-05,
+      "loss": 2.083,
+      "step": 12045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1774376738940985e-05,
+      "loss": 2.067,
+      "step": 12046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.177216080473171e-05,
+      "loss": 2.1245,
+      "step": 12047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.176994497483666e-05,
+      "loss": 2.155,
+      "step": 12048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1767729249313935e-05,
+      "loss": 2.085,
+      "step": 12049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1765513628221635e-05,
+      "loss": 2.1008,
+      "step": 12050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1763298111617873e-05,
+      "loss": 1.9956,
+      "step": 12051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1761082699560737e-05,
+      "loss": 1.9707,
+      "step": 12052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1758867392108333e-05,
+      "loss": 2.0779,
+      "step": 12053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1756652189318754e-05,
+      "loss": 2.0925,
+      "step": 12054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1754437091250084e-05,
+      "loss": 2.0835,
+      "step": 12055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1752222097960422e-05,
+      "loss": 2.0801,
+      "step": 12056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.175000720950784e-05,
+      "loss": 1.9958,
+      "step": 12057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1747792425950434e-05,
+      "loss": 2.0098,
+      "step": 12058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1745577747346274e-05,
+      "loss": 2.0715,
+      "step": 12059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1743363173753446e-05,
+      "loss": 2.0339,
+      "step": 12060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.174114870523002e-05,
+      "loss": 2.0366,
+      "step": 12061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1738934341834066e-05,
+      "loss": 2.0376,
+      "step": 12062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1736720083623657e-05,
+      "loss": 2.0845,
+      "step": 12063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1734505930656857e-05,
+      "loss": 2.0737,
+      "step": 12064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1732291882991732e-05,
+      "loss": 2.1287,
+      "step": 12065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1730077940686345e-05,
+      "loss": 2.1416,
+      "step": 12066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1727864103798743e-05,
+      "loss": 2.0352,
+      "step": 12067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1725650372386993e-05,
+      "loss": 2.0537,
+      "step": 12068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.172343674650914e-05,
+      "loss": 2.0696,
+      "step": 12069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1721223226223242e-05,
+      "loss": 2.0762,
+      "step": 12070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1719009811587341e-05,
+      "loss": 2.0984,
+      "step": 12071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1716796502659477e-05,
+      "loss": 2.0774,
+      "step": 12072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1714583299497699e-05,
+      "loss": 2.0923,
+      "step": 12073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.171237020216004e-05,
+      "loss": 2.0679,
+      "step": 12074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.171015721070454e-05,
+      "loss": 2.0867,
+      "step": 12075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1707944325189236e-05,
+      "loss": 2.0735,
+      "step": 12076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.170573154567214e-05,
+      "loss": 2.0242,
+      "step": 12077
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1703518872211301e-05,
+      "loss": 1.996,
+      "step": 12078
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1701306304864733e-05,
+      "loss": 2.0417,
+      "step": 12079
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1699093843690459e-05,
+      "loss": 2.0149,
+      "step": 12080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1696881488746504e-05,
+      "loss": 2.0928,
+      "step": 12081
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1694669240090872e-05,
+      "loss": 2.0784,
+      "step": 12082
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1692457097781593e-05,
+      "loss": 2.0415,
+      "step": 12083
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1690245061876663e-05,
+      "loss": 2.0457,
+      "step": 12084
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1688033132434102e-05,
+      "loss": 2.0396,
+      "step": 12085
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1685821309511909e-05,
+      "loss": 2.0696,
+      "step": 12086
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1683609593168083e-05,
+      "loss": 2.0198,
+      "step": 12087
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1681397983460634e-05,
+      "loss": 2.0222,
+      "step": 12088
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1679186480447543e-05,
+      "loss": 2.1042,
+      "step": 12089
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1676975084186823e-05,
+      "loss": 2.1106,
+      "step": 12090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1674763794736458e-05,
+      "loss": 2.061,
+      "step": 12091
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1672552612154427e-05,
+      "loss": 2.0486,
+      "step": 12092
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1670341536498729e-05,
+      "loss": 2.0503,
+      "step": 12093
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1668130567827335e-05,
+      "loss": 2.0522,
+      "step": 12094
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1665919706198238e-05,
+      "loss": 2.0681,
+      "step": 12095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1663708951669406e-05,
+      "loss": 2.0436,
+      "step": 12096
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1661498304298815e-05,
+      "loss": 2.0103,
+      "step": 12097
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1659287764144439e-05,
+      "loss": 2.0293,
+      "step": 12098
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.1657077331264244e-05,
+      "loss": 2.0947,
+      "step": 12099
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.16548670057162e-05,
+      "loss": 2.0811,
+      "step": 12100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1652656787558269e-05,
+      "loss": 2.0845,
+      "step": 12101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1650446676848406e-05,
+      "loss": 2.1067,
+      "step": 12102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.164823667364458e-05,
+      "loss": 2.0574,
+      "step": 12103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.164602677800473e-05,
+      "loss": 2.0051,
+      "step": 12104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1643816989986824e-05,
+      "loss": 2.0825,
+      "step": 12105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1641607309648804e-05,
+      "loss": 2.0703,
+      "step": 12106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1639397737048608e-05,
+      "loss": 2.041,
+      "step": 12107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1637188272244195e-05,
+      "loss": 2.0845,
+      "step": 12108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1634978915293494e-05,
+      "loss": 2.0615,
+      "step": 12109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1632769666254451e-05,
+      "loss": 2.0823,
+      "step": 12110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1630560525184995e-05,
+      "loss": 2.0859,
+      "step": 12111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1628351492143059e-05,
+      "loss": 2.113,
+      "step": 12112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1626142567186575e-05,
+      "loss": 2.0769,
+      "step": 12113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1623933750373467e-05,
+      "loss": 2.0923,
+      "step": 12114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1621725041761658e-05,
+      "loss": 2.0454,
+      "step": 12115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1619516441409077e-05,
+      "loss": 2.0237,
+      "step": 12116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1617307949373628e-05,
+      "loss": 2.0332,
+      "step": 12117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.161509956571324e-05,
+      "loss": 2.0691,
+      "step": 12118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1612891290485811e-05,
+      "loss": 2.0591,
+      "step": 12119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1610683123749268e-05,
+      "loss": 2.0428,
+      "step": 12120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1608475065561503e-05,
+      "loss": 2.0537,
+      "step": 12121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1606267115980422e-05,
+      "loss": 2.0579,
+      "step": 12122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1604059275063935e-05,
+      "loss": 2.0493,
+      "step": 12123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1601851542869929e-05,
+      "loss": 2.084,
+      "step": 12124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1599643919456308e-05,
+      "loss": 2.0764,
+      "step": 12125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1597436404880961e-05,
+      "loss": 2.0781,
+      "step": 12126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1595228999201768e-05,
+      "loss": 2.0708,
+      "step": 12127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1593021702476636e-05,
+      "loss": 2.0945,
+      "step": 12128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1590814514763426e-05,
+      "loss": 2.0381,
+      "step": 12129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1588607436120038e-05,
+      "loss": 2.0525,
+      "step": 12130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.158640046660434e-05,
+      "loss": 2.0752,
+      "step": 12131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1584193606274209e-05,
+      "loss": 2.0588,
+      "step": 12132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1581986855187518e-05,
+      "loss": 2.0215,
+      "step": 12133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1579780213402135e-05,
+      "loss": 2.0264,
+      "step": 12134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.157757368097593e-05,
+      "loss": 2.0273,
+      "step": 12135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1575367257966766e-05,
+      "loss": 2.0127,
+      "step": 12136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1573160944432499e-05,
+      "loss": 2.063,
+      "step": 12137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1570954740430995e-05,
+      "loss": 2.0615,
+      "step": 12138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1568748646020099e-05,
+      "loss": 2.0552,
+      "step": 12139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1566542661257675e-05,
+      "loss": 2.0774,
+      "step": 12140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1564336786201566e-05,
+      "loss": 2.1228,
+      "step": 12141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1562131020909614e-05,
+      "loss": 2.1265,
+      "step": 12142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1559925365439672e-05,
+      "loss": 2.0989,
+      "step": 12143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1557719819849573e-05,
+      "loss": 2.0952,
+      "step": 12144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1555514384197162e-05,
+      "loss": 2.0181,
+      "step": 12145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.155330905854027e-05,
+      "loss": 2.0505,
+      "step": 12146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1551103842936723e-05,
+      "loss": 2.0525,
+      "step": 12147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1548898737444362e-05,
+      "loss": 2.0747,
+      "step": 12148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1546693742121005e-05,
+      "loss": 2.0437,
+      "step": 12149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.154448885702448e-05,
+      "loss": 2.0659,
+      "step": 12150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1542284082212609e-05,
+      "loss": 2.0854,
+      "step": 12151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.15400794177432e-05,
+      "loss": 2.1262,
+      "step": 12152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.153787486367408e-05,
+      "loss": 2.0254,
+      "step": 12153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.153567042006305e-05,
+      "loss": 1.9939,
+      "step": 12154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.153346608696793e-05,
+      "loss": 2.0464,
+      "step": 12155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1531261864446519e-05,
+      "loss": 2.0188,
+      "step": 12156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1529057752556617e-05,
+      "loss": 2.0679,
+      "step": 12157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1526853751356034e-05,
+      "loss": 2.0688,
+      "step": 12158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1524649860902555e-05,
+      "loss": 2.0569,
+      "step": 12159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.152244608125399e-05,
+      "loss": 2.0503,
+      "step": 12160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.152024241246812e-05,
+      "loss": 2.0325,
+      "step": 12161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.151803885460273e-05,
+      "loss": 2.0225,
+      "step": 12162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1515835407715616e-05,
+      "loss": 2.0222,
+      "step": 12163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1513632071864552e-05,
+      "loss": 2.021,
+      "step": 12164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1511428847107328e-05,
+      "loss": 2.1282,
+      "step": 12165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1509225733501713e-05,
+      "loss": 2.1184,
+      "step": 12166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1507022731105482e-05,
+      "loss": 2.0891,
+      "step": 12167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.150481983997641e-05,
+      "loss": 2.073,
+      "step": 12168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.150261706017226e-05,
+      "loss": 2.0457,
+      "step": 12169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1500414391750803e-05,
+      "loss": 2.0603,
+      "step": 12170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1498211834769799e-05,
+      "loss": 2.0708,
+      "step": 12171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1496009389287e-05,
+      "loss": 2.0676,
+      "step": 12172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1493807055360179e-05,
+      "loss": 2.0203,
+      "step": 12173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1491604833047074e-05,
+      "loss": 2.0271,
+      "step": 12174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1489402722405448e-05,
+      "loss": 2.0515,
+      "step": 12175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.148720072349304e-05,
+      "loss": 2.0767,
+      "step": 12176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1484998836367593e-05,
+      "loss": 2.1099,
+      "step": 12177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.148279706108686e-05,
+      "loss": 2.0918,
+      "step": 12178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1480595397708568e-05,
+      "loss": 2.0938,
+      "step": 12179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1478393846290466e-05,
+      "loss": 2.106,
+      "step": 12180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1476192406890272e-05,
+      "loss": 2.0422,
+      "step": 12181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.147399107956573e-05,
+      "loss": 2.0737,
+      "step": 12182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1471789864374558e-05,
+      "loss": 2.0947,
+      "step": 12183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1469588761374486e-05,
+      "loss": 2.1072,
+      "step": 12184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.146738777062323e-05,
+      "loss": 2.0234,
+      "step": 12185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1465186892178512e-05,
+      "loss": 2.0684,
+      "step": 12186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1462986126098048e-05,
+      "loss": 2.0632,
+      "step": 12187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1460785472439551e-05,
+      "loss": 2.063,
+      "step": 12188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1458584931260725e-05,
+      "loss": 2.1125,
+      "step": 12189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1456384502619285e-05,
+      "loss": 2.1062,
+      "step": 12190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1454184186572924e-05,
+      "loss": 2.105,
+      "step": 12191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1451983983179358e-05,
+      "loss": 2.0879,
+      "step": 12192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1449783892496272e-05,
+      "loss": 2.0388,
+      "step": 12193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1447583914581362e-05,
+      "loss": 2.0547,
+      "step": 12194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1445384049492327e-05,
+      "loss": 2.0908,
+      "step": 12195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1443184297286844e-05,
+      "loss": 2.1145,
+      "step": 12196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1440984658022614e-05,
+      "loss": 2.062,
+      "step": 12197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1438785131757314e-05,
+      "loss": 2.042,
+      "step": 12198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1436585718548612e-05,
+      "loss": 2.0378,
+      "step": 12199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1434386418454207e-05,
+      "loss": 2.0085,
+      "step": 12200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1432187231531752e-05,
+      "loss": 1.9995,
+      "step": 12201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1429988157838934e-05,
+      "loss": 2.0171,
+      "step": 12202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1427789197433411e-05,
+      "loss": 2.02,
+      "step": 12203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1425590350372855e-05,
+      "loss": 1.9946,
+      "step": 12204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1423391616714924e-05,
+      "loss": 2.084,
+      "step": 12205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1421192996517278e-05,
+      "loss": 2.0837,
+      "step": 12206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1418994489837572e-05,
+      "loss": 2.0764,
+      "step": 12207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1416796096733468e-05,
+      "loss": 2.0906,
+      "step": 12208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.14145978172626e-05,
+      "loss": 2.0894,
+      "step": 12209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1412399651482633e-05,
+      "loss": 2.1069,
+      "step": 12210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1410201599451196e-05,
+      "loss": 2.1001,
+      "step": 12211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1408003661225941e-05,
+      "loss": 2.0952,
+      "step": 12212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1405805836864507e-05,
+      "loss": 2.1047,
+      "step": 12213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1403608126424518e-05,
+      "loss": 2.0945,
+      "step": 12214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1401410529963617e-05,
+      "loss": 2.0537,
+      "step": 12215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1399213047539427e-05,
+      "loss": 2.0623,
+      "step": 12216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1397015679209581e-05,
+      "loss": 2.1226,
+      "step": 12217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1394818425031696e-05,
+      "loss": 2.1494,
+      "step": 12218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1392621285063398e-05,
+      "loss": 2.0747,
+      "step": 12219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1390424259362299e-05,
+      "loss": 2.0581,
+      "step": 12220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1388227347986018e-05,
+      "loss": 2.0623,
+      "step": 12221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1386030550992165e-05,
+      "loss": 2.0825,
+      "step": 12222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.138383386843835e-05,
+      "loss": 2.0784,
+      "step": 12223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.138163730038217e-05,
+      "loss": 2.1052,
+      "step": 12224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1379440846881241e-05,
+      "loss": 2.0535,
+      "step": 12225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1377244507993149e-05,
+      "loss": 2.0591,
+      "step": 12226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1375048283775502e-05,
+      "loss": 2.0798,
+      "step": 12227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.137285217428589e-05,
+      "loss": 2.0808,
+      "step": 12228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1370656179581894e-05,
+      "loss": 2.1045,
+      "step": 12229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1368460299721116e-05,
+      "loss": 2.1091,
+      "step": 12230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1366264534761129e-05,
+      "loss": 2.0322,
+      "step": 12231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1364068884759525e-05,
+      "loss": 2.0164,
+      "step": 12232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1361873349773875e-05,
+      "loss": 2.0581,
+      "step": 12233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1359677929861749e-05,
+      "loss": 2.0725,
+      "step": 12234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1357482625080732e-05,
+      "loss": 2.0964,
+      "step": 12235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1355287435488384e-05,
+      "loss": 2.0955,
+      "step": 12236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1353092361142276e-05,
+      "loss": 2.0554,
+      "step": 12237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.135089740209997e-05,
+      "loss": 2.0627,
+      "step": 12238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1348702558419028e-05,
+      "loss": 2.0669,
+      "step": 12239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1346507830157003e-05,
+      "loss": 2.0332,
+      "step": 12240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.134431321737145e-05,
+      "loss": 2.1001,
+      "step": 12241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1342118720119924e-05,
+      "loss": 2.1155,
+      "step": 12242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1339924338459973e-05,
+      "loss": 2.0613,
+      "step": 12243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1337730072449134e-05,
+      "loss": 2.04,
+      "step": 12244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1335535922144964e-05,
+      "loss": 2.0288,
+      "step": 12245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1333341887604983e-05,
+      "loss": 2.0476,
+      "step": 12246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1331147968886745e-05,
+      "loss": 1.988,
+      "step": 12247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1328954166047775e-05,
+      "loss": 2.0054,
+      "step": 12248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1326760479145597e-05,
+      "loss": 2.064,
+      "step": 12249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.132456690823775e-05,
+      "loss": 2.0972,
+      "step": 12250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1322373453381747e-05,
+      "loss": 2.0613,
+      "step": 12251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1320180114635117e-05,
+      "loss": 1.9869,
+      "step": 12252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1317986892055373e-05,
+      "loss": 2.0413,
+      "step": 12253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.131579378570003e-05,
+      "loss": 2.0723,
+      "step": 12254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1313600795626604e-05,
+      "loss": 2.0149,
+      "step": 12255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1311407921892598e-05,
+      "loss": 2.0261,
+      "step": 12256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.130921516455552e-05,
+      "loss": 2.052,
+      "step": 12257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1307022523672875e-05,
+      "loss": 2.0522,
+      "step": 12258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1304829999302156e-05,
+      "loss": 2.0952,
+      "step": 12259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1302637591500868e-05,
+      "loss": 2.0813,
+      "step": 12260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1300445300326495e-05,
+      "loss": 2.1338,
+      "step": 12261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1298253125836536e-05,
+      "loss": 2.1375,
+      "step": 12262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1296061068088474e-05,
+      "loss": 2.0674,
+      "step": 12263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.129386912713979e-05,
+      "loss": 2.0762,
+      "step": 12264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1291677303047973e-05,
+      "loss": 2.094,
+      "step": 12265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.128948559587049e-05,
+      "loss": 2.0962,
+      "step": 12266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1287294005664829e-05,
+      "loss": 2.084,
+      "step": 12267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1285102532488457e-05,
+      "loss": 2.1199,
+      "step": 12268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1282911176398833e-05,
+      "loss": 2.0583,
+      "step": 12269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1280719937453437e-05,
+      "loss": 2.0464,
+      "step": 12270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.127852881570972e-05,
+      "loss": 2.0081,
+      "step": 12271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1276337811225153e-05,
+      "loss": 2.0186,
+      "step": 12272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1274146924057187e-05,
+      "loss": 2.0618,
+      "step": 12273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1271956154263273e-05,
+      "loss": 2.0781,
+      "step": 12274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1269765501900863e-05,
+      "loss": 2.0864,
+      "step": 12275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1267574967027408e-05,
+      "loss": 2.1064,
+      "step": 12276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1265384549700346e-05,
+      "loss": 2.0186,
+      "step": 12277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1263194249977127e-05,
+      "loss": 2.0139,
+      "step": 12278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1261004067915177e-05,
+      "loss": 2.0918,
+      "step": 12279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1258814003571943e-05,
+      "loss": 2.0476,
+      "step": 12280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1256624057004847e-05,
+      "loss": 2.0347,
+      "step": 12281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1254434228271327e-05,
+      "loss": 2.0293,
+      "step": 12282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1252244517428803e-05,
+      "loss": 2.0928,
+      "step": 12283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1250054924534693e-05,
+      "loss": 2.1194,
+      "step": 12284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.124786544964643e-05,
+      "loss": 2.0413,
+      "step": 12285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1245676092821415e-05,
+      "loss": 2.0488,
+      "step": 12286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1243486854117075e-05,
+      "loss": 2.0679,
+      "step": 12287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1241297733590813e-05,
+      "loss": 2.0957,
+      "step": 12288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1239108731300034e-05,
+      "loss": 2.0676,
+      "step": 12289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1236919847302149e-05,
+      "loss": 2.0781,
+      "step": 12290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1234731081654553e-05,
+      "loss": 2.0381,
+      "step": 12291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1232542434414648e-05,
+      "loss": 2.0649,
+      "step": 12292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1230353905639828e-05,
+      "loss": 2.0671,
+      "step": 12293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1228165495387478e-05,
+      "loss": 2.041,
+      "step": 12294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1225977203715e-05,
+      "loss": 2.0168,
+      "step": 12295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1223789030679763e-05,
+      "loss": 2.0374,
+      "step": 12296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1221600976339163e-05,
+      "loss": 1.9976,
+      "step": 12297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1219413040750576e-05,
+      "loss": 1.9854,
+      "step": 12298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.121722522397137e-05,
+      "loss": 2.0701,
+      "step": 12299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1215037526058929e-05,
+      "loss": 2.0735,
+      "step": 12300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1212849947070613e-05,
+      "loss": 2.0869,
+      "step": 12301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.12106624870638e-05,
+      "loss": 2.0774,
+      "step": 12302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1208475146095837e-05,
+      "loss": 2.1284,
+      "step": 12303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1206287924224104e-05,
+      "loss": 2.1404,
+      "step": 12304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.120410082150595e-05,
+      "loss": 2.0413,
+      "step": 12305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1201913837998718e-05,
+      "loss": 2.0466,
+      "step": 12306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.119972697375978e-05,
+      "loss": 2.0513,
+      "step": 12307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1197540228846463e-05,
+      "loss": 2.0688,
+      "step": 12308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1195353603316131e-05,
+      "loss": 2.0171,
+      "step": 12309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1193167097226114e-05,
+      "loss": 2.0525,
+      "step": 12310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1190980710633753e-05,
+      "loss": 2.0835,
+      "step": 12311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1188794443596383e-05,
+      "loss": 2.1133,
+      "step": 12312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1186608296171334e-05,
+      "loss": 2.0608,
+      "step": 12313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1184422268415945e-05,
+      "loss": 2.0515,
+      "step": 12314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1182236360387536e-05,
+      "loss": 2.0603,
+      "step": 12315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1180050572143422e-05,
+      "loss": 2.0615,
+      "step": 12316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1177864903740939e-05,
+      "loss": 2.1021,
+      "step": 12317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1175679355237385e-05,
+      "loss": 2.1001,
+      "step": 12318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.117349392669009e-05,
+      "loss": 2.0549,
+      "step": 12319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1171308618156359e-05,
+      "loss": 2.0718,
+      "step": 12320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1169123429693493e-05,
+      "loss": 2.0762,
+      "step": 12321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1166938361358803e-05,
+      "loss": 2.0981,
+      "step": 12322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1164753413209584e-05,
+      "loss": 2.1045,
+      "step": 12323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1162568585303142e-05,
+      "loss": 2.1035,
+      "step": 12324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1160383877696764e-05,
+      "loss": 2.0522,
+      "step": 12325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1158199290447745e-05,
+      "loss": 2.0708,
+      "step": 12326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1156014823613372e-05,
+      "loss": 2.0691,
+      "step": 12327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1153830477250931e-05,
+      "loss": 2.0776,
+      "step": 12328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1151646251417704e-05,
+      "loss": 2.0928,
+      "step": 12329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1149462146170973e-05,
+      "loss": 2.0933,
+      "step": 12330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1147278161568005e-05,
+      "loss": 2.1028,
+      "step": 12331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1145094297666081e-05,
+      "loss": 2.0945,
+      "step": 12332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.114291055452246e-05,
+      "loss": 2.0791,
+      "step": 12333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1140726932194422e-05,
+      "loss": 2.0752,
+      "step": 12334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1138543430739225e-05,
+      "loss": 1.9937,
+      "step": 12335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.113636005021412e-05,
+      "loss": 2.0039,
+      "step": 12336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1134176790676374e-05,
+      "loss": 2.0225,
+      "step": 12337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1131993652183233e-05,
+      "loss": 2.0354,
+      "step": 12338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1129810634791956e-05,
+      "loss": 2.0869,
+      "step": 12339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1127627738559786e-05,
+      "loss": 2.0793,
+      "step": 12340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1125444963543958e-05,
+      "loss": 2.0278,
+      "step": 12341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1123262309801731e-05,
+      "loss": 2.0652,
+      "step": 12342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1121079777390322e-05,
+      "loss": 2.0581,
+      "step": 12343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1118897366366984e-05,
+      "loss": 2.0334,
+      "step": 12344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1116715076788937e-05,
+      "loss": 2.0486,
+      "step": 12345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1114532908713411e-05,
+      "loss": 2.0432,
+      "step": 12346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1112350862197633e-05,
+      "loss": 2.0896,
+      "step": 12347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1110168937298819e-05,
+      "loss": 2.0735,
+      "step": 12348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1107987134074198e-05,
+      "loss": 2.0593,
+      "step": 12349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1105805452580981e-05,
+      "loss": 2.0518,
+      "step": 12350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.110362389287637e-05,
+      "loss": 2.0732,
+      "step": 12351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.110144245501759e-05,
+      "loss": 2.0471,
+      "step": 12352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1099261139061831e-05,
+      "loss": 2.0317,
+      "step": 12353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1097079945066311e-05,
+      "loss": 2.0535,
+      "step": 12354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.109489887308822e-05,
+      "loss": 2.0759,
+      "step": 12355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1092717923184751e-05,
+      "loss": 2.0696,
+      "step": 12356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1090537095413106e-05,
+      "loss": 2.0291,
+      "step": 12357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1088356389830465e-05,
+      "loss": 2.033,
+      "step": 12358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1086175806494026e-05,
+      "loss": 2.0684,
+      "step": 12359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1083995345460963e-05,
+      "loss": 2.072,
+      "step": 12360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1081815006788458e-05,
+      "loss": 2.0759,
+      "step": 12361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1079634790533691e-05,
+      "loss": 2.0933,
+      "step": 12362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.107745469675383e-05,
+      "loss": 2.0295,
+      "step": 12363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.107527472550605e-05,
+      "loss": 2.0396,
+      "step": 12364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1073094876847522e-05,
+      "loss": 2.0457,
+      "step": 12365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1070915150835396e-05,
+      "loss": 2.0405,
+      "step": 12366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1068735547526851e-05,
+      "loss": 2.0608,
+      "step": 12367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1066556066979028e-05,
+      "loss": 2.0361,
+      "step": 12368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1064376709249096e-05,
+      "loss": 2.0627,
+      "step": 12369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1062197474394198e-05,
+      "loss": 2.0637,
+      "step": 12370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1060018362471475e-05,
+      "loss": 2.093,
+      "step": 12371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1057839373538087e-05,
+      "loss": 2.0803,
+      "step": 12372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1055660507651164e-05,
+      "loss": 2.0854,
+      "step": 12373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1053481764867851e-05,
+      "loss": 2.084,
+      "step": 12374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1051303145245282e-05,
+      "loss": 2.0151,
+      "step": 12375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1049124648840581e-05,
+      "loss": 2.001,
+      "step": 12376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1046946275710888e-05,
+      "loss": 2.0454,
+      "step": 12377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1044768025913317e-05,
+      "loss": 2.0093,
+      "step": 12378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1042589899505002e-05,
+      "loss": 2.0457,
+      "step": 12379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1040411896543053e-05,
+      "loss": 2.0701,
+      "step": 12380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1038234017084587e-05,
+      "loss": 2.0837,
+      "step": 12381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.103605626118672e-05,
+      "loss": 2.092,
+      "step": 12382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1033878628906555e-05,
+      "loss": 2.0083,
+      "step": 12383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1031701120301207e-05,
+      "loss": 2.0166,
+      "step": 12384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1029523735427774e-05,
+      "loss": 2.0254,
+      "step": 12385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1027346474343347e-05,
+      "loss": 2.043,
+      "step": 12386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1025169337105038e-05,
+      "loss": 2.0808,
+      "step": 12387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1022992323769927e-05,
+      "loss": 2.0908,
+      "step": 12388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1020815434395112e-05,
+      "loss": 2.0833,
+      "step": 12389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1018638669037675e-05,
+      "loss": 2.1245,
+      "step": 12390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1016462027754696e-05,
+      "loss": 2.095,
+      "step": 12391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1014285510603266e-05,
+      "loss": 2.135,
+      "step": 12392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1012109117640446e-05,
+      "loss": 2.0012,
+      "step": 12393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1009932848923323e-05,
+      "loss": 2.0239,
+      "step": 12394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1007756704508964e-05,
+      "loss": 2.0295,
+      "step": 12395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1005580684454428e-05,
+      "loss": 2.0354,
+      "step": 12396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1003404788816787e-05,
+      "loss": 2.0076,
+      "step": 12397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.10012290176531e-05,
+      "loss": 2.0088,
+      "step": 12398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0999053371020422e-05,
+      "loss": 2.1106,
+      "step": 12399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.099687784897581e-05,
+      "loss": 2.0938,
+      "step": 12400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0994702451576307e-05,
+      "loss": 2.0464,
+      "step": 12401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0992527178878972e-05,
+      "loss": 2.0703,
+      "step": 12402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0990352030940836e-05,
+      "loss": 2.0859,
+      "step": 12403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.098817700781895e-05,
+      "loss": 2.0686,
+      "step": 12404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.098600210957035e-05,
+      "loss": 2.0535,
+      "step": 12405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0983827336252063e-05,
+      "loss": 2.0493,
+      "step": 12406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0981652687921129e-05,
+      "loss": 2.0671,
+      "step": 12407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0979478164634567e-05,
+      "loss": 2.0684,
+      "step": 12408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.097730376644941e-05,
+      "loss": 2.0266,
+      "step": 12409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0975129493422676e-05,
+      "loss": 2.0815,
+      "step": 12410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0972955345611374e-05,
+      "loss": 2.0688,
+      "step": 12411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0970781323072535e-05,
+      "loss": 2.0967,
+      "step": 12412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0968607425863152e-05,
+      "loss": 2.1008,
+      "step": 12413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0966433654040252e-05,
+      "loss": 2.1099,
+      "step": 12414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0964260007660823e-05,
+      "loss": 2.0356,
+      "step": 12415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0962086486781876e-05,
+      "loss": 2.0479,
+      "step": 12416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0959913091460406e-05,
+      "loss": 2.1111,
+      "step": 12417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0957739821753405e-05,
+      "loss": 2.1013,
+      "step": 12418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0955566677717874e-05,
+      "loss": 2.0889,
+      "step": 12419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0953393659410792e-05,
+      "loss": 2.0957,
+      "step": 12420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0951220766889145e-05,
+      "loss": 2.0356,
+      "step": 12421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.094904800020992e-05,
+      "loss": 2.0464,
+      "step": 12422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0946875359430085e-05,
+      "loss": 2.0696,
+      "step": 12423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0944702844606629e-05,
+      "loss": 2.0637,
+      "step": 12424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0942530455796509e-05,
+      "loss": 2.0295,
+      "step": 12425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0940358193056706e-05,
+      "loss": 1.9824,
+      "step": 12426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0938186056444182e-05,
+      "loss": 2.0886,
+      "step": 12427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0936014046015888e-05,
+      "loss": 2.0793,
+      "step": 12428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0933842161828799e-05,
+      "loss": 2.1111,
+      "step": 12429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0931670403939854e-05,
+      "loss": 2.0774,
+      "step": 12430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.092949877240602e-05,
+      "loss": 2.072,
+      "step": 12431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0927327267284235e-05,
+      "loss": 2.0503,
+      "step": 12432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0925155888631447e-05,
+      "loss": 2.0486,
+      "step": 12433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0922984636504599e-05,
+      "loss": 2.1064,
+      "step": 12434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.092081351096063e-05,
+      "loss": 2.0837,
+      "step": 12435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0918642512056473e-05,
+      "loss": 2.095,
+      "step": 12436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0916471639849066e-05,
+      "loss": 2.0671,
+      "step": 12437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0914300894395325e-05,
+      "loss": 2.0486,
+      "step": 12438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0912130275752188e-05,
+      "loss": 2.0869,
+      "step": 12439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0909959783976567e-05,
+      "loss": 2.0747,
+      "step": 12440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0907789419125393e-05,
+      "loss": 2.0706,
+      "step": 12441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0905619181255572e-05,
+      "loss": 2.0696,
+      "step": 12442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0903449070424015e-05,
+      "loss": 2.002,
+      "step": 12443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0901279086687636e-05,
+      "loss": 2.0364,
+      "step": 12444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0899109230103334e-05,
+      "loss": 2.0637,
+      "step": 12445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.089693950072802e-05,
+      "loss": 2.0601,
+      "step": 12446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0894769898618588e-05,
+      "loss": 2.0371,
+      "step": 12447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0892600423831927e-05,
+      "loss": 2.0762,
+      "step": 12448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0890431076424942e-05,
+      "loss": 2.0806,
+      "step": 12449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0888261856454507e-05,
+      "loss": 2.0984,
+      "step": 12450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.088609276397752e-05,
+      "loss": 2.0403,
+      "step": 12451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0883923799050854e-05,
+      "loss": 2.0325,
+      "step": 12452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.088175496173139e-05,
+      "loss": 2.0571,
+      "step": 12453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0879586252076009e-05,
+      "loss": 2.0862,
+      "step": 12454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0877417670141573e-05,
+      "loss": 2.1074,
+      "step": 12455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0875249215984961e-05,
+      "loss": 2.1079,
+      "step": 12456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0873080889663034e-05,
+      "loss": 2.0374,
+      "step": 12457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0870912691232649e-05,
+      "loss": 2.0295,
+      "step": 12458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.086874462075067e-05,
+      "loss": 2.0867,
+      "step": 12459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0866576678273946e-05,
+      "loss": 2.0933,
+      "step": 12460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0864408863859342e-05,
+      "loss": 2.0972,
+      "step": 12461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0862241177563694e-05,
+      "loss": 2.0918,
+      "step": 12462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0860073619443848e-05,
+      "loss": 2.0657,
+      "step": 12463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0857906189556652e-05,
+      "loss": 2.0537,
+      "step": 12464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0855738887958936e-05,
+      "loss": 2.0728,
+      "step": 12465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0853571714707546e-05,
+      "loss": 2.071,
+      "step": 12466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0851404669859303e-05,
+      "loss": 2.0947,
+      "step": 12467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.084923775347104e-05,
+      "loss": 2.1172,
+      "step": 12468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0847070965599582e-05,
+      "loss": 2.064,
+      "step": 12469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.084490430630175e-05,
+      "loss": 2.0679,
+      "step": 12470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.084273777563436e-05,
+      "loss": 2.0396,
+      "step": 12471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0840571373654236e-05,
+      "loss": 2.042,
+      "step": 12472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0838405100418173e-05,
+      "loss": 2.075,
+      "step": 12473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0836238955982996e-05,
+      "loss": 2.0693,
+      "step": 12474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0834072940405496e-05,
+      "loss": 2.0679,
+      "step": 12475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0831907053742482e-05,
+      "loss": 2.0627,
+      "step": 12476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0829741296050754e-05,
+      "loss": 2.0723,
+      "step": 12477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0827575667387096e-05,
+      "loss": 2.105,
+      "step": 12478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0825410167808312e-05,
+      "loss": 2.0693,
+      "step": 12479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0823244797371174e-05,
+      "loss": 2.0466,
+      "step": 12480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0821079556132487e-05,
+      "loss": 2.0234,
+      "step": 12481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0818914444149016e-05,
+      "loss": 2.0479,
+      "step": 12482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0816749461477538e-05,
+      "loss": 2.1018,
+      "step": 12483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0814584608174838e-05,
+      "loss": 2.1223,
+      "step": 12484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0812419884297676e-05,
+      "loss": 2.0996,
+      "step": 12485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0810255289902827e-05,
+      "loss": 2.1021,
+      "step": 12486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0808090825047055e-05,
+      "loss": 2.0237,
+      "step": 12487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0805926489787114e-05,
+      "loss": 2.0376,
+      "step": 12488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.080376228417977e-05,
+      "loss": 2.0674,
+      "step": 12489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0801598208281764e-05,
+      "loss": 2.0676,
+      "step": 12490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0799434262149863e-05,
+      "loss": 2.0889,
+      "step": 12491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0797270445840803e-05,
+      "loss": 2.0801,
+      "step": 12492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0795106759411325e-05,
+      "loss": 2.0332,
+      "step": 12493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0792943202918181e-05,
+      "loss": 2.0876,
+      "step": 12494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0790779776418093e-05,
+      "loss": 2.0693,
+      "step": 12495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0788616479967809e-05,
+      "loss": 2.1086,
+      "step": 12496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0786453313624051e-05,
+      "loss": 2.094,
+      "step": 12497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0784290277443542e-05,
+      "loss": 2.1069,
+      "step": 12498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0782127371483016e-05,
+      "loss": 2.0938,
+      "step": 12499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0779964595799177e-05,
+      "loss": 2.0825,
+      "step": 12500
+    },
+    {
+      "Bleu_1": 0.7736878350624731,
+      "Bleu_2": 0.6132061183467616,
+      "Bleu_3": 0.47063800552087776,
+      "Bleu_4": 0.3568792411421486,
+      "CIDEr": 1.2034099695225449,
+      "METEOR": 0.27846207762327846,
+      "ROUGE_L": 0.5593376653221662,
+      "epoch": 0.62,
+      "step": 12500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.077780195044876e-05,
+      "loss": 2.0835,
+      "step": 12501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0775639435488465e-05,
+      "loss": 2.0557,
+      "step": 12502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0773477050975005e-05,
+      "loss": 2.0686,
+      "step": 12503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0771314796965085e-05,
+      "loss": 2.0701,
+      "step": 12504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.076915267351541e-05,
+      "loss": 2.0176,
+      "step": 12505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0766990680682676e-05,
+      "loss": 2.05,
+      "step": 12506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0764828818523584e-05,
+      "loss": 2.0623,
+      "step": 12507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0762667087094817e-05,
+      "loss": 2.0173,
+      "step": 12508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0760505486453078e-05,
+      "loss": 2.0305,
+      "step": 12509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0758344016655036e-05,
+      "loss": 2.0454,
+      "step": 12510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0756182677757387e-05,
+      "loss": 2.0891,
+      "step": 12511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0754021469816805e-05,
+      "loss": 2.0986,
+      "step": 12512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0751860392889958e-05,
+      "loss": 2.0833,
+      "step": 12513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0749699447033529e-05,
+      "loss": 2.0764,
+      "step": 12514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0747538632304175e-05,
+      "loss": 2.0742,
+      "step": 12515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0745377948758573e-05,
+      "loss": 2.0557,
+      "step": 12516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0743217396453379e-05,
+      "loss": 2.0618,
+      "step": 12517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0741056975445245e-05,
+      "loss": 2.0642,
+      "step": 12518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0738896685790836e-05,
+      "loss": 2.0813,
+      "step": 12519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0736736527546792e-05,
+      "loss": 2.0591,
+      "step": 12520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.073457650076977e-05,
+      "loss": 2.0105,
+      "step": 12521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0732416605516411e-05,
+      "loss": 2.0522,
+      "step": 12522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0730256841843355e-05,
+      "loss": 2.0933,
+      "step": 12523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0728097209807242e-05,
+      "loss": 2.071,
+      "step": 12524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0725937709464698e-05,
+      "loss": 2.0649,
+      "step": 12525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0723778340872366e-05,
+      "loss": 2.082,
+      "step": 12526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0721619104086866e-05,
+      "loss": 2.0522,
+      "step": 12527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0719459999164814e-05,
+      "loss": 2.0747,
+      "step": 12528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0717301026162845e-05,
+      "loss": 2.146,
+      "step": 12529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0715142185137562e-05,
+      "loss": 2.1055,
+      "step": 12530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0712983476145588e-05,
+      "loss": 2.0979,
+      "step": 12531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0710824899243533e-05,
+      "loss": 2.0701,
+      "step": 12532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.070866645448799e-05,
+      "loss": 2.0811,
+      "step": 12533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0706508141935576e-05,
+      "loss": 2.0886,
+      "step": 12534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.070434996164288e-05,
+      "loss": 2.0244,
+      "step": 12535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.070219191366651e-05,
+      "loss": 2.0469,
+      "step": 12536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0700033998063046e-05,
+      "loss": 2.0862,
+      "step": 12537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0697876214889083e-05,
+      "loss": 2.0378,
+      "step": 12538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0695718564201205e-05,
+      "loss": 2.0525,
+      "step": 12539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0693561046055996e-05,
+      "loss": 2.0789,
+      "step": 12540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.069140366051003e-05,
+      "loss": 2.0483,
+      "step": 12541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0689246407619888e-05,
+      "loss": 2.0366,
+      "step": 12542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0687089287442133e-05,
+      "loss": 2.0286,
+      "step": 12543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0684932300033343e-05,
+      "loss": 2.0637,
+      "step": 12544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0682775445450072e-05,
+      "loss": 2.1067,
+      "step": 12545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0680618723748892e-05,
+      "loss": 2.1187,
+      "step": 12546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0678462134986349e-05,
+      "loss": 2.0813,
+      "step": 12547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0676305679219007e-05,
+      "loss": 2.0811,
+      "step": 12548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0674149356503415e-05,
+      "loss": 1.9841,
+      "step": 12549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0671993166896109e-05,
+      "loss": 1.979,
+      "step": 12550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0669837110453649e-05,
+      "loss": 2.0334,
+      "step": 12551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0667681187232561e-05,
+      "loss": 2.0662,
+      "step": 12552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0665525397289394e-05,
+      "loss": 2.0913,
+      "step": 12553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0663369740680675e-05,
+      "loss": 2.1008,
+      "step": 12554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0661214217462926e-05,
+      "loss": 2.0957,
+      "step": 12555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0659058827692686e-05,
+      "loss": 2.1106,
+      "step": 12556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.065690357142647e-05,
+      "loss": 2.0845,
+      "step": 12557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0654748448720802e-05,
+      "loss": 2.0874,
+      "step": 12558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0652593459632196e-05,
+      "loss": 2.0288,
+      "step": 12559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0650438604217158e-05,
+      "loss": 2.072,
+      "step": 12560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0648283882532204e-05,
+      "loss": 2.0562,
+      "step": 12561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0646129294633837e-05,
+      "loss": 2.053,
+      "step": 12562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0643974840578559e-05,
+      "loss": 2.0466,
+      "step": 12563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0641820520422869e-05,
+      "loss": 2.0706,
+      "step": 12564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0639666334223252e-05,
+      "loss": 2.0264,
+      "step": 12565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0637512282036215e-05,
+      "loss": 2.0486,
+      "step": 12566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0635358363918233e-05,
+      "loss": 2.0674,
+      "step": 12567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0633204579925794e-05,
+      "loss": 2.0776,
+      "step": 12568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0631050930115384e-05,
+      "loss": 2.0315,
+      "step": 12569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0628897414543468e-05,
+      "loss": 2.053,
+      "step": 12570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0626744033266531e-05,
+      "loss": 2.1238,
+      "step": 12571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0624590786341033e-05,
+      "loss": 2.137,
+      "step": 12572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.062243767382345e-05,
+      "loss": 2.082,
+      "step": 12573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.062028469577024e-05,
+      "loss": 2.1099,
+      "step": 12574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0618131852237862e-05,
+      "loss": 2.0737,
+      "step": 12575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0615979143282774e-05,
+      "loss": 2.1077,
+      "step": 12576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0613826568961424e-05,
+      "loss": 2.0981,
+      "step": 12577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0611674129330265e-05,
+      "loss": 2.0933,
+      "step": 12578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0609521824445743e-05,
+      "loss": 2.0256,
+      "step": 12579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0607369654364293e-05,
+      "loss": 2.0593,
+      "step": 12580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0605217619142364e-05,
+      "loss": 2.0728,
+      "step": 12581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0603065718836377e-05,
+      "loss": 2.0923,
+      "step": 12582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0600913953502777e-05,
+      "loss": 2.0916,
+      "step": 12583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0598762323197986e-05,
+      "loss": 2.0962,
+      "step": 12584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.059661082797842e-05,
+      "loss": 2.0947,
+      "step": 12585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0594459467900512e-05,
+      "loss": 2.0942,
+      "step": 12586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0592308243020668e-05,
+      "loss": 2.0974,
+      "step": 12587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0590157153395313e-05,
+      "loss": 2.083,
+      "step": 12588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0588006199080851e-05,
+      "loss": 2.002,
+      "step": 12589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.058585538013368e-05,
+      "loss": 2.0037,
+      "step": 12590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0583704696610217e-05,
+      "loss": 2.1309,
+      "step": 12591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0581554148566855e-05,
+      "loss": 2.1089,
+      "step": 12592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0579403736059985e-05,
+      "loss": 2.0918,
+      "step": 12593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.057725345914601e-05,
+      "loss": 2.1094,
+      "step": 12594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0575103317881305e-05,
+      "loss": 2.0674,
+      "step": 12595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0572953312322268e-05,
+      "loss": 2.0686,
+      "step": 12596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0570803442525269e-05,
+      "loss": 2.0481,
+      "step": 12597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0568653708546697e-05,
+      "loss": 2.0637,
+      "step": 12598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0566504110442922e-05,
+      "loss": 2.0322,
+      "step": 12599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0564354648270304e-05,
+      "loss": 2.0278,
+      "step": 12600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0562205322085226e-05,
+      "loss": 2.0469,
+      "step": 12601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0560056131944042e-05,
+      "loss": 2.0667,
+      "step": 12602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0557907077903118e-05,
+      "loss": 2.1023,
+      "step": 12603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.055575816001881e-05,
+      "loss": 2.085,
+      "step": 12604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0553609378347459e-05,
+      "loss": 2.0781,
+      "step": 12605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0551460732945432e-05,
+      "loss": 2.0859,
+      "step": 12606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.054931222386906e-05,
+      "loss": 2.0701,
+      "step": 12607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0547163851174698e-05,
+      "loss": 2.0347,
+      "step": 12608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0545015614918673e-05,
+      "loss": 2.113,
+      "step": 12609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0542867515157329e-05,
+      "loss": 2.1011,
+      "step": 12610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0540719551946991e-05,
+      "loss": 2.0518,
+      "step": 12611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.053857172534399e-05,
+      "loss": 2.0681,
+      "step": 12612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0536424035404648e-05,
+      "loss": 2.0942,
+      "step": 12613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0534276482185292e-05,
+      "loss": 2.1055,
+      "step": 12614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0532129065742226e-05,
+      "loss": 2.1233,
+      "step": 12615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0529981786131782e-05,
+      "loss": 2.1011,
+      "step": 12616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0527834643410253e-05,
+      "loss": 2.0366,
+      "step": 12617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0525687637633956e-05,
+      "loss": 2.0735,
+      "step": 12618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.052354076885919e-05,
+      "loss": 2.0786,
+      "step": 12619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0521394037142248e-05,
+      "loss": 2.0747,
+      "step": 12620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0519247442539439e-05,
+      "loss": 2.1113,
+      "step": 12621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.051710098510704e-05,
+      "loss": 2.0771,
+      "step": 12622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0514954664901356e-05,
+      "loss": 2.0186,
+      "step": 12623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.051280848197866e-05,
+      "loss": 2.0457,
+      "step": 12624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0510662436395232e-05,
+      "loss": 2.0754,
+      "step": 12625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0508516528207355e-05,
+      "loss": 2.0811,
+      "step": 12626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.05063707574713e-05,
+      "loss": 2.0652,
+      "step": 12627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.050422512424334e-05,
+      "loss": 2.0813,
+      "step": 12628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0502079628579743e-05,
+      "loss": 2.1121,
+      "step": 12629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0499934270536764e-05,
+      "loss": 2.0918,
+      "step": 12630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0497789050170676e-05,
+      "loss": 2.0461,
+      "step": 12631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0495643967537718e-05,
+      "loss": 2.0281,
+      "step": 12632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0493499022694157e-05,
+      "loss": 2.0762,
+      "step": 12633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0491354215696239e-05,
+      "loss": 2.0566,
+      "step": 12634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0489209546600196e-05,
+      "loss": 2.1001,
+      "step": 12635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0487065015462287e-05,
+      "loss": 2.0776,
+      "step": 12636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0484920622338738e-05,
+      "loss": 2.0344,
+      "step": 12637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0482776367285793e-05,
+      "loss": 2.0283,
+      "step": 12638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0480632250359678e-05,
+      "loss": 2.0457,
+      "step": 12639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.047848827161661e-05,
+      "loss": 2.0288,
+      "step": 12640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.047634443111283e-05,
+      "loss": 2.0754,
+      "step": 12641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0474200728904544e-05,
+      "loss": 2.0515,
+      "step": 12642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0472057165047978e-05,
+      "loss": 2.0881,
+      "step": 12643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0469913739599337e-05,
+      "loss": 2.1008,
+      "step": 12644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0467770452614833e-05,
+      "loss": 2.0947,
+      "step": 12645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.046562730415067e-05,
+      "loss": 2.1018,
+      "step": 12646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0463484294263053e-05,
+      "loss": 2.0842,
+      "step": 12647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0461341423008174e-05,
+      "loss": 2.0959,
+      "step": 12648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0459198690442237e-05,
+      "loss": 2.0378,
+      "step": 12649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0457056096621417e-05,
+      "loss": 2.032,
+      "step": 12650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0454913641601919e-05,
+      "loss": 2.04,
+      "step": 12651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0452771325439911e-05,
+      "loss": 2.0754,
+      "step": 12652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0450629148191586e-05,
+      "loss": 2.1016,
+      "step": 12653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0448487109913112e-05,
+      "loss": 2.1057,
+      "step": 12654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.044634521066066e-05,
+      "loss": 2.0774,
+      "step": 12655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0444203450490404e-05,
+      "loss": 2.0356,
+      "step": 12656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0442061829458506e-05,
+      "loss": 2.0361,
+      "step": 12657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.043992034762113e-05,
+      "loss": 2.0183,
+      "step": 12658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0437779005034436e-05,
+      "loss": 2.0212,
+      "step": 12659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0435637801754567e-05,
+      "loss": 2.0281,
+      "step": 12660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0433496737837686e-05,
+      "loss": 2.0298,
+      "step": 12661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0431355813339935e-05,
+      "loss": 2.0425,
+      "step": 12662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0429215028317456e-05,
+      "loss": 2.0981,
+      "step": 12663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0427074382826396e-05,
+      "loss": 2.0767,
+      "step": 12664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0424933876922876e-05,
+      "loss": 2.0696,
+      "step": 12665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0422793510663045e-05,
+      "loss": 2.0662,
+      "step": 12666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0420653284103017e-05,
+      "loss": 2.0552,
+      "step": 12667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0418513197298931e-05,
+      "loss": 2.0715,
+      "step": 12668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0416373250306895e-05,
+      "loss": 2.071,
+      "step": 12669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.041423344318304e-05,
+      "loss": 2.0759,
+      "step": 12670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0412093775983472e-05,
+      "loss": 2.0383,
+      "step": 12671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0409954248764294e-05,
+      "loss": 2.061,
+      "step": 12672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0407814861581628e-05,
+      "loss": 2.0415,
+      "step": 12673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0405675614491565e-05,
+      "loss": 2.0234,
+      "step": 12674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0403536507550214e-05,
+      "loss": 2.093,
+      "step": 12675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0401397540813666e-05,
+      "loss": 2.1018,
+      "step": 12676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0399258714338006e-05,
+      "loss": 2.0408,
+      "step": 12677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0397120028179337e-05,
+      "loss": 2.0454,
+      "step": 12678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0394981482393726e-05,
+      "loss": 2.033,
+      "step": 12679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0392843077037271e-05,
+      "loss": 2.0437,
+      "step": 12680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0390704812166038e-05,
+      "loss": 1.9927,
+      "step": 12681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0388566687836104e-05,
+      "loss": 2.02,
+      "step": 12682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0386428704103538e-05,
+      "loss": 1.9941,
+      "step": 12683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0384290861024408e-05,
+      "loss": 2.0283,
+      "step": 12684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0382153158654775e-05,
+      "loss": 1.9873,
+      "step": 12685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.03800155970507e-05,
+      "loss": 2.0127,
+      "step": 12686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0377878176268235e-05,
+      "loss": 2.0916,
+      "step": 12687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0375740896363432e-05,
+      "loss": 2.135,
+      "step": 12688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0373603757392336e-05,
+      "loss": 2.0286,
+      "step": 12689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0371466759411002e-05,
+      "loss": 2.0598,
+      "step": 12690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0369329902475458e-05,
+      "loss": 2.0601,
+      "step": 12691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0367193186641744e-05,
+      "loss": 2.0723,
+      "step": 12692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0365056611965897e-05,
+      "loss": 2.0593,
+      "step": 12693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0362920178503938e-05,
+      "loss": 2.0647,
+      "step": 12694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0360783886311902e-05,
+      "loss": 2.0925,
+      "step": 12695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0358647735445806e-05,
+      "loss": 2.0977,
+      "step": 12696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.035651172596167e-05,
+      "loss": 2.0061,
+      "step": 12697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0354375857915503e-05,
+      "loss": 2.031,
+      "step": 12698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0352240131363323e-05,
+      "loss": 2.0684,
+      "step": 12699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.035010454636113e-05,
+      "loss": 2.0605,
+      "step": 12700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0347969102964935e-05,
+      "loss": 2.0701,
+      "step": 12701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.034583380123073e-05,
+      "loss": 2.0649,
+      "step": 12702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0343698641214517e-05,
+      "loss": 2.0247,
+      "step": 12703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0341563622972282e-05,
+      "loss": 2.0115,
+      "step": 12704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.033942874656002e-05,
+      "loss": 2.0193,
+      "step": 12705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0337294012033711e-05,
+      "loss": 2.0396,
+      "step": 12706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0335159419449335e-05,
+      "loss": 2.0442,
+      "step": 12707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0333024968862875e-05,
+      "loss": 2.0618,
+      "step": 12708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0330890660330296e-05,
+      "loss": 2.0474,
+      "step": 12709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0328756493907578e-05,
+      "loss": 2.0669,
+      "step": 12710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0326622469650681e-05,
+      "loss": 2.019,
+      "step": 12711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0324488587615563e-05,
+      "loss": 2.0164,
+      "step": 12712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0322354847858193e-05,
+      "loss": 2.1274,
+      "step": 12713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0320221250434516e-05,
+      "loss": 2.1221,
+      "step": 12714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0318087795400492e-05,
+      "loss": 1.9807,
+      "step": 12715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0315954482812061e-05,
+      "loss": 1.9604,
+      "step": 12716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.031382131272517e-05,
+      "loss": 2.1277,
+      "step": 12717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0311688285195758e-05,
+      "loss": 2.093,
+      "step": 12718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0309555400279762e-05,
+      "loss": 2.0879,
+      "step": 12719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0307422658033114e-05,
+      "loss": 2.0833,
+      "step": 12720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0305290058511747e-05,
+      "loss": 2.0576,
+      "step": 12721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0303157601771575e-05,
+      "loss": 2.0771,
+      "step": 12722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.030102528786853e-05,
+      "loss": 2.0173,
+      "step": 12723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.029889311685852e-05,
+      "loss": 2.0391,
+      "step": 12724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0296761088797472e-05,
+      "loss": 2.0574,
+      "step": 12725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0294629203741287e-05,
+      "loss": 2.0591,
+      "step": 12726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0292497461745868e-05,
+      "loss": 2.0964,
+      "step": 12727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0290365862867125e-05,
+      "loss": 2.1013,
+      "step": 12728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0288234407160949e-05,
+      "loss": 2.0398,
+      "step": 12729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0286103094683245e-05,
+      "loss": 2.0459,
+      "step": 12730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0283971925489897e-05,
+      "loss": 2.0432,
+      "step": 12731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0281840899636793e-05,
+      "loss": 2.0684,
+      "step": 12732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0279710017179817e-05,
+      "loss": 2.0742,
+      "step": 12733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.027757927817485e-05,
+      "loss": 2.075,
+      "step": 12734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0275448682677765e-05,
+      "loss": 2.0024,
+      "step": 12735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0273318230744444e-05,
+      "loss": 2.0107,
+      "step": 12736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0271187922430743e-05,
+      "loss": 2.0806,
+      "step": 12737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0269057757792538e-05,
+      "loss": 2.0635,
+      "step": 12738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0266927736885676e-05,
+      "loss": 2.0723,
+      "step": 12739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0264797859766033e-05,
+      "loss": 2.0498,
+      "step": 12740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0262668126489451e-05,
+      "loss": 2.0437,
+      "step": 12741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0260538537111776e-05,
+      "loss": 2.0808,
+      "step": 12742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0258409091688865e-05,
+      "loss": 2.0403,
+      "step": 12743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0256279790276549e-05,
+      "loss": 2.0403,
+      "step": 12744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0254150632930678e-05,
+      "loss": 2.0979,
+      "step": 12745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.025202161970708e-05,
+      "loss": 2.0708,
+      "step": 12746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0249892750661582e-05,
+      "loss": 2.1077,
+      "step": 12747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0247764025850021e-05,
+      "loss": 2.1042,
+      "step": 12748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0245635445328209e-05,
+      "loss": 2.052,
+      "step": 12749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0243507009151978e-05,
+      "loss": 2.0603,
+      "step": 12750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0241378717377136e-05,
+      "loss": 2.092,
+      "step": 12751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0239250570059494e-05,
+      "loss": 2.1069,
+      "step": 12752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0237122567254864e-05,
+      "loss": 2.05,
+      "step": 12753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0234994709019049e-05,
+      "loss": 2.0422,
+      "step": 12754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0232866995407851e-05,
+      "loss": 2.0908,
+      "step": 12755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0230739426477067e-05,
+      "loss": 2.123,
+      "step": 12756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0228612002282483e-05,
+      "loss": 2.0198,
+      "step": 12757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.02264847228799e-05,
+      "loss": 2.0266,
+      "step": 12758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0224357588325091e-05,
+      "loss": 2.052,
+      "step": 12759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.022223059867385e-05,
+      "loss": 2.0764,
+      "step": 12760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0220103753981949e-05,
+      "loss": 2.0161,
+      "step": 12761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0217977054305152e-05,
+      "loss": 2.0447,
+      "step": 12762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0215850499699252e-05,
+      "loss": 2.0383,
+      "step": 12763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.021372409021999e-05,
+      "loss": 2.0674,
+      "step": 12764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.021159782592315e-05,
+      "loss": 2.092,
+      "step": 12765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.020947170686448e-05,
+      "loss": 2.075,
+      "step": 12766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0207345733099734e-05,
+      "loss": 2.0305,
+      "step": 12767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0205219904684668e-05,
+      "loss": 2.0291,
+      "step": 12768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0203094221675028e-05,
+      "loss": 2.0439,
+      "step": 12769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0200968684126555e-05,
+      "loss": 2.1086,
+      "step": 12770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0198843292094996e-05,
+      "loss": 2.0759,
+      "step": 12771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0196718045636073e-05,
+      "loss": 2.0757,
+      "step": 12772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0194592944805536e-05,
+      "loss": 2.0886,
+      "step": 12773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0192467989659095e-05,
+      "loss": 2.0859,
+      "step": 12774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.019034318025249e-05,
+      "loss": 2.105,
+      "step": 12775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0188218516641435e-05,
+      "loss": 2.0691,
+      "step": 12776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0186093998881643e-05,
+      "loss": 2.0815,
+      "step": 12777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0183969627028836e-05,
+      "loss": 2.0686,
+      "step": 12778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0181845401138713e-05,
+      "loss": 2.0891,
+      "step": 12779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0179721321266992e-05,
+      "loss": 2.084,
+      "step": 12780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0177597387469364e-05,
+      "loss": 2.0581,
+      "step": 12781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0175473599801525e-05,
+      "loss": 2.0654,
+      "step": 12782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0173349958319179e-05,
+      "loss": 2.043,
+      "step": 12783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0171226463078008e-05,
+      "loss": 2.0896,
+      "step": 12784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0169103114133704e-05,
+      "loss": 2.0793,
+      "step": 12785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0166979911541943e-05,
+      "loss": 2.0925,
+      "step": 12786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.016485685535841e-05,
+      "loss": 2.0261,
+      "step": 12787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0162733945638775e-05,
+      "loss": 2.0642,
+      "step": 12788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0160611182438709e-05,
+      "loss": 2.0811,
+      "step": 12789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0158488565813882e-05,
+      "loss": 2.0701,
+      "step": 12790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0156366095819954e-05,
+      "loss": 2.0364,
+      "step": 12791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0154243772512587e-05,
+      "loss": 2.0671,
+      "step": 12792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.015212159594744e-05,
+      "loss": 2.0474,
+      "step": 12793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0149999566180151e-05,
+      "loss": 2.0503,
+      "step": 12794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0147877683266385e-05,
+      "loss": 2.0793,
+      "step": 12795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.014575594726177e-05,
+      "loss": 2.0647,
+      "step": 12796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.014363435822196e-05,
+      "loss": 2.0933,
+      "step": 12797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0141512916202587e-05,
+      "loss": 2.0918,
+      "step": 12798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0139391621259275e-05,
+      "loss": 2.0371,
+      "step": 12799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0137270473447661e-05,
+      "loss": 2.0161,
+      "step": 12800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0135149472823364e-05,
+      "loss": 2.0847,
+      "step": 12801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0133028619442012e-05,
+      "loss": 2.0842,
+      "step": 12802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0130907913359219e-05,
+      "loss": 2.0369,
+      "step": 12803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0128787354630594e-05,
+      "loss": 2.0476,
+      "step": 12804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0126666943311751e-05,
+      "loss": 2.0386,
+      "step": 12805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0124546679458295e-05,
+      "loss": 2.0383,
+      "step": 12806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0122426563125824e-05,
+      "loss": 2.1118,
+      "step": 12807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0120306594369941e-05,
+      "loss": 2.104,
+      "step": 12808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0118186773246231e-05,
+      "loss": 2.0984,
+      "step": 12809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0116067099810296e-05,
+      "loss": 2.0813,
+      "step": 12810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0113947574117709e-05,
+      "loss": 2.061,
+      "step": 12811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0111828196224063e-05,
+      "loss": 2.0918,
+      "step": 12812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0109708966184932e-05,
+      "loss": 2.0652,
+      "step": 12813
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0107589884055884e-05,
+      "loss": 2.0664,
+      "step": 12814
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0105470949892503e-05,
+      "loss": 2.0317,
+      "step": 12815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0103352163750342e-05,
+      "loss": 2.0342,
+      "step": 12816
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0101233525684974e-05,
+      "loss": 2.0469,
+      "step": 12817
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0099115035751955e-05,
+      "loss": 2.0784,
+      "step": 12818
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.009699669400683e-05,
+      "loss": 2.0195,
+      "step": 12819
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0094878500505167e-05,
+      "loss": 2.0393,
+      "step": 12820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0092760455302498e-05,
+      "loss": 2.0417,
+      "step": 12821
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.009064255845438e-05,
+      "loss": 2.0911,
+      "step": 12822
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0088524810016341e-05,
+      "loss": 2.0291,
+      "step": 12823
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0086407210043921e-05,
+      "loss": 2.02,
+      "step": 12824
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0084289758592653e-05,
+      "loss": 2.0674,
+      "step": 12825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0082172455718059e-05,
+      "loss": 2.0803,
+      "step": 12826
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0080055301475672e-05,
+      "loss": 2.0515,
+      "step": 12827
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.007793829592101e-05,
+      "loss": 2.064,
+      "step": 12828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0075821439109576e-05,
+      "loss": 2.05,
+      "step": 12829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0073704731096902e-05,
+      "loss": 2.0725,
+      "step": 12830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0071588171938477e-05,
+      "loss": 2.1248,
+      "step": 12831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0069471761689824e-05,
+      "loss": 2.1113,
+      "step": 12832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.006735550040643e-05,
+      "loss": 1.9963,
+      "step": 12833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0065239388143791e-05,
+      "loss": 2.0056,
+      "step": 12834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0063123424957412e-05,
+      "loss": 2.0942,
+      "step": 12835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0061007610902765e-05,
+      "loss": 2.0955,
+      "step": 12836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0058891946035352e-05,
+      "loss": 2.0542,
+      "step": 12837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.005677643041064e-05,
+      "loss": 2.0601,
+      "step": 12838
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0054661064084114e-05,
+      "loss": 2.1179,
+      "step": 12839
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0052545847111242e-05,
+      "loss": 2.083,
+      "step": 12840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0050430779547498e-05,
+      "loss": 2.0632,
+      "step": 12841
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0048315861448343e-05,
+      "loss": 2.0332,
+      "step": 12842
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0046201092869243e-05,
+      "loss": 2.0779,
+      "step": 12843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0044086473865644e-05,
+      "loss": 2.0569,
+      "step": 12844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0041972004493017e-05,
+      "loss": 1.9934,
+      "step": 12845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0039857684806796e-05,
+      "loss": 1.9954,
+      "step": 12846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.003774351486244e-05,
+      "loss": 2.1052,
+      "step": 12847
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0035629494715378e-05,
+      "loss": 2.1252,
+      "step": 12848
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0033515624421051e-05,
+      "loss": 2.1106,
+      "step": 12849
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.00314019040349e-05,
+      "loss": 2.0635,
+      "step": 12850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0029288333612345e-05,
+      "loss": 2.032,
+      "step": 12851
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0027174913208823e-05,
+      "loss": 2.0613,
+      "step": 12852
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.002506164287975e-05,
+      "loss": 2.083,
+      "step": 12853
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0022948522680535e-05,
+      "loss": 2.0974,
+      "step": 12854
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.002083555266661e-05,
+      "loss": 2.0562,
+      "step": 12855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0018722732893368e-05,
+      "loss": 2.1047,
+      "step": 12856
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0016610063416231e-05,
+      "loss": 2.0632,
+      "step": 12857
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.001449754429059e-05,
+      "loss": 2.0586,
+      "step": 12858
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0012385175571847e-05,
+      "loss": 2.0342,
+      "step": 12859
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0010272957315397e-05,
+      "loss": 2.0388,
+      "step": 12860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0008160889576625e-05,
+      "loss": 2.0288,
+      "step": 12861
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.000604897241093e-05,
+      "loss": 2.0212,
+      "step": 12862
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0003937205873688e-05,
+      "loss": 2.0635,
+      "step": 12863
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.0001825590020269e-05,
+      "loss": 2.072,
+      "step": 12864
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.999714124906062e-06,
+      "loss": 2.1367,
+      "step": 12865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.997602810586425e-06,
+      "loss": 2.1174,
+      "step": 12866
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.995491647116737e-06,
+      "loss": 2.053,
+      "step": 12867
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.993380634552356e-06,
+      "loss": 2.0476,
+      "step": 12868
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.991269772948632e-06,
+      "loss": 2.0669,
+      "step": 12869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.989159062360934e-06,
+      "loss": 2.1084,
+      "step": 12870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.9870485028446e-06,
+      "loss": 2.1201,
+      "step": 12871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.984938094454989e-06,
+      "loss": 2.1199,
+      "step": 12872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.982827837247436e-06,
+      "loss": 2.0417,
+      "step": 12873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.980717731277281e-06,
+      "loss": 2.0205,
+      "step": 12874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.978607776599861e-06,
+      "loss": 2.0811,
+      "step": 12875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.97649797327051e-06,
+      "loss": 2.0881,
+      "step": 12876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.974388321344547e-06,
+      "loss": 2.0525,
+      "step": 12877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.972278820877304e-06,
+      "loss": 2.0583,
+      "step": 12878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.970169471924092e-06,
+      "loss": 2.0823,
+      "step": 12879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.968060274540235e-06,
+      "loss": 2.0544,
+      "step": 12880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.965951228781034e-06,
+      "loss": 2.0671,
+      "step": 12881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.963842334701809e-06,
+      "loss": 2.0476,
+      "step": 12882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.961733592357853e-06,
+      "loss": 2.0283,
+      "step": 12883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.959625001804467e-06,
+      "loss": 2.083,
+      "step": 12884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.95751656309695e-06,
+      "loss": 2.0293,
+      "step": 12885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.955408276290585e-06,
+      "loss": 2.0481,
+      "step": 12886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.953300141440674e-06,
+      "loss": 2.0308,
+      "step": 12887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.95119215860249e-06,
+      "loss": 2.0308,
+      "step": 12888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.94908432783131e-06,
+      "loss": 2.0916,
+      "step": 12889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.94697664918242e-06,
+      "loss": 2.1038,
+      "step": 12890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.944869122711078e-06,
+      "loss": 2.0669,
+      "step": 12891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.942761748472564e-06,
+      "loss": 2.0608,
+      "step": 12892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.940654526522135e-06,
+      "loss": 2.0144,
+      "step": 12893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.938547456915049e-06,
+      "loss": 2.0439,
+      "step": 12894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.936440539706566e-06,
+      "loss": 2.02,
+      "step": 12895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.934333774951932e-06,
+      "loss": 2.0442,
+      "step": 12896
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.932227162706405e-06,
+      "loss": 2.0742,
+      "step": 12897
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.93012070302522e-06,
+      "loss": 2.0776,
+      "step": 12898
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.928014395963612e-06,
+      "loss": 2.0627,
+      "step": 12899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.925908241576828e-06,
+      "loss": 2.0725,
+      "step": 12900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.923802239920089e-06,
+      "loss": 2.0464,
+      "step": 12901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.921696391048631e-06,
+      "loss": 2.0515,
+      "step": 12902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.919590695017676e-06,
+      "loss": 2.0735,
+      "step": 12903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.917485151882435e-06,
+      "loss": 2.0681,
+      "step": 12904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.915379761698133e-06,
+      "loss": 1.999,
+      "step": 12905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.913274524519975e-06,
+      "loss": 1.9832,
+      "step": 12906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.911169440403174e-06,
+      "loss": 2.0632,
+      "step": 12907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.909064509402927e-06,
+      "loss": 2.1157,
+      "step": 12908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.906959731574445e-06,
+      "loss": 2.0288,
+      "step": 12909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.904855106972908e-06,
+      "loss": 2.0212,
+      "step": 12910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.902750635653516e-06,
+      "loss": 2.0928,
+      "step": 12911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.900646317671457e-06,
+      "loss": 2.0808,
+      "step": 12912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.898542153081912e-06,
+      "loss": 2.0393,
+      "step": 12913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.89643814194006e-06,
+      "loss": 2.0632,
+      "step": 12914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.89433428430108e-06,
+      "loss": 2.0974,
+      "step": 12915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.892230580220134e-06,
+      "loss": 2.0657,
+      "step": 12916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.8901270297524e-06,
+      "loss": 2.0896,
+      "step": 12917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.888023632953034e-06,
+      "loss": 2.0874,
+      "step": 12918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.885920389877202e-06,
+      "loss": 1.9736,
+      "step": 12919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.883817300580055e-06,
+      "loss": 1.9937,
+      "step": 12920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.881714365116737e-06,
+      "loss": 2.0769,
+      "step": 12921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.879611583542406e-06,
+      "loss": 2.0557,
+      "step": 12922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.8775089559122e-06,
+      "loss": 2.0669,
+      "step": 12923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.875406482281261e-06,
+      "loss": 2.0562,
+      "step": 12924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.873304162704721e-06,
+      "loss": 2.0393,
+      "step": 12925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.871201997237709e-06,
+      "loss": 2.0593,
+      "step": 12926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.86909998593536e-06,
+      "loss": 2.1167,
+      "step": 12927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.866998128852783e-06,
+      "loss": 2.0874,
+      "step": 12928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.864896426045113e-06,
+      "loss": 2.0833,
+      "step": 12929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.862794877567453e-06,
+      "loss": 2.0818,
+      "step": 12930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.860693483474915e-06,
+      "loss": 2.0461,
+      "step": 12931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.858592243822614e-06,
+      "loss": 2.0535,
+      "step": 12932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.856491158665642e-06,
+      "loss": 2.136,
+      "step": 12933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.854390228059107e-06,
+      "loss": 2.1428,
+      "step": 12934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.852289452058096e-06,
+      "loss": 2.0789,
+      "step": 12935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.8501888307177e-06,
+      "loss": 2.0596,
+      "step": 12936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.848088364093013e-06,
+      "loss": 1.9814,
+      "step": 12937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.845988052239104e-06,
+      "loss": 2.02,
+      "step": 12938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.843887895211067e-06,
+      "loss": 2.0779,
+      "step": 12939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.841787893063967e-06,
+      "loss": 2.0627,
+      "step": 12940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.83968804585287e-06,
+      "loss": 2.1006,
+      "step": 12941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.837588353632853e-06,
+      "loss": 2.1094,
+      "step": 12942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.835488816458966e-06,
+      "loss": 2.1155,
+      "step": 12943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.833389434386278e-06,
+      "loss": 2.1138,
+      "step": 12944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.831290207469839e-06,
+      "loss": 2.0869,
+      "step": 12945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.829191135764693e-06,
+      "loss": 2.0955,
+      "step": 12946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.827092219325894e-06,
+      "loss": 1.9756,
+      "step": 12947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.824993458208479e-06,
+      "loss": 2.0037,
+      "step": 12948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.822894852467488e-06,
+      "loss": 2.0012,
+      "step": 12949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.820796402157955e-06,
+      "loss": 2.0029,
+      "step": 12950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.818698107334906e-06,
+      "loss": 2.0259,
+      "step": 12951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.816599968053367e-06,
+      "loss": 2.0283,
+      "step": 12952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.814501984368362e-06,
+      "loss": 2.0657,
+      "step": 12953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.812404156334906e-06,
+      "loss": 2.0784,
+      "step": 12954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.810306484008012e-06,
+      "loss": 2.0891,
+      "step": 12955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.808208967442691e-06,
+      "loss": 2.0593,
+      "step": 12956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.806111606693946e-06,
+      "loss": 2.05,
+      "step": 12957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.804014401816778e-06,
+      "loss": 2.0845,
+      "step": 12958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.801917352866184e-06,
+      "loss": 2.1042,
+      "step": 12959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.799820459897164e-06,
+      "loss": 2.1128,
+      "step": 12960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.797723722964688e-06,
+      "loss": 2.0984,
+      "step": 12961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.795627142123764e-06,
+      "loss": 2.0984,
+      "step": 12962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.793530717429346e-06,
+      "loss": 2.0596,
+      "step": 12963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.79143444893644e-06,
+      "loss": 2.0698,
+      "step": 12964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.789338336699995e-06,
+      "loss": 2.0046,
+      "step": 12965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.787242380774986e-06,
+      "loss": 2.0015,
+      "step": 12966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.78514658121638e-06,
+      "loss": 1.9724,
+      "step": 12967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.783050938079134e-06,
+      "loss": 1.9634,
+      "step": 12968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.780955451418205e-06,
+      "loss": 2.0774,
+      "step": 12969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.778860121288546e-06,
+      "loss": 2.0986,
+      "step": 12970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.776764947745103e-06,
+      "loss": 2.0413,
+      "step": 12971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.774669930842817e-06,
+      "loss": 2.0762,
+      "step": 12972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.772575070636632e-06,
+      "loss": 2.0703,
+      "step": 12973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.77048036718148e-06,
+      "loss": 2.071,
+      "step": 12974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.7683858205323e-06,
+      "loss": 2.0396,
+      "step": 12975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.766291430744001e-06,
+      "loss": 2.0549,
+      "step": 12976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.764197197871529e-06,
+      "loss": 2.0598,
+      "step": 12977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.76210312196978e-06,
+      "loss": 2.0818,
+      "step": 12978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.760009203093692e-06,
+      "loss": 2.0945,
+      "step": 12979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.757915441298158e-06,
+      "loss": 2.1096,
+      "step": 12980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.755821836638088e-06,
+      "loss": 2.061,
+      "step": 12981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.753728389168387e-06,
+      "loss": 2.0659,
+      "step": 12982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.751635098943954e-06,
+      "loss": 2.0225,
+      "step": 12983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.749541966019681e-06,
+      "loss": 2.0449,
+      "step": 12984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.747448990450464e-06,
+      "loss": 2.0901,
+      "step": 12985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.745356172291173e-06,
+      "loss": 2.0562,
+      "step": 12986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.74326351159671e-06,
+      "loss": 2.0525,
+      "step": 12987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.741171008421933e-06,
+      "loss": 2.0603,
+      "step": 12988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.739078662821736e-06,
+      "loss": 2.0713,
+      "step": 12989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.736986474850975e-06,
+      "loss": 2.053,
+      "step": 12990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.734894444564515e-06,
+      "loss": 2.0515,
+      "step": 12991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.732802572017219e-06,
+      "loss": 2.0686,
+      "step": 12992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.730710857263947e-06,
+      "loss": 2.0747,
+      "step": 12993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.728619300359549e-06,
+      "loss": 2.0854,
+      "step": 12994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.726527901358875e-06,
+      "loss": 2.0586,
+      "step": 12995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.724436660316768e-06,
+      "loss": 2.0771,
+      "step": 12996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.722345577288071e-06,
+      "loss": 2.0371,
+      "step": 12997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.720254652327615e-06,
+      "loss": 2.0447,
+      "step": 12998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.718163885490237e-06,
+      "loss": 2.0635,
+      "step": 12999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.716073276830772e-06,
+      "loss": 2.0679,
+      "step": 13000
+    },
+    {
+      "Bleu_1": 0.7736998722496875,
+      "Bleu_2": 0.612936083933621,
+      "Bleu_3": 0.4704500297325531,
+      "Bleu_4": 0.3567349384462963,
+      "CIDEr": 1.200267469561839,
+      "METEOR": 0.2783179394659513,
+      "ROUGE_L": 0.5597286678681063,
+      "epoch": 0.65,
+      "step": 13000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.71398282640402e-06,
+      "loss": 2.0706,
+      "step": 13001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.71189253426483e-06,
+      "loss": 2.0767,
+      "step": 13002
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.709802400467989e-06,
+      "loss": 2.0627,
+      "step": 13003
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.707712425068338e-06,
+      "loss": 2.0698,
+      "step": 13004
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.70562260812066e-06,
+      "loss": 2.0452,
+      "step": 13005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.703532949679768e-06,
+      "loss": 2.0825,
+      "step": 13006
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.701443449800461e-06,
+      "loss": 2.0737,
+      "step": 13007
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.699354108537531e-06,
+      "loss": 2.0847,
+      "step": 13008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.697264925945772e-06,
+      "loss": 2.0854,
+      "step": 13009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.695175902079967e-06,
+      "loss": 2.0806,
+      "step": 13010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.693087036994899e-06,
+      "loss": 2.0554,
+      "step": 13011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.690998330745348e-06,
+      "loss": 2.0464,
+      "step": 13012
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.688909783386086e-06,
+      "loss": 2.052,
+      "step": 13013
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.686821394971883e-06,
+      "loss": 2.0554,
+      "step": 13014
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.684733165557512e-06,
+      "loss": 2.0994,
+      "step": 13015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.682645095197715e-06,
+      "loss": 2.0796,
+      "step": 13016
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.680557183947273e-06,
+      "loss": 2.0413,
+      "step": 13017
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.678469431860916e-06,
+      "loss": 2.0693,
+      "step": 13018
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.676381838993416e-06,
+      "loss": 2.072,
+      "step": 13019
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.674294405399501e-06,
+      "loss": 2.051,
+      "step": 13020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.672207131133915e-06,
+      "loss": 2.0684,
+      "step": 13021
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.670120016251398e-06,
+      "loss": 2.0645,
+      "step": 13022
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.66803306080668e-06,
+      "loss": 2.0439,
+      "step": 13023
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.665946264854488e-06,
+      "loss": 2.0444,
+      "step": 13024
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.663859628449548e-06,
+      "loss": 2.022,
+      "step": 13025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.66177315164658e-06,
+      "loss": 1.9958,
+      "step": 13026
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.659686834500297e-06,
+      "loss": 2.032,
+      "step": 13027
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.65760067706541e-06,
+      "loss": 2.0439,
+      "step": 13028
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.655514679396627e-06,
+      "loss": 2.0532,
+      "step": 13029
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.653428841548653e-06,
+      "loss": 2.0759,
+      "step": 13030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.651343163576185e-06,
+      "loss": 2.0283,
+      "step": 13031
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.649257645533921e-06,
+      "loss": 2.0493,
+      "step": 13032
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.64717228747654e-06,
+      "loss": 2.0381,
+      "step": 13033
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.645087089458744e-06,
+      "loss": 2.0732,
+      "step": 13034
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.643002051535195e-06,
+      "loss": 2.0098,
+      "step": 13035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.640917173760596e-06,
+      "loss": 2.0273,
+      "step": 13036
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.6388324561896e-06,
+      "loss": 2.0154,
+      "step": 13037
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.636747898876884e-06,
+      "loss": 2.063,
+      "step": 13038
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.634663501877111e-06,
+      "loss": 2.0288,
+      "step": 13039
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.632579265244943e-06,
+      "loss": 2.0588,
+      "step": 13040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.63049518903504e-06,
+      "loss": 2.0833,
+      "step": 13041
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.628411273302047e-06,
+      "loss": 2.0789,
+      "step": 13042
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.62632751810062e-06,
+      "loss": 2.0444,
+      "step": 13043
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.624243923485397e-06,
+      "loss": 2.0488,
+      "step": 13044
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.622160489511023e-06,
+      "loss": 2.0806,
+      "step": 13045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.620077216232128e-06,
+      "loss": 2.0598,
+      "step": 13046
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.617994103703355e-06,
+      "loss": 2.0835,
+      "step": 13047
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.61591115197931e-06,
+      "loss": 2.0701,
+      "step": 13048
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.613828361114641e-06,
+      "loss": 2.0911,
+      "step": 13049
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.611745731163944e-06,
+      "loss": 2.0928,
+      "step": 13050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.609663262181855e-06,
+      "loss": 2.0085,
+      "step": 13051
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.607580954222966e-06,
+      "loss": 2.021,
+      "step": 13052
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.605498807341892e-06,
+      "loss": 2.1018,
+      "step": 13053
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.603416821593235e-06,
+      "loss": 2.1006,
+      "step": 13054
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.601334997031588e-06,
+      "loss": 2.0378,
+      "step": 13055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.599253333711551e-06,
+      "loss": 2.0703,
+      "step": 13056
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.597171831687713e-06,
+      "loss": 2.0698,
+      "step": 13057
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.595090491014644e-06,
+      "loss": 2.1108,
+      "step": 13058
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.59300931174695e-06,
+      "loss": 2.0044,
+      "step": 13059
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.59092829393918e-06,
+      "loss": 2.0183,
+      "step": 13060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.588847437645934e-06,
+      "loss": 2.0845,
+      "step": 13061
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.58676674292176e-06,
+      "loss": 2.0535,
+      "step": 13062
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.584686209821227e-06,
+      "loss": 1.9546,
+      "step": 13063
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.582605838398896e-06,
+      "loss": 1.9568,
+      "step": 13064
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.580525628709315e-06,
+      "loss": 2.041,
+      "step": 13065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.578445580807056e-06,
+      "loss": 2.0569,
+      "step": 13066
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.576365694746647e-06,
+      "loss": 2.0891,
+      "step": 13067
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.574285970582633e-06,
+      "loss": 2.0864,
+      "step": 13068
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.572206408369555e-06,
+      "loss": 2.0603,
+      "step": 13069
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.570127008161949e-06,
+      "loss": 2.043,
+      "step": 13070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.56804777001434e-06,
+      "loss": 2.0806,
+      "step": 13071
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.565968693981265e-06,
+      "loss": 2.0757,
+      "step": 13072
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.563889780117226e-06,
+      "loss": 2.0007,
+      "step": 13073
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.56181102847676e-06,
+      "loss": 2.031,
+      "step": 13074
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.559732439114359e-06,
+      "loss": 2.1033,
+      "step": 13075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.557654012084556e-06,
+      "loss": 2.1489,
+      "step": 13076
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.555575747441836e-06,
+      "loss": 2.0989,
+      "step": 13077
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.553497645240706e-06,
+      "loss": 2.1045,
+      "step": 13078
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.551419705535663e-06,
+      "loss": 2.0969,
+      "step": 13079
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.549341928381196e-06,
+      "loss": 2.0923,
+      "step": 13080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.547264313831793e-06,
+      "loss": 2.0327,
+      "step": 13081
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.54518686194194e-06,
+      "loss": 2.0476,
+      "step": 13082
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.543109572766111e-06,
+      "loss": 2.0693,
+      "step": 13083
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.541032446358784e-06,
+      "loss": 2.061,
+      "step": 13084
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.538955482774426e-06,
+      "loss": 2.0459,
+      "step": 13085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.536878682067506e-06,
+      "loss": 2.0676,
+      "step": 13086
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.534802044292492e-06,
+      "loss": 2.0859,
+      "step": 13087
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.532725569503821e-06,
+      "loss": 2.1035,
+      "step": 13088
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.53064925775597e-06,
+      "loss": 2.0332,
+      "step": 13089
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.528573109103366e-06,
+      "loss": 2.0298,
+      "step": 13090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.526497123600479e-06,
+      "loss": 2.0867,
+      "step": 13091
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.524421301301726e-06,
+      "loss": 2.0833,
+      "step": 13092
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.522345642261553e-06,
+      "loss": 2.043,
+      "step": 13093
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.52027014653439e-06,
+      "loss": 2.0481,
+      "step": 13094
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.518194814174664e-06,
+      "loss": 2.0889,
+      "step": 13095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.516119645236803e-06,
+      "loss": 2.0891,
+      "step": 13096
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.51404463977522e-06,
+      "loss": 2.0061,
+      "step": 13097
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.511969797844333e-06,
+      "loss": 2.0007,
+      "step": 13098
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.50989511949855e-06,
+      "loss": 2.1199,
+      "step": 13099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.507820604792279e-06,
+      "loss": 2.0923,
+      "step": 13100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.50574625377992e-06,
+      "loss": 2.0596,
+      "step": 13101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.503672066515878e-06,
+      "loss": 2.0466,
+      "step": 13102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.501598043054527e-06,
+      "loss": 2.0674,
+      "step": 13103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.499524183450282e-06,
+      "loss": 2.0962,
+      "step": 13104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.4974504877575e-06,
+      "loss": 2.1174,
+      "step": 13105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.49537695603059e-06,
+      "loss": 2.1013,
+      "step": 13106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.493303588323904e-06,
+      "loss": 2.0439,
+      "step": 13107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.491230384691826e-06,
+      "loss": 2.0298,
+      "step": 13108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.48915734518872e-06,
+      "loss": 2.0908,
+      "step": 13109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.487084469868948e-06,
+      "loss": 2.0994,
+      "step": 13110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.48501175878687e-06,
+      "loss": 2.0232,
+      "step": 13111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.482939211996841e-06,
+      "loss": 2.0303,
+      "step": 13112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.48086682955321e-06,
+      "loss": 2.0781,
+      "step": 13113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.478794611510324e-06,
+      "loss": 2.0925,
+      "step": 13114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.476722557922524e-06,
+      "loss": 2.0464,
+      "step": 13115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.474650668844147e-06,
+      "loss": 2.0432,
+      "step": 13116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.472578944329533e-06,
+      "loss": 2.0823,
+      "step": 13117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.470507384432992e-06,
+      "loss": 2.1099,
+      "step": 13118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.46843598920887e-06,
+      "loss": 2.0789,
+      "step": 13119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.466364758711466e-06,
+      "loss": 2.0752,
+      "step": 13120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.464293692995118e-06,
+      "loss": 2.0676,
+      "step": 13121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.462222792114122e-06,
+      "loss": 2.0854,
+      "step": 13122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.460152056122785e-06,
+      "loss": 2.1011,
+      "step": 13123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.458081485075415e-06,
+      "loss": 2.1025,
+      "step": 13124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.45601107902631e-06,
+      "loss": 2.0688,
+      "step": 13125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.45394083802976e-06,
+      "loss": 2.0959,
+      "step": 13126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.451870762140064e-06,
+      "loss": 2.0483,
+      "step": 13127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.449800851411492e-06,
+      "loss": 2.04,
+      "step": 13128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.447731105898344e-06,
+      "loss": 2.0488,
+      "step": 13129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.445661525654874e-06,
+      "loss": 2.0635,
+      "step": 13130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.443592110735379e-06,
+      "loss": 1.9968,
+      "step": 13131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.441522861194106e-06,
+      "loss": 1.9834,
+      "step": 13132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.43945377708533e-06,
+      "loss": 2.0552,
+      "step": 13133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.437384858463308e-06,
+      "loss": 2.0947,
+      "step": 13134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.435316105382288e-06,
+      "loss": 2.0388,
+      "step": 13135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.433247517896539e-06,
+      "loss": 2.0601,
+      "step": 13136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.43117909606029e-06,
+      "loss": 2.1125,
+      "step": 13137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.429110839927791e-06,
+      "loss": 2.0947,
+      "step": 13138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.427042749553273e-06,
+      "loss": 2.1106,
+      "step": 13139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.424974824990977e-06,
+      "loss": 2.093,
+      "step": 13140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.422907066295127e-06,
+      "loss": 2.0876,
+      "step": 13141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.420839473519955e-06,
+      "loss": 2.0657,
+      "step": 13142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.418772046719663e-06,
+      "loss": 2.0159,
+      "step": 13143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.41670478594849e-06,
+      "loss": 2.0198,
+      "step": 13144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.41463769126063e-06,
+      "loss": 2.1064,
+      "step": 13145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.412570762710304e-06,
+      "loss": 2.0715,
+      "step": 13146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.410504000351703e-06,
+      "loss": 2.104,
+      "step": 13147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.40843740423903e-06,
+      "loss": 2.1165,
+      "step": 13148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.406370974426479e-06,
+      "loss": 2.0776,
+      "step": 13149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.404304710968238e-06,
+      "loss": 2.0676,
+      "step": 13150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.402238613918496e-06,
+      "loss": 2.0486,
+      "step": 13151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.400172683331432e-06,
+      "loss": 2.0227,
+      "step": 13152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.398106919261222e-06,
+      "loss": 2.0986,
+      "step": 13153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.396041321762041e-06,
+      "loss": 2.0925,
+      "step": 13154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.393975890888055e-06,
+      "loss": 2.0688,
+      "step": 13155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.391910626693426e-06,
+      "loss": 2.1228,
+      "step": 13156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.389845529232315e-06,
+      "loss": 2.0073,
+      "step": 13157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.387780598558877e-06,
+      "loss": 2.0088,
+      "step": 13158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.38571583472727e-06,
+      "loss": 2.0605,
+      "step": 13159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.383651237791615e-06,
+      "loss": 2.0481,
+      "step": 13160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.381586807806086e-06,
+      "loss": 2.0735,
+      "step": 13161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.379522544824793e-06,
+      "loss": 2.085,
+      "step": 13162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.377458448901892e-06,
+      "loss": 2.1023,
+      "step": 13163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.375394520091496e-06,
+      "loss": 2.1101,
+      "step": 13164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.373330758447732e-06,
+      "loss": 2.0427,
+      "step": 13165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.371267164024721e-06,
+      "loss": 2.0388,
+      "step": 13166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.369203736876577e-06,
+      "loss": 2.0842,
+      "step": 13167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.367140477057414e-06,
+      "loss": 2.0542,
+      "step": 13168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.365077384621337e-06,
+      "loss": 2.0881,
+      "step": 13169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.363014459622449e-06,
+      "loss": 2.0945,
+      "step": 13170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.360951702114848e-06,
+      "loss": 2.062,
+      "step": 13171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.358889112152625e-06,
+      "loss": 2.0703,
+      "step": 13172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.356826689789869e-06,
+      "loss": 2.0989,
+      "step": 13173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.354764435080674e-06,
+      "loss": 2.0515,
+      "step": 13174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.352702348079103e-06,
+      "loss": 2.0381,
+      "step": 13175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.35064042883925e-06,
+      "loss": 2.0259,
+      "step": 13176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.348578677415166e-06,
+      "loss": 2.0742,
+      "step": 13177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.346517093860943e-06,
+      "loss": 2.0813,
+      "step": 13178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.344455678230626e-06,
+      "loss": 2.0378,
+      "step": 13179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.342394430578275e-06,
+      "loss": 2.0603,
+      "step": 13180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.340333350957945e-06,
+      "loss": 2.0,
+      "step": 13181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.338272439423691e-06,
+      "loss": 2.0303,
+      "step": 13182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.33621169602955e-06,
+      "loss": 2.0398,
+      "step": 13183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.334151120829572e-06,
+      "loss": 2.041,
+      "step": 13184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.332090713877782e-06,
+      "loss": 2.0383,
+      "step": 13185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.330030475228219e-06,
+      "loss": 2.0327,
+      "step": 13186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.327970404934911e-06,
+      "loss": 2.0591,
+      "step": 13187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.325910503051877e-06,
+      "loss": 2.0408,
+      "step": 13188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.323850769633146e-06,
+      "loss": 2.0361,
+      "step": 13189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.321791204732708e-06,
+      "loss": 2.052,
+      "step": 13190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.3197318084046e-06,
+      "loss": 2.1184,
+      "step": 13191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.31767258070281e-06,
+      "loss": 2.1104,
+      "step": 13192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.31561352168135e-06,
+      "loss": 2.0703,
+      "step": 13193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.313554631394206e-06,
+      "loss": 2.0691,
+      "step": 13194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.311495909895374e-06,
+      "loss": 2.073,
+      "step": 13195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.309437357238843e-06,
+      "loss": 2.0791,
+      "step": 13196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.307378973478596e-06,
+      "loss": 2.0371,
+      "step": 13197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.305320758668612e-06,
+      "loss": 2.0288,
+      "step": 13198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.30326271286287e-06,
+      "loss": 2.0271,
+      "step": 13199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.301204836115322e-06,
+      "loss": 2.0137,
+      "step": 13200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.299147128479957e-06,
+      "loss": 2.0303,
+      "step": 13201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.297089590010716e-06,
+      "loss": 2.0352,
+      "step": 13202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.295032220761573e-06,
+      "loss": 2.0222,
+      "step": 13203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.292975020786467e-06,
+      "loss": 2.0503,
+      "step": 13204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.290917990139344e-06,
+      "loss": 2.0911,
+      "step": 13205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.288861128874164e-06,
+      "loss": 2.0928,
+      "step": 13206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.286804437044847e-06,
+      "loss": 2.0549,
+      "step": 13207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.284747914705345e-06,
+      "loss": 2.0332,
+      "step": 13208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.282691561909572e-06,
+      "loss": 2.0864,
+      "step": 13209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.280635378711461e-06,
+      "loss": 2.0916,
+      "step": 13210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.278579365164932e-06,
+      "loss": 2.0757,
+      "step": 13211
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.2765235213239e-06,
+      "loss": 2.1016,
+      "step": 13212
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.274467847242282e-06,
+      "loss": 2.0393,
+      "step": 13213
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.272412342973983e-06,
+      "loss": 2.0308,
+      "step": 13214
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.2703570085729e-06,
+      "loss": 2.0815,
+      "step": 13215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.268301844092943e-06,
+      "loss": 2.0896,
+      "step": 13216
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.266246849587994e-06,
+      "loss": 2.0029,
+      "step": 13217
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.264192025111959e-06,
+      "loss": 2.0125,
+      "step": 13218
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.262137370718708e-06,
+      "loss": 2.0806,
+      "step": 13219
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.260082886462126e-06,
+      "loss": 2.0798,
+      "step": 13220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.258028572396095e-06,
+      "loss": 2.0256,
+      "step": 13221
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.255974428574481e-06,
+      "loss": 2.0167,
+      "step": 13222
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.253920455051154e-06,
+      "loss": 2.0764,
+      "step": 13223
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.251866651879977e-06,
+      "loss": 2.0632,
+      "step": 13224
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.24981301911481e-06,
+      "loss": 2.0776,
+      "step": 13225
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.247759556809503e-06,
+      "loss": 2.0918,
+      "step": 13226
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.24570626501791e-06,
+      "loss": 2.0647,
+      "step": 13227
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.243653143793874e-06,
+      "loss": 2.0728,
+      "step": 13228
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.241600193191242e-06,
+      "loss": 2.0188,
+      "step": 13229
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.239547413263833e-06,
+      "loss": 2.0425,
+      "step": 13230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.237494804065503e-06,
+      "loss": 2.0532,
+      "step": 13231
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.235442365650054e-06,
+      "loss": 2.0559,
+      "step": 13232
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.233390098071331e-06,
+      "loss": 2.0847,
+      "step": 13233
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.231338001383139e-06,
+      "loss": 2.0854,
+      "step": 13234
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.229286075639293e-06,
+      "loss": 2.0469,
+      "step": 13235
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.227234320893609e-06,
+      "loss": 2.0383,
+      "step": 13236
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.225182737199886e-06,
+      "loss": 2.0444,
+      "step": 13237
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.223131324611926e-06,
+      "loss": 2.0464,
+      "step": 13238
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.221080083183524e-06,
+      "loss": 2.0791,
+      "step": 13239
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.219029012968473e-06,
+      "loss": 2.0996,
+      "step": 13240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.216978114020558e-06,
+      "loss": 2.0542,
+      "step": 13241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.214927386393567e-06,
+      "loss": 2.0776,
+      "step": 13242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.212876830141271e-06,
+      "loss": 2.075,
+      "step": 13243
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.210826445317453e-06,
+      "loss": 2.1084,
+      "step": 13244
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.208776231975863e-06,
+      "loss": 2.0449,
+      "step": 13245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.20672619017029e-06,
+      "loss": 2.071,
+      "step": 13246
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.20467631995447e-06,
+      "loss": 2.0945,
+      "step": 13247
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.202626621382183e-06,
+      "loss": 2.0793,
+      "step": 13248
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.200577094507159e-06,
+      "loss": 2.0864,
+      "step": 13249
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.198527739383153e-06,
+      "loss": 2.1106,
+      "step": 13250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.196478556063907e-06,
+      "loss": 2.0437,
+      "step": 13251
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.19442954460316e-06,
+      "loss": 2.0483,
+      "step": 13252
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.192380705054642e-06,
+      "loss": 2.0693,
+      "step": 13253
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.190332037472082e-06,
+      "loss": 2.0923,
+      "step": 13254
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.188283541909203e-06,
+      "loss": 2.0417,
+      "step": 13255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.186235218419727e-06,
+      "loss": 2.0576,
+      "step": 13256
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.184187067057366e-06,
+      "loss": 2.0583,
+      "step": 13257
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.182139087875834e-06,
+      "loss": 2.0759,
+      "step": 13258
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.18009128092884e-06,
+      "loss": 2.0464,
+      "step": 13259
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.17804364627007e-06,
+      "loss": 2.0576,
+      "step": 13260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.175996183953243e-06,
+      "loss": 2.0547,
+      "step": 13261
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.173948894032026e-06,
+      "loss": 2.0938,
+      "step": 13262
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.171901776560132e-06,
+      "loss": 2.0662,
+      "step": 13263
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.169854831591228e-06,
+      "loss": 2.0325,
+      "step": 13264
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.167808059178996e-06,
+      "loss": 2.0769,
+      "step": 13265
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.165761459377112e-06,
+      "loss": 2.1023,
+      "step": 13266
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.163715032239246e-06,
+      "loss": 2.0608,
+      "step": 13267
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.161668777819061e-06,
+      "loss": 2.0437,
+      "step": 13268
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.159622696170226e-06,
+      "loss": 2.0291,
+      "step": 13269
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.157576787346379e-06,
+      "loss": 2.0232,
+      "step": 13270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.155531051401195e-06,
+      "loss": 2.0481,
+      "step": 13271
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.153485488388296e-06,
+      "loss": 2.0776,
+      "step": 13272
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.151440098361341e-06,
+      "loss": 2.0159,
+      "step": 13273
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.149394881373968e-06,
+      "loss": 2.0117,
+      "step": 13274
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.147349837479805e-06,
+      "loss": 2.0237,
+      "step": 13275
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.14530496673249e-06,
+      "loss": 2.0332,
+      "step": 13276
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.143260269185626e-06,
+      "loss": 2.0815,
+      "step": 13277
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.14121574489286e-06,
+      "loss": 2.063,
+      "step": 13278
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.139171393907783e-06,
+      "loss": 2.0071,
+      "step": 13279
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.137127216284027e-06,
+      "loss": 2.0212,
+      "step": 13280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.135083212075183e-06,
+      "loss": 2.0427,
+      "step": 13281
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.133039381334857e-06,
+      "loss": 2.0393,
+      "step": 13282
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.130995724116648e-06,
+      "loss": 2.0681,
+      "step": 13283
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.128952240474147e-06,
+      "loss": 2.0291,
+      "step": 13284
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.126908930460942e-06,
+      "loss": 2.1016,
+      "step": 13285
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.124865794130624e-06,
+      "loss": 2.0781,
+      "step": 13286
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.122822831536751e-06,
+      "loss": 2.0999,
+      "step": 13287
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.120780042732923e-06,
+      "loss": 2.0984,
+      "step": 13288
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.118737427772688e-06,
+      "loss": 2.0493,
+      "step": 13289
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.116694986709631e-06,
+      "loss": 2.0237,
+      "step": 13290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.114652719597296e-06,
+      "loss": 2.1074,
+      "step": 13291
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.112610626489245e-06,
+      "loss": 2.1396,
+      "step": 13292
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.110568707439032e-06,
+      "loss": 2.0945,
+      "step": 13293
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.108526962500203e-06,
+      "loss": 2.094,
+      "step": 13294
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.1064853917263e-06,
+      "loss": 2.0535,
+      "step": 13295
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.104443995170856e-06,
+      "loss": 2.0378,
+      "step": 13296
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.102402772887413e-06,
+      "loss": 2.0647,
+      "step": 13297
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.100361724929494e-06,
+      "loss": 2.0615,
+      "step": 13298
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.098320851350622e-06,
+      "loss": 2.0601,
+      "step": 13299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.096280152204323e-06,
+      "loss": 2.0813,
+      "step": 13300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.09423962754411e-06,
+      "loss": 2.1165,
+      "step": 13301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.092199277423483e-06,
+      "loss": 2.1035,
+      "step": 13302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.090159101895966e-06,
+      "loss": 2.0647,
+      "step": 13303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.088119101015039e-06,
+      "loss": 2.1001,
+      "step": 13304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.086079274834221e-06,
+      "loss": 2.0671,
+      "step": 13305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.084039623406988e-06,
+      "loss": 2.0835,
+      "step": 13306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.082000146786833e-06,
+      "loss": 2.0349,
+      "step": 13307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.07996084502724e-06,
+      "loss": 2.0178,
+      "step": 13308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.077921718181683e-06,
+      "loss": 2.072,
+      "step": 13309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.07588276630364e-06,
+      "loss": 2.0901,
+      "step": 13310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.073843989446577e-06,
+      "loss": 2.0645,
+      "step": 13311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.071805387663961e-06,
+      "loss": 2.0879,
+      "step": 13312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.069766961009253e-06,
+      "loss": 2.0417,
+      "step": 13313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.067728709535904e-06,
+      "loss": 2.084,
+      "step": 13314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.06569063329737e-06,
+      "loss": 2.0693,
+      "step": 13315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.063652732347098e-06,
+      "loss": 2.0784,
+      "step": 13316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.061615006738516e-06,
+      "loss": 2.1313,
+      "step": 13317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.059577456525083e-06,
+      "loss": 2.1255,
+      "step": 13318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.057540081760206e-06,
+      "loss": 2.0906,
+      "step": 13319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.05550288249734e-06,
+      "loss": 2.0952,
+      "step": 13320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.053465858789888e-06,
+      "loss": 2.0757,
+      "step": 13321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.051429010691274e-06,
+      "loss": 2.0818,
+      "step": 13322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.049392338254915e-06,
+      "loss": 2.0356,
+      "step": 13323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.047355841534219e-06,
+      "loss": 2.0364,
+      "step": 13324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.045319520582589e-06,
+      "loss": 2.0662,
+      "step": 13325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.043283375453426e-06,
+      "loss": 2.0886,
+      "step": 13326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.041247406200128e-06,
+      "loss": 2.0693,
+      "step": 13327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.03921161287608e-06,
+      "loss": 2.0771,
+      "step": 13328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.037175995534673e-06,
+      "loss": 2.1221,
+      "step": 13329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.03514055422929e-06,
+      "loss": 2.1191,
+      "step": 13330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.033105289013312e-06,
+      "loss": 2.0764,
+      "step": 13331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.031070199940093e-06,
+      "loss": 2.0735,
+      "step": 13332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.029035287063024e-06,
+      "loss": 2.1052,
+      "step": 13333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.027000550435445e-06,
+      "loss": 2.1323,
+      "step": 13334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.024965990110742e-06,
+      "loss": 2.1208,
+      "step": 13335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.022931606142247e-06,
+      "loss": 2.0974,
+      "step": 13336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.020897398583316e-06,
+      "loss": 2.1116,
+      "step": 13337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.018863367487294e-06,
+      "loss": 2.0972,
+      "step": 13338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.01682951290752e-06,
+      "loss": 2.0425,
+      "step": 13339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.014795834897334e-06,
+      "loss": 2.0291,
+      "step": 13340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.012762333510067e-06,
+      "loss": 2.1292,
+      "step": 13341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.010729008799032e-06,
+      "loss": 2.1255,
+      "step": 13342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.008695860817565e-06,
+      "loss": 2.0654,
+      "step": 13343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.00666288961898e-06,
+      "loss": 2.0776,
+      "step": 13344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.004630095256585e-06,
+      "loss": 2.105,
+      "step": 13345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.002597477783697e-06,
+      "loss": 2.0977,
+      "step": 13346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.0005650372536e-06,
+      "loss": 2.0615,
+      "step": 13347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.99853277371962e-06,
+      "loss": 2.0491,
+      "step": 13348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.996500687235021e-06,
+      "loss": 2.1023,
+      "step": 13349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.994468777853118e-06,
+      "loss": 2.1011,
+      "step": 13350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.99243704562718e-06,
+      "loss": 2.0857,
+      "step": 13351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.99040549061049e-06,
+      "loss": 2.0757,
+      "step": 13352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.988374112856323e-06,
+      "loss": 2.0569,
+      "step": 13353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.986342912417952e-06,
+      "loss": 2.0664,
+      "step": 13354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.984311889348641e-06,
+      "loss": 2.0559,
+      "step": 13355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.982281043701657e-06,
+      "loss": 2.0725,
+      "step": 13356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.980250375530242e-06,
+      "loss": 2.0637,
+      "step": 13357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.978219884887668e-06,
+      "loss": 2.0688,
+      "step": 13358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.976189571827159e-06,
+      "loss": 2.0676,
+      "step": 13359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.974159436401982e-06,
+      "loss": 2.0815,
+      "step": 13360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.972129478665357e-06,
+      "loss": 2.0847,
+      "step": 13361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.970099698670522e-06,
+      "loss": 2.0852,
+      "step": 13362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.96807009647071e-06,
+      "loss": 2.0496,
+      "step": 13363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.966040672119139e-06,
+      "loss": 2.0686,
+      "step": 13364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.964011425669033e-06,
+      "loss": 2.0627,
+      "step": 13365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.961982357173605e-06,
+      "loss": 2.0276,
+      "step": 13366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.959953466686064e-06,
+      "loss": 2.0171,
+      "step": 13367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.957924754259616e-06,
+      "loss": 2.0024,
+      "step": 13368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.95589621994746e-06,
+      "loss": 2.0891,
+      "step": 13369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.953867863802797e-06,
+      "loss": 2.0986,
+      "step": 13370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.95183968587882e-06,
+      "loss": 2.0508,
+      "step": 13371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.9498116862287e-06,
+      "loss": 2.0684,
+      "step": 13372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.947783864905638e-06,
+      "loss": 2.105,
+      "step": 13373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.945756221962795e-06,
+      "loss": 2.1331,
+      "step": 13374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.943728757453359e-06,
+      "loss": 2.0327,
+      "step": 13375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.941701471430487e-06,
+      "loss": 2.075,
+      "step": 13376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.939674363947346e-06,
+      "loss": 2.021,
+      "step": 13377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.937647435057092e-06,
+      "loss": 2.0293,
+      "step": 13378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.935620684812881e-06,
+      "loss": 2.1035,
+      "step": 13379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.933594113267864e-06,
+      "loss": 2.0925,
+      "step": 13380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.93156772047518e-06,
+      "loss": 2.0728,
+      "step": 13381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.929541506487976e-06,
+      "loss": 2.0947,
+      "step": 13382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.927515471359383e-06,
+      "loss": 2.0779,
+      "step": 13383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.92548961514253e-06,
+      "loss": 2.0889,
+      "step": 13384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.923463937890547e-06,
+      "loss": 2.0654,
+      "step": 13385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.921438439656557e-06,
+      "loss": 2.0488,
+      "step": 13386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.91941312049366e-06,
+      "loss": 2.0464,
+      "step": 13387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.917387980454992e-06,
+      "loss": 2.0491,
+      "step": 13388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.915363019593637e-06,
+      "loss": 2.1301,
+      "step": 13389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.913338237962718e-06,
+      "loss": 2.1228,
+      "step": 13390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.911313635615317e-06,
+      "loss": 2.0591,
+      "step": 13391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.909289212604532e-06,
+      "loss": 2.0681,
+      "step": 13392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.907264968983452e-06,
+      "loss": 2.1331,
+      "step": 13393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.90524090480516e-06,
+      "loss": 2.1428,
+      "step": 13394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.903217020122734e-06,
+      "loss": 2.1443,
+      "step": 13395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.901193314989249e-06,
+      "loss": 2.1169,
+      "step": 13396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.899169789457773e-06,
+      "loss": 2.0669,
+      "step": 13397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.89714644358137e-06,
+      "loss": 2.0649,
+      "step": 13398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.895123277413104e-06,
+      "loss": 2.0908,
+      "step": 13399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.893100291006029e-06,
+      "loss": 2.063,
+      "step": 13400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.891077484413193e-06,
+      "loss": 2.0266,
+      "step": 13401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.889054857687641e-06,
+      "loss": 2.0762,
+      "step": 13402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.887032410882422e-06,
+      "loss": 2.0469,
+      "step": 13403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.885010144050558e-06,
+      "loss": 2.0315,
+      "step": 13404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.882988057245095e-06,
+      "loss": 2.0967,
+      "step": 13405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.880966150519046e-06,
+      "loss": 2.0706,
+      "step": 13406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.878944423925451e-06,
+      "loss": 2.0103,
+      "step": 13407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.876922877517312e-06,
+      "loss": 2.0005,
+      "step": 13408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.874901511347647e-06,
+      "loss": 2.0203,
+      "step": 13409
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.872880325469464e-06,
+      "loss": 2.0303,
+      "step": 13410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.870859319935765e-06,
+      "loss": 2.082,
+      "step": 13411
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.86883849479955e-06,
+      "loss": 2.0918,
+      "step": 13412
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.866817850113814e-06,
+      "loss": 2.1301,
+      "step": 13413
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.864797385931544e-06,
+      "loss": 2.1436,
+      "step": 13414
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.862777102305725e-06,
+      "loss": 1.9941,
+      "step": 13415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.860756999289337e-06,
+      "loss": 2.0061,
+      "step": 13416
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.858737076935354e-06,
+      "loss": 2.1091,
+      "step": 13417
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.85671733529675e-06,
+      "loss": 2.052,
+      "step": 13418
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.854697774426478e-06,
+      "loss": 2.1091,
+      "step": 13419
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.852678394377518e-06,
+      "loss": 2.0762,
+      "step": 13420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.850659195202804e-06,
+      "loss": 2.0906,
+      "step": 13421
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.848640176955311e-06,
+      "loss": 2.0991,
+      "step": 13422
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.846621339687966e-06,
+      "loss": 2.0291,
+      "step": 13423
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.844602683453718e-06,
+      "loss": 2.0342,
+      "step": 13424
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.842584208305504e-06,
+      "loss": 2.0947,
+      "step": 13425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.840565914296258e-06,
+      "loss": 2.1006,
+      "step": 13426
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.8385478014789e-06,
+      "loss": 2.083,
+      "step": 13427
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.836529869906367e-06,
+      "loss": 2.0522,
+      "step": 13428
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.834512119631557e-06,
+      "loss": 2.0769,
+      "step": 13429
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.832494550707403e-06,
+      "loss": 2.0801,
+      "step": 13430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.830477163186793e-06,
+      "loss": 2.0981,
+      "step": 13431
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.828459957122656e-06,
+      "loss": 2.0771,
+      "step": 13432
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.826442932567868e-06,
+      "loss": 2.0161,
+      "step": 13433
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.824426089575334e-06,
+      "loss": 2.0278,
+      "step": 13434
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.822409428197942e-06,
+      "loss": 2.0742,
+      "step": 13435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.820392948488573e-06,
+      "loss": 2.0598,
+      "step": 13436
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.818376650500112e-06,
+      "loss": 2.0925,
+      "step": 13437
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.816360534285432e-06,
+      "loss": 2.1052,
+      "step": 13438
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.814344599897402e-06,
+      "loss": 2.0361,
+      "step": 13439
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.81232884738889e-06,
+      "loss": 2.0476,
+      "step": 13440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.810313276812754e-06,
+      "loss": 2.0332,
+      "step": 13441
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.808297888221852e-06,
+      "loss": 2.0571,
+      "step": 13442
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.806282681669041e-06,
+      "loss": 2.0286,
+      "step": 13443
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.804267657207147e-06,
+      "loss": 2.0247,
+      "step": 13444
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.802252814889042e-06,
+      "loss": 2.1194,
+      "step": 13445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.800238154767534e-06,
+      "loss": 2.116,
+      "step": 13446
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.798223676895476e-06,
+      "loss": 2.0781,
+      "step": 13447
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.796209381325685e-06,
+      "loss": 2.1143,
+      "step": 13448
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.794195268110986e-06,
+      "loss": 2.0413,
+      "step": 13449
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.792181337304193e-06,
+      "loss": 2.0325,
+      "step": 13450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.790167588958126e-06,
+      "loss": 2.0461,
+      "step": 13451
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.78815402312559e-06,
+      "loss": 2.0544,
+      "step": 13452
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.786140639859383e-06,
+      "loss": 2.0396,
+      "step": 13453
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.784127439212314e-06,
+      "loss": 2.0715,
+      "step": 13454
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.78211442123717e-06,
+      "loss": 2.0588,
+      "step": 13455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.780101585986741e-06,
+      "loss": 2.0823,
+      "step": 13456
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.778088933513814e-06,
+      "loss": 2.0959,
+      "step": 13457
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.776076463871171e-06,
+      "loss": 2.1157,
+      "step": 13458
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.774064177111572e-06,
+      "loss": 2.0491,
+      "step": 13459
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.772052073287807e-06,
+      "loss": 2.0818,
+      "step": 13460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.77004015245262e-06,
+      "loss": 2.0898,
+      "step": 13461
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.768028414658792e-06,
+      "loss": 2.0867,
+      "step": 13462
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.766016859959067e-06,
+      "loss": 2.0823,
+      "step": 13463
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.764005488406197e-06,
+      "loss": 2.1016,
+      "step": 13464
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.761994300052926e-06,
+      "loss": 2.0767,
+      "step": 13465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.759983294952e-06,
+      "loss": 2.126,
+      "step": 13466
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.757972473156152e-06,
+      "loss": 2.0728,
+      "step": 13467
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.755961834718114e-06,
+      "loss": 2.0522,
+      "step": 13468
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.753951379690613e-06,
+      "loss": 2.0388,
+      "step": 13469
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.751941108126372e-06,
+      "loss": 2.0435,
+      "step": 13470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.749931020078108e-06,
+      "loss": 2.095,
+      "step": 13471
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.747921115598529e-06,
+      "loss": 2.0952,
+      "step": 13472
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.745911394740353e-06,
+      "loss": 2.0652,
+      "step": 13473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.743901857556261e-06,
+      "loss": 2.0566,
+      "step": 13474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.74189250409898e-06,
+      "loss": 1.9917,
+      "step": 13475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.739883334421175e-06,
+      "loss": 1.9897,
+      "step": 13476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.73787434857556e-06,
+      "loss": 2.075,
+      "step": 13477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.735865546614796e-06,
+      "loss": 2.0835,
+      "step": 13478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.733856928591574e-06,
+      "loss": 2.0872,
+      "step": 13479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.731848494558561e-06,
+      "loss": 2.0894,
+      "step": 13480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.729840244568433e-06,
+      "loss": 2.1023,
+      "step": 13481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.727832178673849e-06,
+      "loss": 2.1111,
+      "step": 13482
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.72582429692747e-06,
+      "loss": 2.0403,
+      "step": 13483
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.72381659938195e-06,
+      "loss": 2.0452,
+      "step": 13484
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.721809086089938e-06,
+      "loss": 2.1184,
+      "step": 13485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.71980175710408e-06,
+      "loss": 2.1055,
+      "step": 13486
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.717794612477014e-06,
+      "loss": 2.0337,
+      "step": 13487
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.715787652261384e-06,
+      "loss": 2.054,
+      "step": 13488
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.713780876509797e-06,
+      "loss": 2.0801,
+      "step": 13489
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.711774285274907e-06,
+      "loss": 2.0881,
+      "step": 13490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.709767878609311e-06,
+      "loss": 2.0806,
+      "step": 13491
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.707761656565646e-06,
+      "loss": 2.1074,
+      "step": 13492
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.705755619196505e-06,
+      "loss": 2.1062,
+      "step": 13493
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.703749766554502e-06,
+      "loss": 2.0818,
+      "step": 13494
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.701744098692237e-06,
+      "loss": 2.084,
+      "step": 13495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.699738615662304e-06,
+      "loss": 2.0745,
+      "step": 13496
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.697733317517298e-06,
+      "loss": 2.1306,
+      "step": 13497
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.69572820430981e-06,
+      "loss": 2.1201,
+      "step": 13498
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.693723276092408e-06,
+      "loss": 2.0757,
+      "step": 13499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.691718532917685e-06,
+      "loss": 2.054,
+      "step": 13500
+    },
+    {
+      "Bleu_1": 0.7726067538133373,
+      "Bleu_2": 0.6109768733362893,
+      "Bleu_3": 0.46770382455861814,
+      "Bleu_4": 0.3539354861276422,
+      "CIDEr": 1.1978050069056165,
+      "METEOR": 0.27726135472942437,
+      "ROUGE_L": 0.557683306288871,
+      "epoch": 0.68,
+      "step": 13500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.689713974838194e-06,
+      "loss": 2.0835,
+      "step": 13501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.687709601906525e-06,
+      "loss": 2.092,
+      "step": 13502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.68570541417522e-06,
+      "loss": 2.0881,
+      "step": 13503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.68370141169685e-06,
+      "loss": 2.0972,
+      "step": 13504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.681697594523959e-06,
+      "loss": 2.0903,
+      "step": 13505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.679693962709101e-06,
+      "loss": 2.0781,
+      "step": 13506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.677690516304814e-06,
+      "loss": 2.0415,
+      "step": 13507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.675687255363639e-06,
+      "loss": 2.0603,
+      "step": 13508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.673684179938112e-06,
+      "loss": 2.0933,
+      "step": 13509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.671681290080754e-06,
+      "loss": 2.0886,
+      "step": 13510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.669678585844094e-06,
+      "loss": 2.0212,
+      "step": 13511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.66767606728065e-06,
+      "loss": 2.0181,
+      "step": 13512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.665673734442941e-06,
+      "loss": 2.0825,
+      "step": 13513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.663671587383458e-06,
+      "loss": 2.0718,
+      "step": 13514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.661669626154726e-06,
+      "loss": 2.1179,
+      "step": 13515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.659667850809224e-06,
+      "loss": 2.1235,
+      "step": 13516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.65766626139947e-06,
+      "loss": 2.1235,
+      "step": 13517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.655664857977928e-06,
+      "loss": 2.136,
+      "step": 13518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.653663640597107e-06,
+      "loss": 1.9692,
+      "step": 13519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.651662609309468e-06,
+      "loss": 1.9895,
+      "step": 13520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.649661764167492e-06,
+      "loss": 2.033,
+      "step": 13521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.647661105223649e-06,
+      "loss": 2.0774,
+      "step": 13522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.645660632530405e-06,
+      "loss": 2.0305,
+      "step": 13523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.643660346140218e-06,
+      "loss": 2.0176,
+      "step": 13524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.641660246105546e-06,
+      "loss": 2.0149,
+      "step": 13525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.639660332478835e-06,
+      "loss": 2.012,
+      "step": 13526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.637660605312534e-06,
+      "loss": 2.021,
+      "step": 13527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.635661064659081e-06,
+      "loss": 2.0454,
+      "step": 13528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.633661710570917e-06,
+      "loss": 1.9932,
+      "step": 13529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.631662543100472e-06,
+      "loss": 2.011,
+      "step": 13530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.629663562300156e-06,
+      "loss": 2.0588,
+      "step": 13531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.627664768222417e-06,
+      "loss": 2.0781,
+      "step": 13532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.625666160919644e-06,
+      "loss": 2.0259,
+      "step": 13533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.623667740444273e-06,
+      "loss": 2.0552,
+      "step": 13534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.62166950684869e-06,
+      "loss": 2.0225,
+      "step": 13535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.619671460185306e-06,
+      "loss": 2.0483,
+      "step": 13536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.617673600506519e-06,
+      "loss": 2.0679,
+      "step": 13537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.615675927864713e-06,
+      "loss": 2.0657,
+      "step": 13538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.613678442312284e-06,
+      "loss": 2.0479,
+      "step": 13539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.611681143901603e-06,
+      "loss": 2.0659,
+      "step": 13540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.60968403268506e-06,
+      "loss": 2.0195,
+      "step": 13541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.607687108715014e-06,
+      "loss": 2.0181,
+      "step": 13542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.605690372043837e-06,
+      "loss": 2.072,
+      "step": 13543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.603693822723892e-06,
+      "loss": 2.0962,
+      "step": 13544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.601697460807548e-06,
+      "loss": 2.054,
+      "step": 13545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.599701286347126e-06,
+      "loss": 2.0178,
+      "step": 13546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.597705299395008e-06,
+      "loss": 2.062,
+      "step": 13547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.595709500003509e-06,
+      "loss": 2.0767,
+      "step": 13548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.59371388822499e-06,
+      "loss": 2.0679,
+      "step": 13549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.591718464111768e-06,
+      "loss": 2.0828,
+      "step": 13550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.589723227716171e-06,
+      "loss": 2.0747,
+      "step": 13551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.58772817909053e-06,
+      "loss": 2.0669,
+      "step": 13552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.585733318287155e-06,
+      "loss": 2.0647,
+      "step": 13553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.583738645358364e-06,
+      "loss": 2.0735,
+      "step": 13554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.581744160356464e-06,
+      "loss": 2.0486,
+      "step": 13555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.579749863333755e-06,
+      "loss": 2.0291,
+      "step": 13556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.577755754342538e-06,
+      "loss": 2.0979,
+      "step": 13557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.575761833435109e-06,
+      "loss": 2.0818,
+      "step": 13558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.573768100663752e-06,
+      "loss": 2.0515,
+      "step": 13559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.571774556080752e-06,
+      "loss": 2.0879,
+      "step": 13560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.56978119973838e-06,
+      "loss": 2.0508,
+      "step": 13561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.567788031688927e-06,
+      "loss": 2.0566,
+      "step": 13562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.565795051984638e-06,
+      "loss": 2.1025,
+      "step": 13563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.563802260677803e-06,
+      "loss": 2.1116,
+      "step": 13564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.561809657820659e-06,
+      "loss": 2.0669,
+      "step": 13565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.559817243465467e-06,
+      "loss": 2.0369,
+      "step": 13566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.557825017664479e-06,
+      "loss": 2.0818,
+      "step": 13567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.555832980469935e-06,
+      "loss": 2.0752,
+      "step": 13568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.553841131934075e-06,
+      "loss": 2.0735,
+      "step": 13569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.551849472109138e-06,
+      "loss": 2.0684,
+      "step": 13570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.549858001047337e-06,
+      "loss": 2.0754,
+      "step": 13571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.547866718800918e-06,
+      "loss": 2.0833,
+      "step": 13572
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.545875625422078e-06,
+      "loss": 2.0188,
+      "step": 13573
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.543884720963053e-06,
+      "loss": 2.0256,
+      "step": 13574
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.541894005476038e-06,
+      "loss": 2.1113,
+      "step": 13575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.539903479013238e-06,
+      "loss": 2.1167,
+      "step": 13576
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.537913141626855e-06,
+      "loss": 2.0361,
+      "step": 13577
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.535922993369077e-06,
+      "loss": 2.0659,
+      "step": 13578
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.533933034292112e-06,
+      "loss": 2.031,
+      "step": 13579
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.531943264448124e-06,
+      "loss": 2.0352,
+      "step": 13580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.529953683889304e-06,
+      "loss": 2.0295,
+      "step": 13581
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.527964292667819e-06,
+      "loss": 2.0269,
+      "step": 13582
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.525975090835845e-06,
+      "loss": 2.0679,
+      "step": 13583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.523986078445539e-06,
+      "loss": 2.0532,
+      "step": 13584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.521997255549074e-06,
+      "loss": 2.0747,
+      "step": 13585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.520008622198583e-06,
+      "loss": 2.0635,
+      "step": 13586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.518020178446242e-06,
+      "loss": 2.0706,
+      "step": 13587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.516031924344168e-06,
+      "loss": 2.0837,
+      "step": 13588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.514043859944527e-06,
+      "loss": 2.0774,
+      "step": 13589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.512055985299436e-06,
+      "loss": 2.0842,
+      "step": 13590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.510068300461029e-06,
+      "loss": 2.0806,
+      "step": 13591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.508080805481432e-06,
+      "loss": 2.0908,
+      "step": 13592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.506093500412765e-06,
+      "loss": 2.03,
+      "step": 13593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.50410638530714e-06,
+      "loss": 2.0249,
+      "step": 13594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.502119460216671e-06,
+      "loss": 2.0862,
+      "step": 13595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.500132725193462e-06,
+      "loss": 2.082,
+      "step": 13596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.498146180289608e-06,
+      "loss": 2.1033,
+      "step": 13597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.496159825557209e-06,
+      "loss": 2.0989,
+      "step": 13598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.494173661048356e-06,
+      "loss": 2.0413,
+      "step": 13599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.492187686815133e-06,
+      "loss": 2.0337,
+      "step": 13600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.490201902909608e-06,
+      "loss": 2.0688,
+      "step": 13601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.488216309383878e-06,
+      "loss": 2.0786,
+      "step": 13602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.48623090628999e-06,
+      "loss": 2.0356,
+      "step": 13603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.48424569368003e-06,
+      "loss": 2.0688,
+      "step": 13604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.482260671606043e-06,
+      "loss": 2.0662,
+      "step": 13605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.480275840120083e-06,
+      "loss": 2.0686,
+      "step": 13606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.478291199274213e-06,
+      "loss": 2.0803,
+      "step": 13607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.476306749120467e-06,
+      "loss": 2.0603,
+      "step": 13608
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.474322489710887e-06,
+      "loss": 2.0457,
+      "step": 13609
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.472338421097513e-06,
+      "loss": 2.0698,
+      "step": 13610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.470354543332366e-06,
+      "loss": 2.0649,
+      "step": 13611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.46837085646748e-06,
+      "loss": 2.0715,
+      "step": 13612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.466387360554867e-06,
+      "loss": 1.9617,
+      "step": 13613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.464404055646549e-06,
+      "loss": 2.0046,
+      "step": 13614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.462420941794535e-06,
+      "loss": 2.0073,
+      "step": 13615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.460438019050818e-06,
+      "loss": 2.0264,
+      "step": 13616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.458455287467418e-06,
+      "loss": 2.0339,
+      "step": 13617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.456472747096308e-06,
+      "loss": 2.0457,
+      "step": 13618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.454490397989502e-06,
+      "loss": 2.0652,
+      "step": 13619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.452508240198965e-06,
+      "loss": 2.0403,
+      "step": 13620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.450526273776679e-06,
+      "loss": 2.0337,
+      "step": 13621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.448544498774629e-06,
+      "loss": 2.0659,
+      "step": 13622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.446562915244777e-06,
+      "loss": 2.0684,
+      "step": 13623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.44458152323909e-06,
+      "loss": 2.071,
+      "step": 13624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.442600322809528e-06,
+      "loss": 2.0474,
+      "step": 13625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.440619314008043e-06,
+      "loss": 2.04,
+      "step": 13626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.43863849688659e-06,
+      "loss": 2.0129,
+      "step": 13627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.43665787149711e-06,
+      "loss": 2.0317,
+      "step": 13628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.43467743789154e-06,
+      "loss": 2.0476,
+      "step": 13629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.432697196121822e-06,
+      "loss": 2.0566,
+      "step": 13630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.430717146239874e-06,
+      "loss": 2.0977,
+      "step": 13631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.428737288297639e-06,
+      "loss": 2.0886,
+      "step": 13632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.426757622347012e-06,
+      "loss": 2.0605,
+      "step": 13633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.424778148439933e-06,
+      "loss": 2.0842,
+      "step": 13634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.422798866628287e-06,
+      "loss": 2.0808,
+      "step": 13635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.420819776964001e-06,
+      "loss": 2.0874,
+      "step": 13636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.418840879498957e-06,
+      "loss": 2.0732,
+      "step": 13637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.416862174285059e-06,
+      "loss": 2.0781,
+      "step": 13638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.414883661374192e-06,
+      "loss": 2.0876,
+      "step": 13639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.41290534081824e-06,
+      "loss": 2.0681,
+      "step": 13640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.410927212669085e-06,
+      "loss": 2.1191,
+      "step": 13641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.408949276978604e-06,
+      "loss": 2.1104,
+      "step": 13642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.406971533798652e-06,
+      "loss": 2.1189,
+      "step": 13643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.404993983181114e-06,
+      "loss": 2.1118,
+      "step": 13644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.403016625177826e-06,
+      "loss": 2.0334,
+      "step": 13645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.401039459840666e-06,
+      "loss": 2.0386,
+      "step": 13646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.399062487221464e-06,
+      "loss": 2.033,
+      "step": 13647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.397085707372065e-06,
+      "loss": 2.0356,
+      "step": 13648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.395109120344322e-06,
+      "loss": 2.1238,
+      "step": 13649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.39313272619005e-06,
+      "loss": 2.0913,
+      "step": 13650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.391156524961102e-06,
+      "loss": 2.1138,
+      "step": 13651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.389180516709276e-06,
+      "loss": 2.1367,
+      "step": 13652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.387204701486404e-06,
+      "loss": 2.0967,
+      "step": 13653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.385229079344298e-06,
+      "loss": 2.1006,
+      "step": 13654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.383253650334766e-06,
+      "loss": 2.0837,
+      "step": 13655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.38127841450961e-06,
+      "loss": 2.0703,
+      "step": 13656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.379303371920632e-06,
+      "loss": 2.0898,
+      "step": 13657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.377328522619615e-06,
+      "loss": 2.0735,
+      "step": 13658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.375353866658363e-06,
+      "loss": 2.0725,
+      "step": 13659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.373379404088639e-06,
+      "loss": 2.0879,
+      "step": 13660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.371405134962245e-06,
+      "loss": 2.0171,
+      "step": 13661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.369431059330934e-06,
+      "loss": 2.0208,
+      "step": 13662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.367457177246482e-06,
+      "loss": 2.0803,
+      "step": 13663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.365483488760652e-06,
+      "loss": 2.0913,
+      "step": 13664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.363509993925201e-06,
+      "loss": 2.0574,
+      "step": 13665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.36153669279188e-06,
+      "loss": 2.0635,
+      "step": 13666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.359563585412438e-06,
+      "loss": 2.0996,
+      "step": 13667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.357590671838616e-06,
+      "loss": 2.0779,
+      "step": 13668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.355617952122151e-06,
+      "loss": 2.0183,
+      "step": 13669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.353645426314783e-06,
+      "loss": 2.0515,
+      "step": 13670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.351673094468229e-06,
+      "loss": 2.0645,
+      "step": 13671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.349700956634222e-06,
+      "loss": 2.0889,
+      "step": 13672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.347729012864461e-06,
+      "loss": 2.0208,
+      "step": 13673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.345757263210679e-06,
+      "loss": 2.0308,
+      "step": 13674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.343785707724563e-06,
+      "loss": 2.0742,
+      "step": 13675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.341814346457835e-06,
+      "loss": 2.062,
+      "step": 13676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.339843179462175e-06,
+      "loss": 2.0994,
+      "step": 13677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.337872206789284e-06,
+      "loss": 2.0894,
+      "step": 13678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.335901428490843e-06,
+      "loss": 2.0916,
+      "step": 13679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.333930844618536e-06,
+      "loss": 2.0852,
+      "step": 13680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.331960455224042e-06,
+      "loss": 2.0896,
+      "step": 13681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.329990260359026e-06,
+      "loss": 2.0671,
+      "step": 13682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.32802026007516e-06,
+      "loss": 2.1287,
+      "step": 13683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.326050454424099e-06,
+      "loss": 2.1399,
+      "step": 13684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.324080843457505e-06,
+      "loss": 2.0615,
+      "step": 13685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.322111427227025e-06,
+      "loss": 2.1006,
+      "step": 13686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.320142205784312e-06,
+      "loss": 2.0388,
+      "step": 13687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.31817317918099e-06,
+      "loss": 2.0679,
+      "step": 13688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.31620434746871e-06,
+      "loss": 2.0923,
+      "step": 13689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.31423571069909e-06,
+      "loss": 2.0962,
+      "step": 13690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.312267268923774e-06,
+      "loss": 2.0261,
+      "step": 13691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.310299022194363e-06,
+      "loss": 2.0361,
+      "step": 13692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.30833097056248e-06,
+      "loss": 2.0825,
+      "step": 13693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.306363114079733e-06,
+      "loss": 2.0751,
+      "step": 13694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.304395452797731e-06,
+      "loss": 2.041,
+      "step": 13695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.302427986768068e-06,
+      "loss": 2.0703,
+      "step": 13696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.300460716042341e-06,
+      "loss": 2.1116,
+      "step": 13697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.298493640672141e-06,
+      "loss": 2.1121,
+      "step": 13698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.29652676070905e-06,
+      "loss": 2.0562,
+      "step": 13699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.294560076204646e-06,
+      "loss": 2.0671,
+      "step": 13700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.292593587210507e-06,
+      "loss": 1.9514,
+      "step": 13701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.290627293778204e-06,
+      "loss": 1.9646,
+      "step": 13702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.288661195959288e-06,
+      "loss": 2.0813,
+      "step": 13703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.286695293805334e-06,
+      "loss": 2.0903,
+      "step": 13704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.284729587367877e-06,
+      "loss": 2.0659,
+      "step": 13705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.282764076698489e-06,
+      "loss": 2.0588,
+      "step": 13706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.280798761848692e-06,
+      "loss": 1.9902,
+      "step": 13707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.278833642870033e-06,
+      "loss": 2.0229,
+      "step": 13708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.276868719814048e-06,
+      "loss": 2.0911,
+      "step": 13709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.274903992732258e-06,
+      "loss": 2.0801,
+      "step": 13710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.272939461676187e-06,
+      "loss": 2.0615,
+      "step": 13711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.270975126697365e-06,
+      "loss": 2.0642,
+      "step": 13712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.269010987847278e-06,
+      "loss": 2.0823,
+      "step": 13713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.267047045177462e-06,
+      "loss": 2.0627,
+      "step": 13714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.265083298739395e-06,
+      "loss": 1.9839,
+      "step": 13715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.263119748584589e-06,
+      "loss": 1.9731,
+      "step": 13716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.26115639476454e-06,
+      "loss": 2.0745,
+      "step": 13717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.259193237330715e-06,
+      "loss": 2.0723,
+      "step": 13718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.257230276334617e-06,
+      "loss": 2.0957,
+      "step": 13719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.255267511827704e-06,
+      "loss": 2.0898,
+      "step": 13720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.253304943861466e-06,
+      "loss": 2.0715,
+      "step": 13721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.251342572487351e-06,
+      "loss": 2.0952,
+      "step": 13722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.249380397756829e-06,
+      "loss": 2.0688,
+      "step": 13723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.247418419721355e-06,
+      "loss": 2.0974,
+      "step": 13724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.245456638432379e-06,
+      "loss": 2.0659,
+      "step": 13725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.243495053941346e-06,
+      "loss": 2.0742,
+      "step": 13726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.241533666299704e-06,
+      "loss": 2.0859,
+      "step": 13727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.239572475558867e-06,
+      "loss": 2.1047,
+      "step": 13728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.237611481770293e-06,
+      "loss": 2.0571,
+      "step": 13729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.235650684985378e-06,
+      "loss": 2.0652,
+      "step": 13730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.23369008525557e-06,
+      "loss": 2.0874,
+      "step": 13731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.231729682632265e-06,
+      "loss": 2.0989,
+      "step": 13732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.229769477166874e-06,
+      "loss": 2.0642,
+      "step": 13733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.227809468910806e-06,
+      "loss": 2.0576,
+      "step": 13734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.225849657915458e-06,
+      "loss": 2.0608,
+      "step": 13735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.223890044232225e-06,
+      "loss": 2.0649,
+      "step": 13736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.221930627912494e-06,
+      "loss": 2.0042,
+      "step": 13737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.219971409007646e-06,
+      "loss": 2.064,
+      "step": 13738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.218012387569066e-06,
+      "loss": 2.1118,
+      "step": 13739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.21605356364812e-06,
+      "loss": 2.1431,
+      "step": 13740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.214094937296182e-06,
+      "loss": 2.0059,
+      "step": 13741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.212136508564613e-06,
+      "loss": 1.9983,
+      "step": 13742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.210178277504763e-06,
+      "loss": 2.0215,
+      "step": 13743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.208220244168e-06,
+      "loss": 2.0059,
+      "step": 13744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.206262408605649e-06,
+      "loss": 2.0647,
+      "step": 13745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.204304770869078e-06,
+      "loss": 2.075,
+      "step": 13746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.202347331009603e-06,
+      "loss": 1.9951,
+      "step": 13747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.200390089078563e-06,
+      "loss": 2.0415,
+      "step": 13748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.198433045127288e-06,
+      "loss": 2.0579,
+      "step": 13749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.196476199207092e-06,
+      "loss": 2.0815,
+      "step": 13750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.194519551369295e-06,
+      "loss": 2.0525,
+      "step": 13751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.192563101665211e-06,
+      "loss": 2.0571,
+      "step": 13752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.190606850146138e-06,
+      "loss": 2.1108,
+      "step": 13753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.188650796863384e-06,
+      "loss": 2.1042,
+      "step": 13754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.18669494186824e-06,
+      "loss": 1.9937,
+      "step": 13755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.184739285211995e-06,
+      "loss": 2.009,
+      "step": 13756
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.182783826945938e-06,
+      "loss": 2.0957,
+      "step": 13757
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.180828567121345e-06,
+      "loss": 2.104,
+      "step": 13758
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.178873505789495e-06,
+      "loss": 2.1401,
+      "step": 13759
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.176918643001647e-06,
+      "loss": 2.134,
+      "step": 13760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.174963978809081e-06,
+      "loss": 2.0974,
+      "step": 13761
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.173009513263037e-06,
+      "loss": 2.1064,
+      "step": 13762
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.171055246414787e-06,
+      "loss": 2.0232,
+      "step": 13763
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.169101178315567e-06,
+      "loss": 2.0813,
+      "step": 13764
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.167147309016624e-06,
+      "loss": 2.0715,
+      "step": 13765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.165193638569196e-06,
+      "loss": 2.0718,
+      "step": 13766
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.163240167024515e-06,
+      "loss": 2.0378,
+      "step": 13767
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.161286894433806e-06,
+      "loss": 2.0374,
+      "step": 13768
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.1593338208483e-06,
+      "loss": 2.0422,
+      "step": 13769
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.157380946319206e-06,
+      "loss": 2.0481,
+      "step": 13770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.155428270897736e-06,
+      "loss": 2.0813,
+      "step": 13771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.153475794635104e-06,
+      "loss": 2.0623,
+      "step": 13772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.151523517582502e-06,
+      "loss": 2.1042,
+      "step": 13773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.149571439791137e-06,
+      "loss": 2.0947,
+      "step": 13774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.147619561312187e-06,
+      "loss": 2.0435,
+      "step": 13775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.14566788219685e-06,
+      "loss": 2.0493,
+      "step": 13776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.143716402496292e-06,
+      "loss": 2.0549,
+      "step": 13777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.141765122261708e-06,
+      "loss": 2.0774,
+      "step": 13778
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.139814041544252e-06,
+      "loss": 2.0642,
+      "step": 13779
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.137863160395093e-06,
+      "loss": 2.0759,
+      "step": 13780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.135912478865393e-06,
+      "loss": 2.0615,
+      "step": 13781
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.133961997006306e-06,
+      "loss": 2.0447,
+      "step": 13782
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.132011714868978e-06,
+      "loss": 2.0547,
+      "step": 13783
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.13006163250456e-06,
+      "loss": 2.0688,
+      "step": 13784
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.128111749964177e-06,
+      "loss": 2.1079,
+      "step": 13785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.126162067298976e-06,
+      "loss": 2.0923,
+      "step": 13786
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.12421258456008e-06,
+      "loss": 2.0435,
+      "step": 13787
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.12226330179861e-06,
+      "loss": 2.074,
+      "step": 13788
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.120314219065696e-06,
+      "loss": 2.0598,
+      "step": 13789
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.118365336412426e-06,
+      "loss": 2.0823,
+      "step": 13790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.116416653889933e-06,
+      "loss": 2.0686,
+      "step": 13791
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.114468171549296e-06,
+      "loss": 2.1191,
+      "step": 13792
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.112519889441635e-06,
+      "loss": 2.0439,
+      "step": 13793
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.110571807618023e-06,
+      "loss": 2.074,
+      "step": 13794
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.108623926129555e-06,
+      "loss": 1.9812,
+      "step": 13795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.106676245027307e-06,
+      "loss": 1.9856,
+      "step": 13796
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.10472876436236e-06,
+      "loss": 1.9954,
+      "step": 13797
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.102781484185779e-06,
+      "loss": 1.9912,
+      "step": 13798
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.10083440454864e-06,
+      "loss": 2.0444,
+      "step": 13799
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.098887525501982e-06,
+      "loss": 2.0647,
+      "step": 13800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.096940847096884e-06,
+      "loss": 2.0583,
+      "step": 13801
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.094994369384373e-06,
+      "loss": 2.0437,
+      "step": 13802
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.093048092415514e-06,
+      "loss": 2.052,
+      "step": 13803
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.091102016241332e-06,
+      "loss": 2.0442,
+      "step": 13804
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.089156140912864e-06,
+      "loss": 2.0344,
+      "step": 13805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.087210466481137e-06,
+      "loss": 2.0537,
+      "step": 13806
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.085264992997178e-06,
+      "loss": 2.0498,
+      "step": 13807
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.083319720512001e-06,
+      "loss": 2.0354,
+      "step": 13808
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.081374649076622e-06,
+      "loss": 2.0659,
+      "step": 13809
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.079429778742044e-06,
+      "loss": 2.0657,
+      "step": 13810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.077485109559272e-06,
+      "loss": 2.0789,
+      "step": 13811
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.075540641579304e-06,
+      "loss": 2.0911,
+      "step": 13812
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.073596374853129e-06,
+      "loss": 2.0935,
+      "step": 13813
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.071652309431737e-06,
+      "loss": 2.113,
+      "step": 13814
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.069708445366095e-06,
+      "loss": 2.009,
+      "step": 13815
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.067764782707201e-06,
+      "loss": 2.0244,
+      "step": 13816
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.065821321506003e-06,
+      "loss": 2.1018,
+      "step": 13817
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.063878061813485e-06,
+      "loss": 2.1055,
+      "step": 13818
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.061935003680593e-06,
+      "loss": 2.0698,
+      "step": 13819
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.059992147158287e-06,
+      "loss": 2.0886,
+      "step": 13820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.058049492297516e-06,
+      "loss": 2.0176,
+      "step": 13821
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.056107039149224e-06,
+      "loss": 2.0408,
+      "step": 13822
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.054164787764349e-06,
+      "loss": 2.1025,
+      "step": 13823
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.052222738193824e-06,
+      "loss": 2.1003,
+      "step": 13824
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.050280890488577e-06,
+      "loss": 2.0198,
+      "step": 13825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.048339244699532e-06,
+      "loss": 2.033,
+      "step": 13826
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.046397800877603e-06,
+      "loss": 2.0828,
+      "step": 13827
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.044456559073707e-06,
+      "loss": 2.0803,
+      "step": 13828
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.042515519338755e-06,
+      "loss": 2.0784,
+      "step": 13829
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.040574681723627e-06,
+      "loss": 2.0789,
+      "step": 13830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.038634046279249e-06,
+      "loss": 2.0928,
+      "step": 13831
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.036693613056485e-06,
+      "loss": 2.0811,
+      "step": 13832
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.034753382106242e-06,
+      "loss": 2.0623,
+      "step": 13833
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.032813353479388e-06,
+      "loss": 2.073,
+      "step": 13834
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.030873527226799e-06,
+      "loss": 2.0947,
+      "step": 13835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.028933903399348e-06,
+      "loss": 2.1038,
+      "step": 13836
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.026994482047897e-06,
+      "loss": 2.0698,
+      "step": 13837
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.025055263223307e-06,
+      "loss": 2.0554,
+      "step": 13838
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.02311624697643e-06,
+      "loss": 2.0715,
+      "step": 13839
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.021177433358116e-06,
+      "loss": 2.0518,
+      "step": 13840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.019238822419208e-06,
+      "loss": 2.0967,
+      "step": 13841
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.017300414210543e-06,
+      "loss": 2.0964,
+      "step": 13842
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.015362208782951e-06,
+      "loss": 2.0422,
+      "step": 13843
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.013424206187273e-06,
+      "loss": 2.0591,
+      "step": 13844
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.011486406474306e-06,
+      "loss": 2.0718,
+      "step": 13845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.00954880969489e-06,
+      "loss": 2.062,
+      "step": 13846
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.007611415899819e-06,
+      "loss": 2.0571,
+      "step": 13847
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.005674225139915e-06,
+      "loss": 2.0552,
+      "step": 13848
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.003737237465965e-06,
+      "loss": 2.0222,
+      "step": 13849
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.001800452928768e-06,
+      "loss": 2.021,
+      "step": 13850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.999863871579118e-06,
+      "loss": 2.0723,
+      "step": 13851
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.997927493467796e-06,
+      "loss": 2.0759,
+      "step": 13852
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.995991318645582e-06,
+      "loss": 2.1248,
+      "step": 13853
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.994055347163256e-06,
+      "loss": 2.1052,
+      "step": 13854
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.992119579071568e-06,
+      "loss": 2.0381,
+      "step": 13855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.990184014421303e-06,
+      "loss": 2.0608,
+      "step": 13856
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.988248653263207e-06,
+      "loss": 2.0618,
+      "step": 13857
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.986313495648038e-06,
+      "loss": 2.0452,
+      "step": 13858
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.984378541626544e-06,
+      "loss": 2.062,
+      "step": 13859
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.982443791249452e-06,
+      "loss": 2.0684,
+      "step": 13860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.980509244567523e-06,
+      "loss": 2.0847,
+      "step": 13861
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.978574901631465e-06,
+      "loss": 2.0828,
+      "step": 13862
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.976640762492026e-06,
+      "loss": 2.0935,
+      "step": 13863
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.974706827199908e-06,
+      "loss": 2.093,
+      "step": 13864
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.972773095805832e-06,
+      "loss": 2.0874,
+      "step": 13865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.970839568360512e-06,
+      "loss": 2.0938,
+      "step": 13866
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.968906244914648e-06,
+      "loss": 2.0002,
+      "step": 13867
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.966973125518942e-06,
+      "loss": 1.9927,
+      "step": 13868
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.965040210224091e-06,
+      "loss": 2.0532,
+      "step": 13869
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.963107499080768e-06,
+      "loss": 2.0698,
+      "step": 13870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.961174992139677e-06,
+      "loss": 2.0486,
+      "step": 13871
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.959242689451478e-06,
+      "loss": 2.0959,
+      "step": 13872
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.95731059106686e-06,
+      "loss": 2.0801,
+      "step": 13873
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.955378697036477e-06,
+      "loss": 2.0615,
+      "step": 13874
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.953447007410996e-06,
+      "loss": 2.0229,
+      "step": 13875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.95151552224107e-06,
+      "loss": 2.042,
+      "step": 13876
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.949584241577351e-06,
+      "loss": 2.1204,
+      "step": 13877
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.94765316547049e-06,
+      "loss": 2.1111,
+      "step": 13878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.945722293971122e-06,
+      "loss": 2.0623,
+      "step": 13879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.943791627129881e-06,
+      "loss": 2.0693,
+      "step": 13880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.941861164997401e-06,
+      "loss": 2.0469,
+      "step": 13881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.939930907624301e-06,
+      "loss": 2.0339,
+      "step": 13882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.938000855061204e-06,
+      "loss": 2.0769,
+      "step": 13883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.93607100735872e-06,
+      "loss": 2.0752,
+      "step": 13884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.934141364567461e-06,
+      "loss": 2.1057,
+      "step": 13885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.932211926738031e-06,
+      "loss": 2.0981,
+      "step": 13886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.930282693921015e-06,
+      "loss": 2.0352,
+      "step": 13887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.928353666167022e-06,
+      "loss": 2.0391,
+      "step": 13888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.926424843526622e-06,
+      "loss": 2.0808,
+      "step": 13889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.924496226050413e-06,
+      "loss": 2.1018,
+      "step": 13890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.922567813788956e-06,
+      "loss": 2.1077,
+      "step": 13891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.92063960679283e-06,
+      "loss": 2.0886,
+      "step": 13892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.918711605112597e-06,
+      "loss": 2.0979,
+      "step": 13893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.916783808798815e-06,
+      "loss": 2.0886,
+      "step": 13894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.914856217902042e-06,
+      "loss": 2.0618,
+      "step": 13895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.912928832472824e-06,
+      "loss": 2.0713,
+      "step": 13896
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.911001652561706e-06,
+      "loss": 2.0791,
+      "step": 13897
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.909074678219226e-06,
+      "loss": 2.0688,
+      "step": 13898
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.907147909495912e-06,
+      "loss": 2.0652,
+      "step": 13899
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 7.9052213464423e-06,
+      "loss": 2.0691,
+      "step": 13900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.903294989108912e-06,
+      "loss": 2.1086,
+      "step": 13901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.901368837546247e-06,
+      "loss": 2.0867,
+      "step": 13902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.899442891804843e-06,
+      "loss": 2.1016,
+      "step": 13903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.89751715193518e-06,
+      "loss": 2.0911,
+      "step": 13904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.895591617987779e-06,
+      "loss": 2.0493,
+      "step": 13905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.893666290013121e-06,
+      "loss": 2.0461,
+      "step": 13906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.891741168061701e-06,
+      "loss": 2.0486,
+      "step": 13907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.889816252184001e-06,
+      "loss": 2.0784,
+      "step": 13908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.887891542430502e-06,
+      "loss": 2.0691,
+      "step": 13909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.885967038851677e-06,
+      "loss": 2.0674,
+      "step": 13910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.884042741497995e-06,
+      "loss": 2.0457,
+      "step": 13911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.882118650419911e-06,
+      "loss": 2.0454,
+      "step": 13912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.880194765667892e-06,
+      "loss": 2.0994,
+      "step": 13913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.878271087292384e-06,
+      "loss": 2.0515,
+      "step": 13914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.876347615343837e-06,
+      "loss": 2.074,
+      "step": 13915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.874424349872693e-06,
+      "loss": 2.0847,
+      "step": 13916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.872501290929372e-06,
+      "loss": 2.0593,
+      "step": 13917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.870578438564328e-06,
+      "loss": 2.0886,
+      "step": 13918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.868655792827963e-06,
+      "loss": 2.0886,
+      "step": 13919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.866733353770716e-06,
+      "loss": 2.0642,
+      "step": 13920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.864811121442986e-06,
+      "loss": 2.0618,
+      "step": 13921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.862889095895187e-06,
+      "loss": 2.0833,
+      "step": 13922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.860967277177721e-06,
+      "loss": 2.0735,
+      "step": 13923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.859045665340987e-06,
+      "loss": 2.0557,
+      "step": 13924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.857124260435378e-06,
+      "loss": 2.0464,
+      "step": 13925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.855203062511276e-06,
+      "loss": 2.0664,
+      "step": 13926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.853282071619064e-06,
+      "loss": 2.061,
+      "step": 13927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.851361287809121e-06,
+      "loss": 2.0725,
+      "step": 13928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.849440711131814e-06,
+      "loss": 2.032,
+      "step": 13929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.847520341637507e-06,
+      "loss": 2.0247,
+      "step": 13930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.845600179376569e-06,
+      "loss": 2.075,
+      "step": 13931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.843680224399336e-06,
+      "loss": 2.0881,
+      "step": 13932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.841760476756174e-06,
+      "loss": 2.0332,
+      "step": 13933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.839840936497411e-06,
+      "loss": 2.0112,
+      "step": 13934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.837921603673403e-06,
+      "loss": 2.064,
+      "step": 13935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.836002478334467e-06,
+      "loss": 2.0752,
+      "step": 13936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.834083560530934e-06,
+      "loss": 2.0552,
+      "step": 13937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.832164850313129e-06,
+      "loss": 2.0486,
+      "step": 13938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.830246347731361e-06,
+      "loss": 2.0881,
+      "step": 13939
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.828328052835948e-06,
+      "loss": 2.0642,
+      "step": 13940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.826409965677196e-06,
+      "loss": 2.0593,
+      "step": 13941
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.824492086305389e-06,
+      "loss": 2.062,
+      "step": 13942
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.822574414770846e-06,
+      "loss": 2.0776,
+      "step": 13943
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.820656951123827e-06,
+      "loss": 2.0564,
+      "step": 13944
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.818739695414645e-06,
+      "loss": 2.1042,
+      "step": 13945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.816822647693556e-06,
+      "loss": 2.1104,
+      "step": 13946
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.814905808010838e-06,
+      "loss": 2.1277,
+      "step": 13947
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.812989176416764e-06,
+      "loss": 2.1206,
+      "step": 13948
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.811072752961589e-06,
+      "loss": 2.0442,
+      "step": 13949
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.80915653769557e-06,
+      "loss": 2.0088,
+      "step": 13950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.807240530668958e-06,
+      "loss": 2.0703,
+      "step": 13951
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.805324731932002e-06,
+      "loss": 2.072,
+      "step": 13952
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.803409141534935e-06,
+      "loss": 2.0813,
+      "step": 13953
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.801493759527997e-06,
+      "loss": 2.0857,
+      "step": 13954
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.79957858596141e-06,
+      "loss": 2.0288,
+      "step": 13955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.79766362088541e-06,
+      "loss": 2.0476,
+      "step": 13956
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.795748864350196e-06,
+      "loss": 2.0981,
+      "step": 13957
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.793834316405996e-06,
+      "loss": 2.0833,
+      "step": 13958
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.791919977103004e-06,
+      "loss": 2.1118,
+      "step": 13959
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.790005846491441e-06,
+      "loss": 2.0977,
+      "step": 13960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.788091924621482e-06,
+      "loss": 2.0581,
+      "step": 13961
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.786178211543324e-06,
+      "loss": 2.0793,
+      "step": 13962
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.784264707307155e-06,
+      "loss": 2.0342,
+      "step": 13963
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.782351411963154e-06,
+      "loss": 2.0278,
+      "step": 13964
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.780438325561492e-06,
+      "loss": 2.0798,
+      "step": 13965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.778525448152342e-06,
+      "loss": 2.04,
+      "step": 13966
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.776612779785864e-06,
+      "loss": 2.0745,
+      "step": 13967
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.77470032051221e-06,
+      "loss": 2.0645,
+      "step": 13968
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.772788070381545e-06,
+      "loss": 2.114,
+      "step": 13969
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.770876029444005e-06,
+      "loss": 2.0842,
+      "step": 13970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.76896419774974e-06,
+      "loss": 2.0442,
+      "step": 13971
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.767052575348868e-06,
+      "loss": 2.0535,
+      "step": 13972
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.765141162291547e-06,
+      "loss": 2.0754,
+      "step": 13973
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.76322995862787e-06,
+      "loss": 2.0818,
+      "step": 13974
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.761318964407984e-06,
+      "loss": 2.0864,
+      "step": 13975
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.759408179681984e-06,
+      "loss": 2.0776,
+      "step": 13976
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.757497604499987e-06,
+      "loss": 2.0535,
+      "step": 13977
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.75558723891209e-06,
+      "loss": 2.0505,
+      "step": 13978
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.753677082968395e-06,
+      "loss": 2.0688,
+      "step": 13979
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.751767136718992e-06,
+      "loss": 2.0686,
+      "step": 13980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.749857400213966e-06,
+      "loss": 2.0642,
+      "step": 13981
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.747947873503399e-06,
+      "loss": 2.0762,
+      "step": 13982
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.746038556637368e-06,
+      "loss": 2.1052,
+      "step": 13983
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.744129449665936e-06,
+      "loss": 2.0918,
+      "step": 13984
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.742220552639175e-06,
+      "loss": 2.1077,
+      "step": 13985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.740311865607144e-06,
+      "loss": 2.124,
+      "step": 13986
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.738403388619882e-06,
+      "loss": 2.0588,
+      "step": 13987
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.736495121727458e-06,
+      "loss": 2.095,
+      "step": 13988
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.73458706497989e-06,
+      "loss": 2.0188,
+      "step": 13989
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.732679218427238e-06,
+      "loss": 1.99,
+      "step": 13990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.730771582119522e-06,
+      "loss": 2.114,
+      "step": 13991
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.728864156106764e-06,
+      "loss": 2.0974,
+      "step": 13992
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.72695694043899e-06,
+      "loss": 2.061,
+      "step": 13993
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.72504993516621e-06,
+      "loss": 2.0803,
+      "step": 13994
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.723143140338436e-06,
+      "loss": 2.051,
+      "step": 13995
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.72123655600567e-06,
+      "loss": 2.0632,
+      "step": 13996
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.719330182217915e-06,
+      "loss": 2.0916,
+      "step": 13997
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.717424019025157e-06,
+      "loss": 2.0769,
+      "step": 13998
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.715518066477385e-06,
+      "loss": 2.1079,
+      "step": 13999
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.713612324624584e-06,
+      "loss": 2.1006,
+      "step": 14000
+    },
+    {
+      "Bleu_1": 0.773076169991568,
+      "Bleu_2": 0.6122494501857746,
+      "Bleu_3": 0.46986436892518385,
+      "Bleu_4": 0.35615712489690937,
+      "CIDEr": 1.201508050391189,
+      "METEOR": 0.2780962694862662,
+      "ROUGE_L": 0.5589928239646759,
+      "epoch": 0.7,
+      "step": 14000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.711706793516722e-06,
+      "loss": 2.0471,
+      "step": 14001
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.709801473203777e-06,
+      "loss": 2.0574,
+      "step": 14002
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.707896363735716e-06,
+      "loss": 2.0515,
+      "step": 14003
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.70599146516248e-06,
+      "loss": 2.0339,
+      "step": 14004
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.70408677753405e-06,
+      "loss": 2.0566,
+      "step": 14005
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.702182300900346e-06,
+      "loss": 2.093,
+      "step": 14006
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.70027803531134e-06,
+      "loss": 2.0664,
+      "step": 14007
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.698373980816945e-06,
+      "loss": 2.0347,
+      "step": 14008
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.696470137467103e-06,
+      "loss": 2.0586,
+      "step": 14009
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.694566505311736e-06,
+      "loss": 2.0737,
+      "step": 14010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.692663084400769e-06,
+      "loss": 2.0752,
+      "step": 14011
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.690759874784117e-06,
+      "loss": 2.064,
+      "step": 14012
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.688856876511691e-06,
+      "loss": 2.0427,
+      "step": 14013
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.68695408963338e-06,
+      "loss": 2.0735,
+      "step": 14014
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.685051514199107e-06,
+      "loss": 2.0366,
+      "step": 14015
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.68314915025874e-06,
+      "loss": 2.0493,
+      "step": 14016
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.68124699786219e-06,
+      "loss": 2.1018,
+      "step": 14017
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.67934505705932e-06,
+      "loss": 2.1106,
+      "step": 14018
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.677443327900013e-06,
+      "loss": 2.0332,
+      "step": 14019
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.67554181043414e-06,
+      "loss": 2.0332,
+      "step": 14020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.67364050471156e-06,
+      "loss": 2.0105,
+      "step": 14021
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.671739410782148e-06,
+      "loss": 2.0322,
+      "step": 14022
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.669838528695746e-06,
+      "loss": 2.0815,
+      "step": 14023
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.667937858502201e-06,
+      "loss": 2.064,
+      "step": 14024
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.666037400251362e-06,
+      "loss": 2.0439,
+      "step": 14025
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.664137153993063e-06,
+      "loss": 2.0657,
+      "step": 14026
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.662237119777137e-06,
+      "loss": 2.0901,
+      "step": 14027
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.660337297653413e-06,
+      "loss": 2.0994,
+      "step": 14028
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.658437687671702e-06,
+      "loss": 2.031,
+      "step": 14029
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.656538289881833e-06,
+      "loss": 1.9885,
+      "step": 14030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.654639104333598e-06,
+      "loss": 2.0254,
+      "step": 14031
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.652740131076824e-06,
+      "loss": 2.0566,
+      "step": 14032
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.650841370161287e-06,
+      "loss": 2.0588,
+      "step": 14033
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.648942821636791e-06,
+      "loss": 2.0527,
+      "step": 14034
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.647044485553121e-06,
+      "loss": 2.0876,
+      "step": 14035
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.64514636196006e-06,
+      "loss": 2.084,
+      "step": 14036
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.643248450907384e-06,
+      "loss": 2.0615,
+      "step": 14037
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.64135075244486e-06,
+      "loss": 2.0686,
+      "step": 14038
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.639453266622257e-06,
+      "loss": 2.0376,
+      "step": 14039
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.637555993489331e-06,
+      "loss": 2.0559,
+      "step": 14040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.63565893309584e-06,
+      "loss": 2.1169,
+      "step": 14041
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.633762085491527e-06,
+      "loss": 2.0991,
+      "step": 14042
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.631865450726148e-06,
+      "loss": 2.0535,
+      "step": 14043
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.629969028849415e-06,
+      "loss": 2.0454,
+      "step": 14044
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.628072819911087e-06,
+      "loss": 2.0581,
+      "step": 14045
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.626176823960865e-06,
+      "loss": 2.0908,
+      "step": 14046
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.624281041048489e-06,
+      "loss": 2.0496,
+      "step": 14047
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.622385471223665e-06,
+      "loss": 2.0535,
+      "step": 14048
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.620490114536101e-06,
+      "loss": 2.0559,
+      "step": 14049
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.618594971035503e-06,
+      "loss": 2.0667,
+      "step": 14050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.6167000407715695e-06,
+      "loss": 2.0896,
+      "step": 14051
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.614805323793992e-06,
+      "loss": 2.075,
+      "step": 14052
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.612910820152456e-06,
+      "loss": 2.0422,
+      "step": 14053
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.611016529896643e-06,
+      "loss": 2.0488,
+      "step": 14054
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.60912245307623e-06,
+      "loss": 2.1267,
+      "step": 14055
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.607228589740886e-06,
+      "loss": 2.094,
+      "step": 14056
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.605334939940278e-06,
+      "loss": 2.0452,
+      "step": 14057
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.603441503724066e-06,
+      "loss": 2.0437,
+      "step": 14058
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.601548281141888e-06,
+      "loss": 2.0999,
+      "step": 14059
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.599655272243415e-06,
+      "loss": 2.082,
+      "step": 14060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.597762477078266e-06,
+      "loss": 2.0312,
+      "step": 14061
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.595869895696101e-06,
+      "loss": 2.0491,
+      "step": 14062
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.593977528146532e-06,
+      "loss": 2.0515,
+      "step": 14063
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.59208537447919e-06,
+      "loss": 2.0627,
+      "step": 14064
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.590193434743694e-06,
+      "loss": 2.0972,
+      "step": 14065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.58830170898966e-06,
+      "loss": 2.083,
+      "step": 14066
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.586410197266695e-06,
+      "loss": 2.0632,
+      "step": 14067
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.584518899624401e-06,
+      "loss": 2.0562,
+      "step": 14068
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.582627816112377e-06,
+      "loss": 2.054,
+      "step": 14069
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.580736946780213e-06,
+      "loss": 2.0613,
+      "step": 14070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.578846291677496e-06,
+      "loss": 2.0859,
+      "step": 14071
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.576955850853804e-06,
+      "loss": 2.0886,
+      "step": 14072
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5750656243587205e-06,
+      "loss": 2.0684,
+      "step": 14073
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5731756122417955e-06,
+      "loss": 2.0605,
+      "step": 14074
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.571285814552614e-06,
+      "loss": 2.1169,
+      "step": 14075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.569396231340716e-06,
+      "loss": 2.1125,
+      "step": 14076
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.567506862655668e-06,
+      "loss": 2.0801,
+      "step": 14077
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.565617708547007e-06,
+      "loss": 2.0664,
+      "step": 14078
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.563728769064275e-06,
+      "loss": 2.0254,
+      "step": 14079
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.56184004425701e-06,
+      "loss": 2.0796,
+      "step": 14080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.559951534174742e-06,
+      "loss": 2.0593,
+      "step": 14081
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.558063238866992e-06,
+      "loss": 2.0178,
+      "step": 14082
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.556175158383286e-06,
+      "loss": 2.0315,
+      "step": 14083
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.554287292773121e-06,
+      "loss": 2.0698,
+      "step": 14084
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.552399642086022e-06,
+      "loss": 2.0378,
+      "step": 14085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5505122063714735e-06,
+      "loss": 2.0266,
+      "step": 14086
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.54862498567899e-06,
+      "loss": 2.0229,
+      "step": 14087
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5467379800580465e-06,
+      "loss": 2.0195,
+      "step": 14088
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.544851189558133e-06,
+      "loss": 2.0259,
+      "step": 14089
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.542964614228729e-06,
+      "loss": 2.0171,
+      "step": 14090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5410782541193e-06,
+      "loss": 2.0369,
+      "step": 14091
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.53919210927933e-06,
+      "loss": 2.0623,
+      "step": 14092
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5373061797582684e-06,
+      "loss": 2.0583,
+      "step": 14093
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.535420465605574e-06,
+      "loss": 2.0579,
+      "step": 14094
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.533534966870699e-06,
+      "loss": 2.0239,
+      "step": 14095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.531649683603086e-06,
+      "loss": 2.0066,
+      "step": 14096
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.529764615852176e-06,
+      "loss": 2.073,
+      "step": 14097
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5278797636674085e-06,
+      "loss": 2.0449,
+      "step": 14098
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.525995127098195e-06,
+      "loss": 2.0542,
+      "step": 14099
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 7.5241107061939775e-06,
+      "loss": 2.0305,
+      "step": 14100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.522226501004156e-06,
+      "loss": 2.0679,
+      "step": 14101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.520342511578158e-06,
+      "loss": 2.0701,
+      "step": 14102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.518458737965374e-06,
+      "loss": 2.1006,
+      "step": 14103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.516575180215212e-06,
+      "loss": 2.1011,
+      "step": 14104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.514691838377062e-06,
+      "loss": 1.9856,
+      "step": 14105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.512808712500317e-06,
+      "loss": 1.9825,
+      "step": 14106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.510925802634357e-06,
+      "loss": 2.1077,
+      "step": 14107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.50904310882856e-06,
+      "loss": 2.1042,
+      "step": 14108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.507160631132296e-06,
+      "loss": 2.0798,
+      "step": 14109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.505278369594931e-06,
+      "loss": 2.0544,
+      "step": 14110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.503396324265828e-06,
+      "loss": 2.123,
+      "step": 14111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.501514495194338e-06,
+      "loss": 2.1143,
+      "step": 14112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4996328824298165e-06,
+      "loss": 2.0303,
+      "step": 14113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4977514860215926e-06,
+      "loss": 2.0376,
+      "step": 14114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.495870306019021e-06,
+      "loss": 2.0308,
+      "step": 14115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.493989342471415e-06,
+      "loss": 2.0459,
+      "step": 14116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.492108595428121e-06,
+      "loss": 2.0867,
+      "step": 14117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.490228064938442e-06,
+      "loss": 2.0808,
+      "step": 14118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.488347751051703e-06,
+      "loss": 2.093,
+      "step": 14119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.486467653817208e-06,
+      "loss": 2.1121,
+      "step": 14120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4845877732842625e-06,
+      "loss": 2.0962,
+      "step": 14121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4827081095021635e-06,
+      "loss": 2.1206,
+      "step": 14122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.480828662520203e-06,
+      "loss": 2.0571,
+      "step": 14123
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.478949432387669e-06,
+      "loss": 2.0642,
+      "step": 14124
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.477070419153839e-06,
+      "loss": 2.0908,
+      "step": 14125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.47519162286799e-06,
+      "loss": 2.0977,
+      "step": 14126
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.473313043579392e-06,
+      "loss": 2.0381,
+      "step": 14127
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.471434681337308e-06,
+      "loss": 2.0525,
+      "step": 14128
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.469556536190994e-06,
+      "loss": 2.0522,
+      "step": 14129
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.467678608189708e-06,
+      "loss": 2.032,
+      "step": 14130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.465800897382683e-06,
+      "loss": 2.0356,
+      "step": 14131
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.463923403819181e-06,
+      "loss": 2.05,
+      "step": 14132
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.462046127548412e-06,
+      "loss": 2.0537,
+      "step": 14133
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4601690686196285e-06,
+      "loss": 2.0752,
+      "step": 14134
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.458292227082041e-06,
+      "loss": 2.0613,
+      "step": 14135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.456415602984869e-06,
+      "loss": 2.0984,
+      "step": 14136
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.454539196377326e-06,
+      "loss": 2.0166,
+      "step": 14137
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.45266300730862e-06,
+      "loss": 2.0371,
+      "step": 14138
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.45078703582795e-06,
+      "loss": 2.1238,
+      "step": 14139
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.448911281984512e-06,
+      "loss": 2.1055,
+      "step": 14140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.447035745827496e-06,
+      "loss": 2.1013,
+      "step": 14141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4451604274060845e-06,
+      "loss": 2.1169,
+      "step": 14142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.443285326769458e-06,
+      "loss": 1.9983,
+      "step": 14143
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.441410443966784e-06,
+      "loss": 1.9993,
+      "step": 14144
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.43953577904724e-06,
+      "loss": 2.0627,
+      "step": 14145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.437661332059966e-06,
+      "loss": 2.0771,
+      "step": 14146
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.435787103054144e-06,
+      "loss": 2.1055,
+      "step": 14147
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4339130920788975e-06,
+      "loss": 2.0686,
+      "step": 14148
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.432039299183393e-06,
+      "loss": 2.095,
+      "step": 14149
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.430165724416752e-06,
+      "loss": 2.0645,
+      "step": 14150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.428292367828116e-06,
+      "loss": 2.0806,
+      "step": 14151
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.426419229466607e-06,
+      "loss": 2.0967,
+      "step": 14152
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.424546309381346e-06,
+      "loss": 2.0908,
+      "step": 14153
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.422673607621451e-06,
+      "loss": 2.1184,
+      "step": 14154
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4208011242360355e-06,
+      "loss": 2.1001,
+      "step": 14155
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.418928859274185e-06,
+      "loss": 2.0994,
+      "step": 14156
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.417056812785023e-06,
+      "loss": 2.0977,
+      "step": 14157
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.415184984817618e-06,
+      "loss": 2.0635,
+      "step": 14158
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.413313375421078e-06,
+      "loss": 2.0938,
+      "step": 14159
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.411441984644467e-06,
+      "loss": 2.1348,
+      "step": 14160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.409570812536863e-06,
+      "loss": 2.1199,
+      "step": 14161
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.407699859147347e-06,
+      "loss": 2.1042,
+      "step": 14162
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.405829124524963e-06,
+      "loss": 2.0808,
+      "step": 14163
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.403958608718793e-06,
+      "loss": 2.0464,
+      "step": 14164
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.40208831177787e-06,
+      "loss": 2.083,
+      "step": 14165
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.4002182337512465e-06,
+      "loss": 2.1008,
+      "step": 14166
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.398348374687962e-06,
+      "loss": 2.0981,
+      "step": 14167
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.396478734637053e-06,
+      "loss": 2.0876,
+      "step": 14168
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.394609313647549e-06,
+      "loss": 2.0381,
+      "step": 14169
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3927401117684765e-06,
+      "loss": 2.0571,
+      "step": 14170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.39087112904884e-06,
+      "loss": 2.0881,
+      "step": 14171
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.38900236553767e-06,
+      "loss": 2.0715,
+      "step": 14172
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.387133821283953e-06,
+      "loss": 2.0542,
+      "step": 14173
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.385265496336711e-06,
+      "loss": 2.0298,
+      "step": 14174
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.383397390744922e-06,
+      "loss": 2.0693,
+      "step": 14175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3815295045575805e-06,
+      "loss": 2.0288,
+      "step": 14176
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3796618378236694e-06,
+      "loss": 2.0149,
+      "step": 14177
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.377794390592164e-06,
+      "loss": 1.9937,
+      "step": 14178
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.375927162912039e-06,
+      "loss": 2.0146,
+      "step": 14179
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.374060154832261e-06,
+      "loss": 2.0188,
+      "step": 14180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.372193366401787e-06,
+      "loss": 2.0664,
+      "step": 14181
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.370326797669573e-06,
+      "loss": 2.0688,
+      "step": 14182
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.368460448684566e-06,
+      "loss": 2.0542,
+      "step": 14183
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.366594319495712e-06,
+      "loss": 2.0803,
+      "step": 14184
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.364728410151949e-06,
+      "loss": 2.0295,
+      "step": 14185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.362862720702198e-06,
+      "loss": 2.0693,
+      "step": 14186
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.360997251195402e-06,
+      "loss": 2.0962,
+      "step": 14187
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.359132001680459e-06,
+      "loss": 2.0969,
+      "step": 14188
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.357266972206307e-06,
+      "loss": 2.0659,
+      "step": 14189
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.355402162821835e-06,
+      "loss": 2.0771,
+      "step": 14190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.353537573575953e-06,
+      "loss": 2.074,
+      "step": 14191
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.35167320451756e-06,
+      "loss": 2.084,
+      "step": 14192
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.349809055695542e-06,
+      "loss": 2.1028,
+      "step": 14193
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.347945127158786e-06,
+      "loss": 2.1243,
+      "step": 14194
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.346081418956172e-06,
+      "loss": 2.0789,
+      "step": 14195
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.344217931136575e-06,
+      "loss": 2.0715,
+      "step": 14196
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.342354663748859e-06,
+      "loss": 2.0493,
+      "step": 14197
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3404916168418884e-06,
+      "loss": 2.063,
+      "step": 14198
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.33862879046452e-06,
+      "loss": 2.0581,
+      "step": 14199
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.336766184665608e-06,
+      "loss": 2.0483,
+      "step": 14200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.334903799493982e-06,
+      "loss": 2.0977,
+      "step": 14201
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.333041634998503e-06,
+      "loss": 2.0693,
+      "step": 14202
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.331179691227979e-06,
+      "loss": 2.0874,
+      "step": 14203
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.329317968231264e-06,
+      "loss": 2.0742,
+      "step": 14204
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.32745646605716e-06,
+      "loss": 2.0259,
+      "step": 14205
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.325595184754491e-06,
+      "loss": 2.0491,
+      "step": 14206
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.323734124372062e-06,
+      "loss": 2.1133,
+      "step": 14207
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3218732849586814e-06,
+      "loss": 2.0886,
+      "step": 14208
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.320012666563147e-06,
+      "loss": 2.0288,
+      "step": 14209
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.31815226923425e-06,
+      "loss": 2.0291,
+      "step": 14210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.316292093020779e-06,
+      "loss": 2.1467,
+      "step": 14211
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.314432137971514e-06,
+      "loss": 2.127,
+      "step": 14212
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.312572404135227e-06,
+      "loss": 2.0908,
+      "step": 14213
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.310712891560693e-06,
+      "loss": 2.0835,
+      "step": 14214
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.308853600296676e-06,
+      "loss": 2.0627,
+      "step": 14215
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.306994530391923e-06,
+      "loss": 2.0566,
+      "step": 14216
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.305135681895203e-06,
+      "loss": 2.0769,
+      "step": 14217
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3032770548552416e-06,
+      "loss": 2.0825,
+      "step": 14218
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.3014186493208005e-06,
+      "loss": 2.0737,
+      "step": 14219
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2995604653406e-06,
+      "loss": 2.0583,
+      "step": 14220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.297702502963371e-06,
+      "loss": 2.0608,
+      "step": 14221
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.295844762237839e-06,
+      "loss": 2.073,
+      "step": 14222
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.293987243212722e-06,
+      "loss": 2.0613,
+      "step": 14223
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2921299459367275e-06,
+      "loss": 2.0544,
+      "step": 14224
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.290272870458569e-06,
+      "loss": 2.0708,
+      "step": 14225
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2884160168269295e-06,
+      "loss": 2.0508,
+      "step": 14226
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.286559385090525e-06,
+      "loss": 2.0842,
+      "step": 14227
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.284702975298021e-06,
+      "loss": 2.0859,
+      "step": 14228
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.282846787498115e-06,
+      "loss": 2.0479,
+      "step": 14229
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.280990821739484e-06,
+      "loss": 2.0461,
+      "step": 14230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.279135078070785e-06,
+      "loss": 2.0928,
+      "step": 14231
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.277279556540703e-06,
+      "loss": 2.0981,
+      "step": 14232
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.275424257197873e-06,
+      "loss": 2.0364,
+      "step": 14233
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.273569180090971e-06,
+      "loss": 2.0359,
+      "step": 14234
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.271714325268628e-06,
+      "loss": 2.0461,
+      "step": 14235
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.26985969277949e-06,
+      "loss": 2.0271,
+      "step": 14236
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.268005282672195e-06,
+      "loss": 2.0581,
+      "step": 14237
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.266151094995373e-06,
+      "loss": 2.0564,
+      "step": 14238
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.264297129797646e-06,
+      "loss": 2.0337,
+      "step": 14239
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.262443387127634e-06,
+      "loss": 2.0625,
+      "step": 14240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.260589867033941e-06,
+      "loss": 2.1172,
+      "step": 14241
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.258736569565189e-06,
+      "loss": 2.1101,
+      "step": 14242
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.25688349476996e-06,
+      "loss": 2.0535,
+      "step": 14243
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2550306426968675e-06,
+      "loss": 2.075,
+      "step": 14244
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.253178013394481e-06,
+      "loss": 2.085,
+      "step": 14245
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.251325606911406e-06,
+      "loss": 2.0942,
+      "step": 14246
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2494734232962e-06,
+      "loss": 2.1045,
+      "step": 14247
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.247621462597443e-06,
+      "loss": 2.0784,
+      "step": 14248
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.245769724863699e-06,
+      "loss": 2.0327,
+      "step": 14249
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.243918210143529e-06,
+      "loss": 2.0391,
+      "step": 14250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.242066918485485e-06,
+      "loss": 2.0513,
+      "step": 14251
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.240215849938115e-06,
+      "loss": 2.0762,
+      "step": 14252
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.238365004549964e-06,
+      "loss": 2.0801,
+      "step": 14253
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.236514382369566e-06,
+      "loss": 2.0859,
+      "step": 14254
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.234663983445452e-06,
+      "loss": 2.0864,
+      "step": 14255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.232813807826145e-06,
+      "loss": 2.1079,
+      "step": 14256
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.230963855560171e-06,
+      "loss": 2.0417,
+      "step": 14257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.229114126696025e-06,
+      "loss": 2.0474,
+      "step": 14258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.227264621282237e-06,
+      "loss": 2.0205,
+      "step": 14259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.225415339367288e-06,
+      "loss": 2.0454,
+      "step": 14260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.223566280999691e-06,
+      "loss": 2.0566,
+      "step": 14261
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.221717446227923e-06,
+      "loss": 2.0674,
+      "step": 14262
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.219868835100469e-06,
+      "loss": 2.0852,
+      "step": 14263
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.21802044766581e-06,
+      "loss": 2.0913,
+      "step": 14264
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.216172283972417e-06,
+      "loss": 2.051,
+      "step": 14265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.214324344068756e-06,
+      "loss": 2.0383,
+      "step": 14266
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.212476628003286e-06,
+      "loss": 2.032,
+      "step": 14267
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2106291358244624e-06,
+      "loss": 1.9998,
+      "step": 14268
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.208781867580732e-06,
+      "loss": 2.0674,
+      "step": 14269
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.206934823320539e-06,
+      "loss": 2.052,
+      "step": 14270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.2050880030923194e-06,
+      "loss": 2.0769,
+      "step": 14271
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.203241406944509e-06,
+      "loss": 2.0903,
+      "step": 14272
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.201395034925516e-06,
+      "loss": 2.0688,
+      "step": 14273
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.1995488870837845e-06,
+      "loss": 2.0972,
+      "step": 14274
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.197702963467701e-06,
+      "loss": 2.0505,
+      "step": 14275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.1958572641256954e-06,
+      "loss": 2.0361,
+      "step": 14276
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.194011789106155e-06,
+      "loss": 2.1057,
+      "step": 14277
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.192166538457481e-06,
+      "loss": 2.1047,
+      "step": 14278
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.190321512228058e-06,
+      "loss": 2.0588,
+      "step": 14279
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.188476710466277e-06,
+      "loss": 2.1052,
+      "step": 14280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.18663213322051e-06,
+      "loss": 2.0466,
+      "step": 14281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.18478778053913e-06,
+      "loss": 2.0547,
+      "step": 14282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.182943652470506e-06,
+      "loss": 2.0737,
+      "step": 14283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.181099749062993e-06,
+      "loss": 2.0679,
+      "step": 14284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.179256070364951e-06,
+      "loss": 2.0667,
+      "step": 14285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.177412616424725e-06,
+      "loss": 2.0337,
+      "step": 14286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.175569387290662e-06,
+      "loss": 2.1084,
+      "step": 14287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.173726383011085e-06,
+      "loss": 2.094,
+      "step": 14288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.171883603634344e-06,
+      "loss": 2.0891,
+      "step": 14289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.170041049208744e-06,
+      "loss": 2.0928,
+      "step": 14290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.168198719782624e-06,
+      "loss": 2.072,
+      "step": 14291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.166356615404281e-06,
+      "loss": 2.082,
+      "step": 14292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.1645147361220285e-06,
+      "loss": 2.0801,
+      "step": 14293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.162673081984166e-06,
+      "loss": 2.0647,
+      "step": 14294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.160831653038991e-06,
+      "loss": 2.0613,
+      "step": 14295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.158990449334791e-06,
+      "loss": 2.0593,
+      "step": 14296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.157149470919855e-06,
+      "loss": 2.0276,
+      "step": 14297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.155308717842445e-06,
+      "loss": 2.0417,
+      "step": 14298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.153468190150847e-06,
+      "loss": 2.0845,
+      "step": 14299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.151627887893325e-06,
+      "loss": 2.0737,
+      "step": 14300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.149787811118135e-06,
+      "loss": 2.0557,
+      "step": 14301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.147947959873537e-06,
+      "loss": 2.0852,
+      "step": 14302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.146108334207765e-06,
+      "loss": 2.0713,
+      "step": 14303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.144268934169082e-06,
+      "loss": 2.0627,
+      "step": 14304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.1424297598057e-06,
+      "loss": 2.1001,
+      "step": 14305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.140590811165875e-06,
+      "loss": 2.1045,
+      "step": 14306
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.13875208829781e-06,
+      "loss": 2.0813,
+      "step": 14307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.136913591249732e-06,
+      "loss": 2.0671,
+      "step": 14308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.135075320069855e-06,
+      "loss": 2.0842,
+      "step": 14309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.133237274806381e-06,
+      "loss": 2.0898,
+      "step": 14310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.1313994555075145e-06,
+      "loss": 2.1035,
+      "step": 14311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.129561862221453e-06,
+      "loss": 2.0903,
+      "step": 14312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.1277244949963704e-06,
+      "loss": 2.0637,
+      "step": 14313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.125887353880472e-06,
+      "loss": 2.0625,
+      "step": 14314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.124050438921911e-06,
+      "loss": 2.0962,
+      "step": 14315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.122213750168882e-06,
+      "loss": 2.0647,
+      "step": 14316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.120377287669532e-06,
+      "loss": 2.0247,
+      "step": 14317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.118541051472026e-06,
+      "loss": 2.0051,
+      "step": 14318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.116705041624518e-06,
+      "loss": 2.0837,
+      "step": 14319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.114869258175155e-06,
+      "loss": 2.0308,
+      "step": 14320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.113033701172076e-06,
+      "loss": 2.0605,
+      "step": 14321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.11119837066342e-06,
+      "loss": 2.0354,
+      "step": 14322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.109363266697314e-06,
+      "loss": 2.0469,
+      "step": 14323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.107528389321883e-06,
+      "loss": 2.0522,
+      "step": 14324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.105693738585243e-06,
+      "loss": 2.0686,
+      "step": 14325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.103859314535507e-06,
+      "loss": 2.0625,
+      "step": 14326
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.102025117220785e-06,
+      "loss": 2.103,
+      "step": 14327
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.100191146689161e-06,
+      "loss": 2.1182,
+      "step": 14328
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.098357402988749e-06,
+      "loss": 2.0945,
+      "step": 14329
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.096523886167619e-06,
+      "loss": 2.0959,
+      "step": 14330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.09469059627387e-06,
+      "loss": 2.0752,
+      "step": 14331
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0928575333555636e-06,
+      "loss": 2.1121,
+      "step": 14332
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.091024697460775e-06,
+      "loss": 2.0396,
+      "step": 14333
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.08919208863757e-06,
+      "loss": 2.0396,
+      "step": 14334
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.087359706934005e-06,
+      "loss": 2.0864,
+      "step": 14335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0855275523981285e-06,
+      "loss": 2.1128,
+      "step": 14336
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.083695625077992e-06,
+      "loss": 2.0437,
+      "step": 14337
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.081863925021636e-06,
+      "loss": 2.0371,
+      "step": 14338
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.08003245227709e-06,
+      "loss": 2.1094,
+      "step": 14339
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.078201206892385e-06,
+      "loss": 2.092,
+      "step": 14340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.076370188915546e-06,
+      "loss": 2.1094,
+      "step": 14341
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0745393983945875e-06,
+      "loss": 2.1104,
+      "step": 14342
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.072708835377512e-06,
+      "loss": 2.0801,
+      "step": 14343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.070878499912338e-06,
+      "loss": 2.0632,
+      "step": 14344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.069048392047049e-06,
+      "loss": 2.0339,
+      "step": 14345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.067218511829652e-06,
+      "loss": 2.0662,
+      "step": 14346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.065388859308123e-06,
+      "loss": 2.0278,
+      "step": 14347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.063559434530447e-06,
+      "loss": 2.0408,
+      "step": 14348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.061730237544598e-06,
+      "loss": 2.0735,
+      "step": 14349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.059901268398541e-06,
+      "loss": 2.0586,
+      "step": 14350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.058072527140244e-06,
+      "loss": 2.0803,
+      "step": 14351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.056244013817661e-06,
+      "loss": 2.1091,
+      "step": 14352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.054415728478743e-06,
+      "loss": 2.1172,
+      "step": 14353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.052587671171435e-06,
+      "loss": 2.1113,
+      "step": 14354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.050759841943675e-06,
+      "loss": 2.1121,
+      "step": 14355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0489322408433945e-06,
+      "loss": 2.0759,
+      "step": 14356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.047104867918529e-06,
+      "loss": 2.0505,
+      "step": 14357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.045277723216981e-06,
+      "loss": 2.0366,
+      "step": 14358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.043450806786686e-06,
+      "loss": 2.0117,
+      "step": 14359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.041624118675535e-06,
+      "loss": 2.0122,
+      "step": 14360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.039797658931447e-06,
+      "loss": 2.0637,
+      "step": 14361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.037971427602301e-06,
+      "loss": 2.0303,
+      "step": 14362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.036145424736006e-06,
+      "loss": 2.0486,
+      "step": 14363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.034319650380436e-06,
+      "loss": 2.0515,
+      "step": 14364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.03249410458347e-06,
+      "loss": 2.0947,
+      "step": 14365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0306687873929825e-06,
+      "loss": 2.0837,
+      "step": 14366
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.028843698856842e-06,
+      "loss": 2.0076,
+      "step": 14367
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.027018839022907e-06,
+      "loss": 2.0261,
+      "step": 14368
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.025194207939033e-06,
+      "loss": 2.0051,
+      "step": 14369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.023369805653069e-06,
+      "loss": 2.0317,
+      "step": 14370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.021545632212859e-06,
+      "loss": 2.0693,
+      "step": 14371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0197216876662375e-06,
+      "loss": 2.0859,
+      "step": 14372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.017897972061037e-06,
+      "loss": 2.0925,
+      "step": 14373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.016074485445089e-06,
+      "loss": 2.0959,
+      "step": 14374
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.014251227866194e-06,
+      "loss": 2.0764,
+      "step": 14375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.012428199372186e-06,
+      "loss": 2.0374,
+      "step": 14376
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.010605400010852e-06,
+      "loss": 2.0823,
+      "step": 14377
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.008782829830012e-06,
+      "loss": 2.0662,
+      "step": 14378
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.006960488877447e-06,
+      "loss": 2.114,
+      "step": 14379
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.005138377200951e-06,
+      "loss": 2.0962,
+      "step": 14380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.003316494848304e-06,
+      "loss": 2.0486,
+      "step": 14381
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.0014948418672855e-06,
+      "loss": 2.0571,
+      "step": 14382
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.999673418305666e-06,
+      "loss": 2.1023,
+      "step": 14383
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.997852224211215e-06,
+      "loss": 2.0764,
+      "step": 14384
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.996031259631675e-06,
+      "loss": 2.0349,
+      "step": 14385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.99421052461482e-06,
+      "loss": 2.0867,
+      "step": 14386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.992390019208377e-06,
+      "loss": 2.0498,
+      "step": 14387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.990569743460107e-06,
+      "loss": 2.0938,
+      "step": 14388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.988749697417726e-06,
+      "loss": 2.0676,
+      "step": 14389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.986929881128972e-06,
+      "loss": 2.0833,
+      "step": 14390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.985110294641564e-06,
+      "loss": 2.0938,
+      "step": 14391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.983290938003221e-06,
+      "loss": 2.0801,
+      "step": 14392
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.981471811261652e-06,
+      "loss": 2.0576,
+      "step": 14393
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.979652914464565e-06,
+      "loss": 2.0552,
+      "step": 14394
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.977834247659653e-06,
+      "loss": 2.0925,
+      "step": 14395
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.976015810894612e-06,
+      "loss": 2.1199,
+      "step": 14396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.974197604217127e-06,
+      "loss": 2.0542,
+      "step": 14397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.972379627674879e-06,
+      "loss": 2.0618,
+      "step": 14398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.970561881315547e-06,
+      "loss": 2.0461,
+      "step": 14399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.968744365186786e-06,
+      "loss": 2.0422,
+      "step": 14400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.966927079336277e-06,
+      "loss": 2.0339,
+      "step": 14401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.965110023811656e-06,
+      "loss": 2.0195,
+      "step": 14402
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.963293198660592e-06,
+      "loss": 2.0881,
+      "step": 14403
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.961476603930712e-06,
+      "loss": 2.1121,
+      "step": 14404
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9596602396696675e-06,
+      "loss": 2.0684,
+      "step": 14405
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.957844105925084e-06,
+      "loss": 2.0833,
+      "step": 14406
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.956028202744588e-06,
+      "loss": 2.0701,
+      "step": 14407
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.954212530175801e-06,
+      "loss": 2.0491,
+      "step": 14408
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.952397088266335e-06,
+      "loss": 2.1135,
+      "step": 14409
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.950581877063801e-06,
+      "loss": 2.1016,
+      "step": 14410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.948766896615799e-06,
+      "loss": 2.0828,
+      "step": 14411
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.946952146969924e-06,
+      "loss": 2.0723,
+      "step": 14412
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9451376281737675e-06,
+      "loss": 2.0322,
+      "step": 14413
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.943323340274914e-06,
+      "loss": 2.0225,
+      "step": 14414
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.941509283320932e-06,
+      "loss": 2.0923,
+      "step": 14415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.939695457359409e-06,
+      "loss": 2.0969,
+      "step": 14416
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.937881862437891e-06,
+      "loss": 2.0393,
+      "step": 14417
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.936068498603959e-06,
+      "loss": 2.0388,
+      "step": 14418
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.934255365905151e-06,
+      "loss": 2.0708,
+      "step": 14419
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.932442464389019e-06,
+      "loss": 2.0908,
+      "step": 14420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.930629794103103e-06,
+      "loss": 2.1094,
+      "step": 14421
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.92881735509494e-06,
+      "loss": 2.0908,
+      "step": 14422
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.927005147412058e-06,
+      "loss": 2.0791,
+      "step": 14423
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9251931711019795e-06,
+      "loss": 2.0601,
+      "step": 14424
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.923381426212222e-06,
+      "loss": 2.0491,
+      "step": 14425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9215699127903e-06,
+      "loss": 2.0374,
+      "step": 14426
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9197586308837115e-06,
+      "loss": 2.0574,
+      "step": 14427
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.91794758053996e-06,
+      "loss": 2.0662,
+      "step": 14428
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.916136761806541e-06,
+      "loss": 2.085,
+      "step": 14429
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.91432617473093e-06,
+      "loss": 2.0793,
+      "step": 14430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.912515819360621e-06,
+      "loss": 2.0627,
+      "step": 14431
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9107056957430755e-06,
+      "loss": 2.0818,
+      "step": 14432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.908895803925777e-06,
+      "loss": 2.1047,
+      "step": 14433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.907086143956174e-06,
+      "loss": 2.1125,
+      "step": 14434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.9052767158817275e-06,
+      "loss": 2.0779,
+      "step": 14435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.903467519749888e-06,
+      "loss": 2.0652,
+      "step": 14436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.901658555608103e-06,
+      "loss": 2.0503,
+      "step": 14437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.899849823503805e-06,
+      "loss": 2.0566,
+      "step": 14438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.898041323484427e-06,
+      "loss": 1.9978,
+      "step": 14439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.896233055597399e-06,
+      "loss": 2.0435,
+      "step": 14440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.894425019890136e-06,
+      "loss": 2.0066,
+      "step": 14441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.892617216410053e-06,
+      "loss": 2.0435,
+      "step": 14442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.89080964520456e-06,
+      "loss": 2.0251,
+      "step": 14443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.889002306321059e-06,
+      "loss": 2.0403,
+      "step": 14444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.887195199806933e-06,
+      "loss": 2.0481,
+      "step": 14445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.885388325709591e-06,
+      "loss": 2.0864,
+      "step": 14446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.883581684076398e-06,
+      "loss": 1.9761,
+      "step": 14447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.881775274954747e-06,
+      "loss": 2.0239,
+      "step": 14448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.879969098391998e-06,
+      "loss": 2.0889,
+      "step": 14449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8781631544355185e-06,
+      "loss": 2.0774,
+      "step": 14450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.876357443132668e-06,
+      "loss": 2.0342,
+      "step": 14451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.874551964530799e-06,
+      "loss": 2.0874,
+      "step": 14452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.872746718677259e-06,
+      "loss": 2.0669,
+      "step": 14453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.870941705619392e-06,
+      "loss": 2.0837,
+      "step": 14454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8691369254045175e-06,
+      "loss": 2.0579,
+      "step": 14455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.867332378079986e-06,
+      "loss": 2.0537,
+      "step": 14456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.865528063693096e-06,
+      "loss": 2.0632,
+      "step": 14457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.863723982291186e-06,
+      "loss": 2.053,
+      "step": 14458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.861920133921551e-06,
+      "loss": 2.0918,
+      "step": 14459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.860116518631499e-06,
+      "loss": 2.0771,
+      "step": 14460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.858313136468329e-06,
+      "loss": 2.0901,
+      "step": 14461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.856509987479329e-06,
+      "loss": 2.0967,
+      "step": 14462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.85470707171179e-06,
+      "loss": 2.0591,
+      "step": 14463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.852904389212987e-06,
+      "loss": 2.0481,
+      "step": 14464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.851101940030195e-06,
+      "loss": 2.0278,
+      "step": 14465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.849299724210681e-06,
+      "loss": 2.033,
+      "step": 14466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.847497741801707e-06,
+      "loss": 2.0286,
+      "step": 14467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.845695992850525e-06,
+      "loss": 2.0366,
+      "step": 14468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.843894477404391e-06,
+      "loss": 2.0981,
+      "step": 14469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.842093195510533e-06,
+      "loss": 2.0923,
+      "step": 14470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.840292147216207e-06,
+      "loss": 2.0662,
+      "step": 14471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.838491332568622e-06,
+      "loss": 2.0618,
+      "step": 14472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.836690751615023e-06,
+      "loss": 2.1155,
+      "step": 14473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.834890404402616e-06,
+      "loss": 2.1001,
+      "step": 14474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8330902909786144e-06,
+      "loss": 2.0635,
+      "step": 14475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.831290411390225e-06,
+      "loss": 2.0847,
+      "step": 14476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.829490765684646e-06,
+      "loss": 2.0833,
+      "step": 14477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.827691353909076e-06,
+      "loss": 2.0457,
+      "step": 14478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.825892176110698e-06,
+      "loss": 2.073,
+      "step": 14479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8240932323366935e-06,
+      "loss": 2.1001,
+      "step": 14480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.822294522634241e-06,
+      "loss": 2.062,
+      "step": 14481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.820496047050506e-06,
+      "loss": 2.051,
+      "step": 14482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.818697805632654e-06,
+      "loss": 2.0278,
+      "step": 14483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.816899798427841e-06,
+      "loss": 2.0425,
+      "step": 14484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.815102025483216e-06,
+      "loss": 2.0801,
+      "step": 14485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.813304486845931e-06,
+      "loss": 2.0928,
+      "step": 14486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.811507182563107e-06,
+      "loss": 2.0488,
+      "step": 14487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8097101126818976e-06,
+      "loss": 2.0747,
+      "step": 14488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.807913277249411e-06,
+      "loss": 2.0447,
+      "step": 14489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.806116676312783e-06,
+      "loss": 2.0535,
+      "step": 14490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8043203099191156e-06,
+      "loss": 2.0994,
+      "step": 14491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.802524178115518e-06,
+      "loss": 2.1282,
+      "step": 14492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.8007282809490945e-06,
+      "loss": 2.0242,
+      "step": 14493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.798932618466939e-06,
+      "loss": 2.022,
+      "step": 14494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.797137190716142e-06,
+      "loss": 2.0679,
+      "step": 14495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.795341997743788e-06,
+      "loss": 2.053,
+      "step": 14496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.793547039596949e-06,
+      "loss": 2.0898,
+      "step": 14497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.7917523163226985e-06,
+      "loss": 2.0693,
+      "step": 14498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.789957827968102e-06,
+      "loss": 2.0957,
+      "step": 14499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.788163574580217e-06,
+      "loss": 2.1025,
+      "step": 14500
+    },
+    {
+      "Bleu_1": 0.7736383354355431,
+      "Bleu_2": 0.6133225229545359,
+      "Bleu_3": 0.47126553835975493,
+      "Bleu_4": 0.3574641681467299,
+      "CIDEr": 1.203051978913014,
+      "METEOR": 0.2785106190034526,
+      "ROUGE_L": 0.5601638863414345,
+      "epoch": 0.72,
+      "step": 14500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.786369556206099e-06,
+      "loss": 2.0447,
+      "step": 14501
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.784575772892783e-06,
+      "loss": 2.0566,
+      "step": 14502
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.782782224687323e-06,
+      "loss": 2.0549,
+      "step": 14503
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7809889116367385e-06,
+      "loss": 2.0381,
+      "step": 14504
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.779195833788074e-06,
+      "loss": 2.071,
+      "step": 14505
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.777402991188336e-06,
+      "loss": 2.0852,
+      "step": 14506
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.775610383884546e-06,
+      "loss": 2.1445,
+      "step": 14507
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.773818011923713e-06,
+      "loss": 2.127,
+      "step": 14508
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.772025875352837e-06,
+      "loss": 2.1096,
+      "step": 14509
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.77023397421892e-06,
+      "loss": 2.0991,
+      "step": 14510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.768442308568949e-06,
+      "loss": 2.1035,
+      "step": 14511
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.766650878449908e-06,
+      "loss": 2.0959,
+      "step": 14512
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.764859683908775e-06,
+      "loss": 2.0713,
+      "step": 14513
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7630687249925246e-06,
+      "loss": 2.1003,
+      "step": 14514
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.76127800174812e-06,
+      "loss": 2.1057,
+      "step": 14515
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.759487514222528e-06,
+      "loss": 2.0996,
+      "step": 14516
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.757697262462685e-06,
+      "loss": 2.0879,
+      "step": 14517
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7559072465155595e-06,
+      "loss": 2.0925,
+      "step": 14518
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.754117466428072e-06,
+      "loss": 2.0671,
+      "step": 14519
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.752327922247177e-06,
+      "loss": 2.0874,
+      "step": 14520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7505386140197914e-06,
+      "loss": 2.106,
+      "step": 14521
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7487495417928376e-06,
+      "loss": 2.1201,
+      "step": 14522
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.746960705613236e-06,
+      "loss": 2.0903,
+      "step": 14523
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.745172105527894e-06,
+      "loss": 2.0911,
+      "step": 14524
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.743383741583718e-06,
+      "loss": 2.0891,
+      "step": 14525
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.741595613827609e-06,
+      "loss": 2.0784,
+      "step": 14526
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.739807722306446e-06,
+      "loss": 2.0146,
+      "step": 14527
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7380200670671305e-06,
+      "loss": 2.0459,
+      "step": 14528
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.736232648156526e-06,
+      "loss": 2.0261,
+      "step": 14529
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.73444546562152e-06,
+      "loss": 2.0642,
+      "step": 14530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.732658519508971e-06,
+      "loss": 2.074,
+      "step": 14531
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7308718098657404e-06,
+      "loss": 2.0598,
+      "step": 14532
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.72908533673868e-06,
+      "loss": 2.0596,
+      "step": 14533
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.727299100174642e-06,
+      "loss": 2.0811,
+      "step": 14534
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.725513100220474e-06,
+      "loss": 2.0952,
+      "step": 14535
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7237273369230005e-06,
+      "loss": 2.0991,
+      "step": 14536
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.721941810329059e-06,
+      "loss": 2.0215,
+      "step": 14537
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.720156520485467e-06,
+      "loss": 2.0312,
+      "step": 14538
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.718371467439045e-06,
+      "loss": 2.0591,
+      "step": 14539
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.716586651236604e-06,
+      "loss": 2.0308,
+      "step": 14540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.714802071924953e-06,
+      "loss": 2.0127,
+      "step": 14541
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.713017729550877e-06,
+      "loss": 2.0469,
+      "step": 14542
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.711233624161187e-06,
+      "loss": 2.0981,
+      "step": 14543
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.709449755802648e-06,
+      "loss": 2.0935,
+      "step": 14544
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.707666124522061e-06,
+      "loss": 2.0896,
+      "step": 14545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.705882730366185e-06,
+      "loss": 2.0667,
+      "step": 14546
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.704099573381792e-06,
+      "loss": 2.0876,
+      "step": 14547
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.702316653615647e-06,
+      "loss": 2.0571,
+      "step": 14548
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.700533971114496e-06,
+      "loss": 2.0632,
+      "step": 14549
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.698751525925097e-06,
+      "loss": 2.0693,
+      "step": 14550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.696969318094188e-06,
+      "loss": 2.0813,
+      "step": 14551
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6951873476685045e-06,
+      "loss": 2.0874,
+      "step": 14552
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.693405614694779e-06,
+      "loss": 2.04,
+      "step": 14553
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6916241192197334e-06,
+      "loss": 2.0347,
+      "step": 14554
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.689842861290089e-06,
+      "loss": 2.0427,
+      "step": 14555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.688061840952557e-06,
+      "loss": 2.0276,
+      "step": 14556
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.686281058253833e-06,
+      "loss": 2.0728,
+      "step": 14557
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.684500513240632e-06,
+      "loss": 2.0881,
+      "step": 14558
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.682720205959629e-06,
+      "loss": 2.1006,
+      "step": 14559
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.680940136457527e-06,
+      "loss": 2.1162,
+      "step": 14560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.679160304780997e-06,
+      "loss": 2.0488,
+      "step": 14561
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6773807109767115e-06,
+      "loss": 2.0151,
+      "step": 14562
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.675601355091344e-06,
+      "loss": 2.083,
+      "step": 14563
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.673822237171554e-06,
+      "loss": 2.104,
+      "step": 14564
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.672043357263996e-06,
+      "loss": 2.0664,
+      "step": 14565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6702647154153175e-06,
+      "loss": 2.0823,
+      "step": 14566
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.668486311672164e-06,
+      "loss": 2.0789,
+      "step": 14567
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.666708146081173e-06,
+      "loss": 2.1223,
+      "step": 14568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.664930218688973e-06,
+      "loss": 2.0588,
+      "step": 14569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6631525295421876e-06,
+      "loss": 2.0957,
+      "step": 14570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.66137507868744e-06,
+      "loss": 2.0314,
+      "step": 14571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.659597866171326e-06,
+      "loss": 2.0269,
+      "step": 14572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.657820892040475e-06,
+      "loss": 2.0242,
+      "step": 14573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.656044156341462e-06,
+      "loss": 2.0354,
+      "step": 14574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.654267659120901e-06,
+      "loss": 2.0222,
+      "step": 14575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6524914004253626e-06,
+      "loss": 2.0049,
+      "step": 14576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.650715380301433e-06,
+      "loss": 2.031,
+      "step": 14577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.648939598795687e-06,
+      "loss": 2.0144,
+      "step": 14578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.647164055954692e-06,
+      "loss": 2.0605,
+      "step": 14579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.645388751825009e-06,
+      "loss": 2.054,
+      "step": 14580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.643613686453194e-06,
+      "loss": 2.0615,
+      "step": 14581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.641838859885796e-06,
+      "loss": 2.0713,
+      "step": 14582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.640064272169355e-06,
+      "loss": 2.0332,
+      "step": 14583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.638289923350412e-06,
+      "loss": 2.0439,
+      "step": 14584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.636515813475496e-06,
+      "loss": 2.0752,
+      "step": 14585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.634741942591132e-06,
+      "loss": 2.0767,
+      "step": 14586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.63296831074383e-06,
+      "loss": 2.1116,
+      "step": 14587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.631194917980114e-06,
+      "loss": 2.1248,
+      "step": 14588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.629421764346474e-06,
+      "loss": 2.0615,
+      "step": 14589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.627648849889426e-06,
+      "loss": 2.0439,
+      "step": 14590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.625876174655448e-06,
+      "loss": 2.0205,
+      "step": 14591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6241037386910345e-06,
+      "loss": 2.0474,
+      "step": 14592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.622331542042664e-06,
+      "loss": 2.0769,
+      "step": 14593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.620559584756809e-06,
+      "loss": 2.0857,
+      "step": 14594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6187878668799375e-06,
+      "loss": 2.0718,
+      "step": 14595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.617016388458514e-06,
+      "loss": 2.0735,
+      "step": 14596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.6152451495389814e-06,
+      "loss": 2.1279,
+      "step": 14597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.61347415016781e-06,
+      "loss": 2.1289,
+      "step": 14598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.611703390391417e-06,
+      "loss": 2.0801,
+      "step": 14599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.609932870256261e-06,
+      "loss": 2.0813,
+      "step": 14600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.608162589808758e-06,
+      "loss": 2.0825,
+      "step": 14601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.606392549095336e-06,
+      "loss": 2.0544,
+      "step": 14602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.60462274816241e-06,
+      "loss": 2.0574,
+      "step": 14603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.602853187056388e-06,
+      "loss": 2.072,
+      "step": 14604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.60108386582369e-06,
+      "loss": 2.0786,
+      "step": 14605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.599314784510694e-06,
+      "loss": 2.0657,
+      "step": 14606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.597545943163811e-06,
+      "loss": 2.0447,
+      "step": 14607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.595777341829415e-06,
+      "loss": 2.0522,
+      "step": 14608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.594008980553887e-06,
+      "loss": 2.0264,
+      "step": 14609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.592240859383601e-06,
+      "loss": 2.0405,
+      "step": 14610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.590472978364926e-06,
+      "loss": 2.0693,
+      "step": 14611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.588705337544218e-06,
+      "loss": 2.1001,
+      "step": 14612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.586937936967843e-06,
+      "loss": 2.1003,
+      "step": 14613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.585170776682128e-06,
+      "loss": 2.1042,
+      "step": 14614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.58340385673344e-06,
+      "loss": 2.0356,
+      "step": 14615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.581637177168088e-06,
+      "loss": 2.0701,
+      "step": 14616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.579870738032427e-06,
+      "loss": 2.0168,
+      "step": 14617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.57810453937276e-06,
+      "loss": 2.019,
+      "step": 14618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.576338581235412e-06,
+      "loss": 1.9509,
+      "step": 14619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.574572863666694e-06,
+      "loss": 1.9719,
+      "step": 14620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.572807386712907e-06,
+      "loss": 2.0295,
+      "step": 14621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.57104215042035e-06,
+      "loss": 2.0312,
+      "step": 14622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.569277154835313e-06,
+      "loss": 1.9867,
+      "step": 14623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.567512400004083e-06,
+      "loss": 2.011,
+      "step": 14624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.565747885972938e-06,
+      "loss": 2.0291,
+      "step": 14625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.563983612788148e-06,
+      "loss": 2.0576,
+      "step": 14626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.562219580495981e-06,
+      "loss": 2.011,
+      "step": 14627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.560455789142702e-06,
+      "loss": 2.0422,
+      "step": 14628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.558692238774551e-06,
+      "loss": 2.0496,
+      "step": 14629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5569289294377924e-06,
+      "loss": 2.0374,
+      "step": 14630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.555165861178646e-06,
+      "loss": 2.0918,
+      "step": 14631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5534030340433685e-06,
+      "loss": 2.1179,
+      "step": 14632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.551640448078172e-06,
+      "loss": 2.0331,
+      "step": 14633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5498781033292845e-06,
+      "loss": 2.0393,
+      "step": 14634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.54811599984292e-06,
+      "loss": 2.1116,
+      "step": 14635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.546354137665288e-06,
+      "loss": 2.105,
+      "step": 14636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.544592516842592e-06,
+      "loss": 2.0081,
+      "step": 14637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.54283113742103e-06,
+      "loss": 2.0374,
+      "step": 14638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5410699994467845e-06,
+      "loss": 2.0769,
+      "step": 14639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.539309102966051e-06,
+      "loss": 2.082,
+      "step": 14640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.537548448024996e-06,
+      "loss": 1.9792,
+      "step": 14641
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.535788034669799e-06,
+      "loss": 2.0054,
+      "step": 14642
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.534027862946625e-06,
+      "loss": 2.0955,
+      "step": 14643
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.532267932901617e-06,
+      "loss": 2.1091,
+      "step": 14644
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.53050824458095e-06,
+      "loss": 2.0345,
+      "step": 14645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.52874879803075e-06,
+      "loss": 2.0615,
+      "step": 14646
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.526989593297173e-06,
+      "loss": 2.0774,
+      "step": 14647
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.525230630426339e-06,
+      "loss": 2.0828,
+      "step": 14648
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5234719094643815e-06,
+      "loss": 2.0825,
+      "step": 14649
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.52171343045742e-06,
+      "loss": 2.0781,
+      "step": 14650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.519955193451565e-06,
+      "loss": 2.1069,
+      "step": 14651
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.51819719849293e-06,
+      "loss": 2.1287,
+      "step": 14652
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.516439445627614e-06,
+      "loss": 2.0718,
+      "step": 14653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.514681934901709e-06,
+      "loss": 2.1062,
+      "step": 14654
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.512924666361308e-06,
+      "loss": 2.1121,
+      "step": 14655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.511167640052491e-06,
+      "loss": 2.0977,
+      "step": 14656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.509410856021334e-06,
+      "loss": 2.0896,
+      "step": 14657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.507654314313913e-06,
+      "loss": 2.0989,
+      "step": 14658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.505898014976274e-06,
+      "loss": 2.0723,
+      "step": 14659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5041419580544965e-06,
+      "loss": 2.042,
+      "step": 14660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.5023861435946095e-06,
+      "loss": 2.094,
+      "step": 14661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.500630571642679e-06,
+      "loss": 2.1055,
+      "step": 14662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.498875242244724e-06,
+      "loss": 2.0365,
+      "step": 14663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.4971201554467866e-06,
+      "loss": 2.0461,
+      "step": 14664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.495365311294885e-06,
+      "loss": 2.0859,
+      "step": 14665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.493610709835044e-06,
+      "loss": 2.0828,
+      "step": 14666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.491856351113274e-06,
+      "loss": 2.1099,
+      "step": 14667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.490102235175584e-06,
+      "loss": 2.1084,
+      "step": 14668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.488348362067963e-06,
+      "loss": 2.0457,
+      "step": 14669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.486594731836417e-06,
+      "loss": 2.0254,
+      "step": 14670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.484841344526921e-06,
+      "loss": 2.095,
+      "step": 14671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.483088200185466e-06,
+      "loss": 2.1125,
+      "step": 14672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.4813352988580275e-06,
+      "loss": 2.1111,
+      "step": 14673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.47958264059056e-06,
+      "loss": 2.1252,
+      "step": 14674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.47783022542904e-06,
+      "loss": 2.0967,
+      "step": 14675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.476078053419407e-06,
+      "loss": 2.1301,
+      "step": 14676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.474326124607629e-06,
+      "loss": 1.9924,
+      "step": 14677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.472574439039631e-06,
+      "loss": 2.0564,
+      "step": 14678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.4708229967613555e-06,
+      "loss": 2.0737,
+      "step": 14679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.469071797818734e-06,
+      "loss": 2.0957,
+      "step": 14680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.467320842257686e-06,
+      "loss": 2.0483,
+      "step": 14681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.46557013012413e-06,
+      "loss": 2.0242,
+      "step": 14682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.463819661463983e-06,
+      "loss": 2.0044,
+      "step": 14683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.462069436323131e-06,
+      "loss": 2.0129,
+      "step": 14684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.460319454747492e-06,
+      "loss": 2.0349,
+      "step": 14685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.458569716782938e-06,
+      "loss": 2.0615,
+      "step": 14686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.456820222475375e-06,
+      "loss": 2.0569,
+      "step": 14687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.455070971870668e-06,
+      "loss": 2.0713,
+      "step": 14688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.453321965014688e-06,
+      "loss": 2.0659,
+      "step": 14689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.451573201953307e-06,
+      "loss": 2.084,
+      "step": 14690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.449824682732378e-06,
+      "loss": 2.063,
+      "step": 14691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.44807640739776e-06,
+      "loss": 2.0713,
+      "step": 14692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.4463283759952965e-06,
+      "loss": 2.1025,
+      "step": 14693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.444580588570828e-06,
+      "loss": 2.0923,
+      "step": 14694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.442833045170189e-06,
+      "loss": 2.0183,
+      "step": 14695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.441085745839206e-06,
+      "loss": 2.0264,
+      "step": 14696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.439338690623697e-06,
+      "loss": 2.0991,
+      "step": 14697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.437591879569486e-06,
+      "loss": 2.0938,
+      "step": 14698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.435845312722367e-06,
+      "loss": 2.0596,
+      "step": 14699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.434098990128157e-06,
+      "loss": 2.0452,
+      "step": 14700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4323529118326355e-06,
+      "loss": 2.0435,
+      "step": 14701
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.430607077881606e-06,
+      "loss": 2.0256,
+      "step": 14702
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4288614883208404e-06,
+      "loss": 2.0908,
+      "step": 14703
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.427116143196121e-06,
+      "loss": 2.0989,
+      "step": 14704
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.425371042553213e-06,
+      "loss": 2.0044,
+      "step": 14705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.423626186437881e-06,
+      "loss": 2.0232,
+      "step": 14706
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4218815748958845e-06,
+      "loss": 2.0654,
+      "step": 14707
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.420137207972971e-06,
+      "loss": 2.0442,
+      "step": 14708
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.418393085714886e-06,
+      "loss": 2.124,
+      "step": 14709
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.416649208167365e-06,
+      "loss": 2.1311,
+      "step": 14710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.414905575376143e-06,
+      "loss": 2.0842,
+      "step": 14711
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.413162187386941e-06,
+      "loss": 2.0652,
+      "step": 14712
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.411419044245485e-06,
+      "loss": 2.0649,
+      "step": 14713
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.40967614599747e-06,
+      "loss": 2.0723,
+      "step": 14714
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4079334926886226e-06,
+      "loss": 2.1101,
+      "step": 14715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.406191084364622e-06,
+      "loss": 2.0786,
+      "step": 14716
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4044489210711795e-06,
+      "loss": 2.1277,
+      "step": 14717
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.402707002853967e-06,
+      "loss": 2.1113,
+      "step": 14718
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.40096532975867e-06,
+      "loss": 2.0881,
+      "step": 14719
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.399223901830962e-06,
+      "loss": 2.0659,
+      "step": 14720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3974827191165075e-06,
+      "loss": 2.0786,
+      "step": 14721
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.395741781660969e-06,
+      "loss": 2.0605,
+      "step": 14722
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.394001089510001e-06,
+      "loss": 2.0854,
+      "step": 14723
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.392260642709249e-06,
+      "loss": 2.0815,
+      "step": 14724
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.390520441304358e-06,
+      "loss": 2.0808,
+      "step": 14725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3887804853409595e-06,
+      "loss": 2.0916,
+      "step": 14726
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.387040774864681e-06,
+      "loss": 2.0889,
+      "step": 14727
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.385301309921147e-06,
+      "loss": 2.0879,
+      "step": 14728
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3835620905559735e-06,
+      "loss": 2.0176,
+      "step": 14729
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.381823116814771e-06,
+      "loss": 2.0237,
+      "step": 14730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.380084388743132e-06,
+      "loss": 2.0549,
+      "step": 14731
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.378345906386666e-06,
+      "loss": 2.0422,
+      "step": 14732
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.376607669790949e-06,
+      "loss": 2.0354,
+      "step": 14733
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.37486967900158e-06,
+      "loss": 2.0461,
+      "step": 14734
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.373131934064123e-06,
+      "loss": 2.0649,
+      "step": 14735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.371394435024154e-06,
+      "loss": 2.073,
+      "step": 14736
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.369657181927234e-06,
+      "loss": 2.0269,
+      "step": 14737
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.367920174818924e-06,
+      "loss": 2.0312,
+      "step": 14738
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.366183413744772e-06,
+      "loss": 2.0684,
+      "step": 14739
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.364446898750328e-06,
+      "loss": 2.1025,
+      "step": 14740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.362710629881117e-06,
+      "loss": 2.0471,
+      "step": 14741
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3609746071826835e-06,
+      "loss": 2.0562,
+      "step": 14742
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.359238830700549e-06,
+      "loss": 2.0349,
+      "step": 14743
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3575033004802315e-06,
+      "loss": 1.9937,
+      "step": 14744
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.35576801656725e-06,
+      "loss": 2.083,
+      "step": 14745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.354032979007093e-06,
+      "loss": 2.0701,
+      "step": 14746
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.352298187845281e-06,
+      "loss": 2.1382,
+      "step": 14747
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3505636431272855e-06,
+      "loss": 2.1169,
+      "step": 14748
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.348829344898614e-06,
+      "loss": 2.0928,
+      "step": 14749
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.347095293204731e-06,
+      "loss": 2.0684,
+      "step": 14750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.345361488091118e-06,
+      "loss": 2.0396,
+      "step": 14751
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.343627929603237e-06,
+      "loss": 2.0359,
+      "step": 14752
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.341894617786554e-06,
+      "loss": 2.0854,
+      "step": 14753
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.340161552686518e-06,
+      "loss": 2.0635,
+      "step": 14754
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.338428734348583e-06,
+      "loss": 1.9487,
+      "step": 14755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.336696162818179e-06,
+      "loss": 1.9688,
+      "step": 14756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.334963838140755e-06,
+      "loss": 2.1035,
+      "step": 14757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.333231760361723e-06,
+      "loss": 2.0886,
+      "step": 14758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.331499929526524e-06,
+      "loss": 2.103,
+      "step": 14759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.329768345680555e-06,
+      "loss": 2.0964,
+      "step": 14760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.328037008869235e-06,
+      "loss": 2.0686,
+      "step": 14761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.326305919137965e-06,
+      "loss": 2.0789,
+      "step": 14762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.324575076532138e-06,
+      "loss": 2.0979,
+      "step": 14763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.322844481097144e-06,
+      "loss": 2.083,
+      "step": 14764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3211141328783684e-06,
+      "loss": 2.0181,
+      "step": 14765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.319384031921186e-06,
+      "loss": 2.0476,
+      "step": 14766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.317654178270968e-06,
+      "loss": 2.0647,
+      "step": 14767
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3159245719730726e-06,
+      "loss": 2.051,
+      "step": 14768
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.314195213072866e-06,
+      "loss": 2.0686,
+      "step": 14769
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.312466101615693e-06,
+      "loss": 2.0613,
+      "step": 14770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.310737237646892e-06,
+      "loss": 2.0723,
+      "step": 14771
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.309008621211814e-06,
+      "loss": 2.0703,
+      "step": 14772
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.307280252355774e-06,
+      "loss": 2.0508,
+      "step": 14773
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.305552131124113e-06,
+      "loss": 2.0349,
+      "step": 14774
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.303824257562137e-06,
+      "loss": 2.0447,
+      "step": 14775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.3020966317151604e-06,
+      "loss": 2.0553,
+      "step": 14776
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.30036925362849e-06,
+      "loss": 1.9937,
+      "step": 14777
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.298642123347423e-06,
+      "loss": 2.0505,
+      "step": 14778
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.29691524091725e-06,
+      "loss": 2.1169,
+      "step": 14779
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.295188606383261e-06,
+      "loss": 2.1243,
+      "step": 14780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.293462219790728e-06,
+      "loss": 2.0408,
+      "step": 14781
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.291736081184933e-06,
+      "loss": 2.0364,
+      "step": 14782
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.290010190611133e-06,
+      "loss": 2.1062,
+      "step": 14783
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.288284548114593e-06,
+      "loss": 2.1113,
+      "step": 14784
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.286559153740568e-06,
+      "loss": 2.0286,
+      "step": 14785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2848340075342895e-06,
+      "loss": 2.0295,
+      "step": 14786
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.28310910954102e-06,
+      "loss": 2.0718,
+      "step": 14787
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.281384459805971e-06,
+      "loss": 2.0847,
+      "step": 14788
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.279660058374389e-06,
+      "loss": 2.0625,
+      "step": 14789
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.277935905291481e-06,
+      "loss": 2.063,
+      "step": 14790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2762120006024645e-06,
+      "loss": 2.0537,
+      "step": 14791
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.274488344352549e-06,
+      "loss": 2.0515,
+      "step": 14792
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.272764936586935e-06,
+      "loss": 2.1111,
+      "step": 14793
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.271041777350813e-06,
+      "loss": 2.0903,
+      "step": 14794
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2693188666893754e-06,
+      "loss": 2.0823,
+      "step": 14795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.267596204647801e-06,
+      "loss": 2.0662,
+      "step": 14796
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2658737912712664e-06,
+      "loss": 2.0295,
+      "step": 14797
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.264151626604938e-06,
+      "loss": 2.0376,
+      "step": 14798
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.262429710693979e-06,
+      "loss": 2.0457,
+      "step": 14799
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.260708043583549e-06,
+      "loss": 2.0052,
+      "step": 14800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.258986625318782e-06,
+      "loss": 2.0789,
+      "step": 14801
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2572654559448415e-06,
+      "loss": 2.1196,
+      "step": 14802
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.25554453550684e-06,
+      "loss": 2.0652,
+      "step": 14803
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.253823864049929e-06,
+      "loss": 2.0723,
+      "step": 14804
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.252103441619215e-06,
+      "loss": 2.0647,
+      "step": 14805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.250383268259819e-06,
+      "loss": 2.0527,
+      "step": 14806
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2486633440168526e-06,
+      "loss": 2.0791,
+      "step": 14807
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.246943668935416e-06,
+      "loss": 2.0762,
+      "step": 14808
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2452242430606085e-06,
+      "loss": 2.0652,
+      "step": 14809
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.243505066437523e-06,
+      "loss": 2.0898,
+      "step": 14810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.241786139111229e-06,
+      "loss": 2.0415,
+      "step": 14811
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.240067461126817e-06,
+      "loss": 2.0547,
+      "step": 14812
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.238349032529353e-06,
+      "loss": 2.0583,
+      "step": 14813
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.236630853363902e-06,
+      "loss": 2.0686,
+      "step": 14814
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2349129236755255e-06,
+      "loss": 2.0508,
+      "step": 14815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.23319524350926e-06,
+      "loss": 2.0569,
+      "step": 14816
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2314778129101665e-06,
+      "loss": 2.0339,
+      "step": 14817
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2297606319232634e-06,
+      "loss": 2.0264,
+      "step": 14818
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.228043700593606e-06,
+      "loss": 2.1033,
+      "step": 14819
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2263270189662016e-06,
+      "loss": 2.0598,
+      "step": 14820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.224610587086071e-06,
+      "loss": 2.0928,
+      "step": 14821
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.222894404998226e-06,
+      "loss": 2.0767,
+      "step": 14822
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.221178472747675e-06,
+      "loss": 2.0117,
+      "step": 14823
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.219462790379413e-06,
+      "loss": 2.0439,
+      "step": 14824
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.217747357938437e-06,
+      "loss": 2.0217,
+      "step": 14825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2160321754697185e-06,
+      "loss": 2.0422,
+      "step": 14826
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.214317243018257e-06,
+      "loss": 2.0671,
+      "step": 14827
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.212602560629001e-06,
+      "loss": 2.0447,
+      "step": 14828
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.210888128346942e-06,
+      "loss": 2.0879,
+      "step": 14829
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.209173946217017e-06,
+      "loss": 2.0835,
+      "step": 14830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.207460014284189e-06,
+      "loss": 2.1133,
+      "step": 14831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2057463325934e-06,
+      "loss": 2.0947,
+      "step": 14832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.204032901189592e-06,
+      "loss": 2.0649,
+      "step": 14833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.2023197201176955e-06,
+      "loss": 2.0842,
+      "step": 14834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.20060678942264e-06,
+      "loss": 2.0896,
+      "step": 14835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.198894109149341e-06,
+      "loss": 2.0769,
+      "step": 14836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.197181679342715e-06,
+      "loss": 2.0757,
+      "step": 14837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.195469500047667e-06,
+      "loss": 2.0654,
+      "step": 14838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.193757571309097e-06,
+      "loss": 2.0615,
+      "step": 14839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1920458931719025e-06,
+      "loss": 2.0757,
+      "step": 14840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1903344656809576e-06,
+      "loss": 2.0608,
+      "step": 14841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.188623288881159e-06,
+      "loss": 2.0586,
+      "step": 14842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.186912362817365e-06,
+      "loss": 2.0518,
+      "step": 14843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.185201687534457e-06,
+      "loss": 2.0752,
+      "step": 14844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.183491263077287e-06,
+      "loss": 2.0754,
+      "step": 14845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1817810894907075e-06,
+      "loss": 2.0544,
+      "step": 14846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.18007116681957e-06,
+      "loss": 2.0457,
+      "step": 14847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.178361495108713e-06,
+      "loss": 2.0598,
+      "step": 14848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1766520744029735e-06,
+      "loss": 2.0786,
+      "step": 14849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.174942904747174e-06,
+      "loss": 2.0789,
+      "step": 14850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.173233986186141e-06,
+      "loss": 2.1189,
+      "step": 14851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.171525318764687e-06,
+      "loss": 2.1184,
+      "step": 14852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.169816902527618e-06,
+      "loss": 2.0586,
+      "step": 14853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.168108737519737e-06,
+      "loss": 2.0483,
+      "step": 14854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1664008237858374e-06,
+      "loss": 2.0454,
+      "step": 14855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.164693161370709e-06,
+      "loss": 2.0752,
+      "step": 14856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.162985750319139e-06,
+      "loss": 2.0493,
+      "step": 14857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.161278590675884e-06,
+      "loss": 2.0571,
+      "step": 14858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.159571682485735e-06,
+      "loss": 2.0464,
+      "step": 14859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1578650257934315e-06,
+      "loss": 2.0532,
+      "step": 14860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.156158620643748e-06,
+      "loss": 2.0662,
+      "step": 14861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.154452467081424e-06,
+      "loss": 2.0549,
+      "step": 14862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1527465651511986e-06,
+      "loss": 2.0347,
+      "step": 14863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.151040914897812e-06,
+      "loss": 2.0557,
+      "step": 14864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.149335516365991e-06,
+      "loss": 2.1196,
+      "step": 14865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.14763036960046e-06,
+      "loss": 2.1296,
+      "step": 14866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.145925474645932e-06,
+      "loss": 2.0342,
+      "step": 14867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.144220831547113e-06,
+      "loss": 2.0764,
+      "step": 14868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.142516440348714e-06,
+      "loss": 2.0366,
+      "step": 14869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.140812301095423e-06,
+      "loss": 2.0166,
+      "step": 14870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.139108413831932e-06,
+      "loss": 2.0232,
+      "step": 14871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.137404778602927e-06,
+      "loss": 2.0286,
+      "step": 14872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.135701395453072e-06,
+      "loss": 2.0461,
+      "step": 14873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.133998264427053e-06,
+      "loss": 2.0693,
+      "step": 14874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.1322953855695155e-06,
+      "loss": 2.0142,
+      "step": 14875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.130592758925132e-06,
+      "loss": 2.0547,
+      "step": 14876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.12889038453854e-06,
+      "loss": 2.0618,
+      "step": 14877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.127188262454384e-06,
+      "loss": 2.1145,
+      "step": 14878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.125486392717305e-06,
+      "loss": 2.0347,
+      "step": 14879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.123784775371927e-06,
+      "loss": 2.0686,
+      "step": 14880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.122083410462879e-06,
+      "loss": 2.0752,
+      "step": 14881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.120382298034773e-06,
+      "loss": 2.092,
+      "step": 14882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.118681438132218e-06,
+      "loss": 2.0442,
+      "step": 14883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.11698083079982e-06,
+      "loss": 2.0474,
+      "step": 14884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.115280476082174e-06,
+      "loss": 2.0479,
+      "step": 14885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.113580374023871e-06,
+      "loss": 2.0669,
+      "step": 14886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.111880524669497e-06,
+      "loss": 2.0347,
+      "step": 14887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.110180928063616e-06,
+      "loss": 2.0178,
+      "step": 14888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.108481584250818e-06,
+      "loss": 2.0793,
+      "step": 14889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.106782493275642e-06,
+      "loss": 2.1194,
+      "step": 14890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.10508365518267e-06,
+      "loss": 2.0088,
+      "step": 14891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.103385070016434e-06,
+      "loss": 2.0293,
+      "step": 14892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.101686737821486e-06,
+      "loss": 2.1182,
+      "step": 14893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.099988658642356e-06,
+      "loss": 2.1367,
+      "step": 14894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.098290832523582e-06,
+      "loss": 2.1204,
+      "step": 14895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.096593259509681e-06,
+      "loss": 2.1519,
+      "step": 14896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.09489593964518e-06,
+      "loss": 2.1033,
+      "step": 14897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.09319887297457e-06,
+      "loss": 2.084,
+      "step": 14898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.091502059542377e-06,
+      "loss": 2.0767,
+      "step": 14899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.08980549939308e-06,
+      "loss": 2.0581,
+      "step": 14900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.088109192571185e-06,
+      "loss": 2.0715,
+      "step": 14901
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0864131391211615e-06,
+      "loss": 2.0684,
+      "step": 14902
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.084717339087493e-06,
+      "loss": 2.0371,
+      "step": 14903
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0830217925146524e-06,
+      "loss": 2.0457,
+      "step": 14904
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.081326499447097e-06,
+      "loss": 2.0566,
+      "step": 14905
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0796314599292875e-06,
+      "loss": 2.0767,
+      "step": 14906
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.077936674005675e-06,
+      "loss": 2.082,
+      "step": 14907
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.076242141720703e-06,
+      "loss": 2.1069,
+      "step": 14908
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.074547863118809e-06,
+      "loss": 1.999,
+      "step": 14909
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.07285383824442e-06,
+      "loss": 2.0522,
+      "step": 14910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.071160067141966e-06,
+      "loss": 2.1086,
+      "step": 14911
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.069466549855863e-06,
+      "loss": 2.0793,
+      "step": 14912
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.06777328643051e-06,
+      "loss": 2.0034,
+      "step": 14913
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0660802769103295e-06,
+      "loss": 2.0295,
+      "step": 14914
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.064387521339701e-06,
+      "loss": 2.0383,
+      "step": 14915
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.062695019763032e-06,
+      "loss": 2.0613,
+      "step": 14916
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.061002772224692e-06,
+      "loss": 2.0845,
+      "step": 14917
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.059310778769065e-06,
+      "loss": 2.092,
+      "step": 14918
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.057619039440521e-06,
+      "loss": 2.1252,
+      "step": 14919
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.055927554283424e-06,
+      "loss": 2.1035,
+      "step": 14920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.054236323342127e-06,
+      "loss": 2.0623,
+      "step": 14921
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.052545346660987e-06,
+      "loss": 2.0601,
+      "step": 14922
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.050854624284346e-06,
+      "loss": 2.073,
+      "step": 14923
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.04916415625654e-06,
+      "loss": 2.084,
+      "step": 14924
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0474739426218965e-06,
+      "loss": 2.0835,
+      "step": 14925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.045783983424746e-06,
+      "loss": 2.0959,
+      "step": 14926
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.044094278709406e-06,
+      "loss": 2.041,
+      "step": 14927
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0424048285201774e-06,
+      "loss": 2.053,
+      "step": 14928
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.040715632901376e-06,
+      "loss": 2.0437,
+      "step": 14929
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.039026691897285e-06,
+      "loss": 2.0696,
+      "step": 14930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0373380055522124e-06,
+      "loss": 2.0425,
+      "step": 14931
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.035649573910427e-06,
+      "loss": 2.0598,
+      "step": 14932
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.033961397016214e-06,
+      "loss": 2.0986,
+      "step": 14933
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.032273474913842e-06,
+      "loss": 2.1199,
+      "step": 14934
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.030585807647573e-06,
+      "loss": 2.0913,
+      "step": 14935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0288983952616665e-06,
+      "loss": 2.1033,
+      "step": 14936
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.02721123780037e-06,
+      "loss": 2.0022,
+      "step": 14937
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0255243353079295e-06,
+      "loss": 2.03,
+      "step": 14938
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.023837687828583e-06,
+      "loss": 2.0513,
+      "step": 14939
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0221512954065595e-06,
+      "loss": 2.0979,
+      "step": 14940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.020465158086084e-06,
+      "loss": 2.106,
+      "step": 14941
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0187792759113754e-06,
+      "loss": 2.1062,
+      "step": 14942
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.017093648926632e-06,
+      "loss": 2.0093,
+      "step": 14943
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.015408277176076e-06,
+      "loss": 2.0117,
+      "step": 14944
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.013723160703885e-06,
+      "loss": 1.9983,
+      "step": 14945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.012038299554269e-06,
+      "loss": 1.9897,
+      "step": 14946
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.010353693771395e-06,
+      "loss": 2.0154,
+      "step": 14947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.008669343399448e-06,
+      "loss": 2.0493,
+      "step": 14948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0069852484825954e-06,
+      "loss": 2.0598,
+      "step": 14949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.005301409065003e-06,
+      "loss": 2.075,
+      "step": 14950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.003617825190826e-06,
+      "loss": 2.1262,
+      "step": 14951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.001934496904213e-06,
+      "loss": 2.0991,
+      "step": 14952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.000251424249309e-06,
+      "loss": 2.0383,
+      "step": 14953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.998568607270252e-06,
+      "loss": 2.0593,
+      "step": 14954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.99688604601117e-06,
+      "loss": 2.0691,
+      "step": 14955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.995203740516187e-06,
+      "loss": 2.0789,
+      "step": 14956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.993521690829422e-06,
+      "loss": 2.0906,
+      "step": 14957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.991839896994975e-06,
+      "loss": 2.0686,
+      "step": 14958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.990158359056966e-06,
+      "loss": 2.0447,
+      "step": 14959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.988477077059471e-06,
+      "loss": 2.0598,
+      "step": 14960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9867960510466e-06,
+      "loss": 2.0225,
+      "step": 14961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.98511528106242e-06,
+      "loss": 2.0691,
+      "step": 14962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.983434767151013e-06,
+      "loss": 2.0774,
+      "step": 14963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.981754509356451e-06,
+      "loss": 2.0664,
+      "step": 14964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.980074507722794e-06,
+      "loss": 2.0476,
+      "step": 14965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.978394762294098e-06,
+      "loss": 2.053,
+      "step": 14966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9767152731144195e-06,
+      "loss": 2.0471,
+      "step": 14967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.975036040227786e-06,
+      "loss": 2.0803,
+      "step": 14968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.973357063678248e-06,
+      "loss": 2.0706,
+      "step": 14969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.971678343509822e-06,
+      "loss": 2.0964,
+      "step": 14970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9699998797665476e-06,
+      "loss": 2.0681,
+      "step": 14971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.968321672492418e-06,
+      "loss": 2.0894,
+      "step": 14972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.966643721731468e-06,
+      "loss": 2.0752,
+      "step": 14973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.964966027527679e-06,
+      "loss": 2.095,
+      "step": 14974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.963288589925057e-06,
+      "loss": 2.0537,
+      "step": 14975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.961611408967584e-06,
+      "loss": 2.0415,
+      "step": 14976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.959934484699246e-06,
+      "loss": 2.0759,
+      "step": 14977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9582578171640245e-06,
+      "loss": 2.0701,
+      "step": 14978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.956581406405878e-06,
+      "loss": 2.0784,
+      "step": 14979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.954905252468773e-06,
+      "loss": 2.0664,
+      "step": 14980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.953229355396665e-06,
+      "loss": 2.0317,
+      "step": 14981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.951553715233502e-06,
+      "loss": 2.0737,
+      "step": 14982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.949878332023224e-06,
+      "loss": 2.0535,
+      "step": 14983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.948203205809773e-06,
+      "loss": 2.0688,
+      "step": 14984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.946528336637063e-06,
+      "loss": 2.0439,
+      "step": 14985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.944853724549033e-06,
+      "loss": 2.0627,
+      "step": 14986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.94317936958958e-06,
+      "loss": 2.0259,
+      "step": 14987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.94150527180263e-06,
+      "loss": 2.0569,
+      "step": 14988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.939831431232071e-06,
+      "loss": 2.0449,
+      "step": 14989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.938157847921801e-06,
+      "loss": 2.0276,
+      "step": 14990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.936484521915709e-06,
+      "loss": 2.0779,
+      "step": 14991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.934811453257675e-06,
+      "loss": 2.1025,
+      "step": 14992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.933138641991575e-06,
+      "loss": 2.0308,
+      "step": 14993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.931466088161274e-06,
+      "loss": 2.0247,
+      "step": 14994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.929793791810636e-06,
+      "loss": 2.0557,
+      "step": 14995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.928121752983512e-06,
+      "loss": 2.0742,
+      "step": 14996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.926449971723752e-06,
+      "loss": 2.1047,
+      "step": 14997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.924778448075195e-06,
+      "loss": 2.1125,
+      "step": 14998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.923107182081677e-06,
+      "loss": 2.0376,
+      "step": 14999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.921436173787014e-06,
+      "loss": 2.0435,
+      "step": 15000
+    },
+    {
+      "Bleu_1": 0.7741305844618093,
+      "Bleu_2": 0.6132155320133599,
+      "Bleu_3": 0.4714479784218381,
+      "Bleu_4": 0.3578334169553194,
+      "CIDEr": 1.2050090458347438,
+      "METEOR": 0.27848017130606645,
+      "ROUGE_L": 0.5602285118687599,
+      "epoch": 0.75,
+      "step": 15000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.919765423235047e-06,
+      "loss": 2.0437,
+      "step": 15001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9180949304695664e-06,
+      "loss": 2.0452,
+      "step": 15002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.916424695534398e-06,
+      "loss": 2.082,
+      "step": 15003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.914754718473334e-06,
+      "loss": 2.0806,
+      "step": 15004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.913084999330163e-06,
+      "loss": 2.0664,
+      "step": 15005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.911415538148677e-06,
+      "loss": 2.0588,
+      "step": 15006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.909746334972656e-06,
+      "loss": 2.0791,
+      "step": 15007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.908077389845869e-06,
+      "loss": 2.0754,
+      "step": 15008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.906408702812088e-06,
+      "loss": 2.0867,
+      "step": 15009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.904740273915067e-06,
+      "loss": 2.0896,
+      "step": 15010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.903072103198561e-06,
+      "loss": 2.0586,
+      "step": 15011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.901404190706316e-06,
+      "loss": 2.0801,
+      "step": 15012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.899736536482069e-06,
+      "loss": 2.0422,
+      "step": 15013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.898069140569562e-06,
+      "loss": 2.0757,
+      "step": 15014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8964020030124986e-06,
+      "loss": 2.0654,
+      "step": 15015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.894735123854625e-06,
+      "loss": 2.0576,
+      "step": 15016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.893068503139627e-06,
+      "loss": 2.0398,
+      "step": 15017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8914021409112336e-06,
+      "loss": 2.0332,
+      "step": 15018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.889736037213126e-06,
+      "loss": 2.0547,
+      "step": 15019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.888070192089001e-06,
+      "loss": 2.0593,
+      "step": 15020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.886404605582546e-06,
+      "loss": 2.0674,
+      "step": 15021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.884739277737437e-06,
+      "loss": 2.0735,
+      "step": 15022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.883074208597347e-06,
+      "loss": 2.043,
+      "step": 15023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.881409398205938e-06,
+      "loss": 2.0491,
+      "step": 15024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.879744846606868e-06,
+      "loss": 2.0229,
+      "step": 15025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.878080553843789e-06,
+      "loss": 2.0271,
+      "step": 15026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8764165199603444e-06,
+      "loss": 2.105,
+      "step": 15027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.874752745000173e-06,
+      "loss": 2.1177,
+      "step": 15028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.87308922900691e-06,
+      "loss": 2.0913,
+      "step": 15029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.871425972024162e-06,
+      "loss": 2.1235,
+      "step": 15030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.86976297409557e-06,
+      "loss": 2.1025,
+      "step": 15031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.868100235264718e-06,
+      "loss": 2.1179,
+      "step": 15032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8664377555752354e-06,
+      "loss": 2.0518,
+      "step": 15033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8647755350707036e-06,
+      "loss": 2.0581,
+      "step": 15034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.863113573794712e-06,
+      "loss": 2.019,
+      "step": 15035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.861451871790849e-06,
+      "loss": 2.0283,
+      "step": 15036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8597904291026885e-06,
+      "loss": 2.0386,
+      "step": 15037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.858129245773799e-06,
+      "loss": 2.0613,
+      "step": 15038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.856468321847752e-06,
+      "loss": 2.0933,
+      "step": 15039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8548076573680845e-06,
+      "loss": 2.083,
+      "step": 15040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.853147252378366e-06,
+      "loss": 2.1013,
+      "step": 15041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.85148710692212e-06,
+      "loss": 2.103,
+      "step": 15042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8498272210429e-06,
+      "loss": 2.0657,
+      "step": 15043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.848167594784221e-06,
+      "loss": 2.0715,
+      "step": 15044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.846508228189612e-06,
+      "loss": 2.0549,
+      "step": 15045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.844849121302583e-06,
+      "loss": 2.0833,
+      "step": 15046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8431902741666405e-06,
+      "loss": 2.0571,
+      "step": 15047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8415316868253e-06,
+      "loss": 2.0776,
+      "step": 15048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8398733593220405e-06,
+      "loss": 2.0881,
+      "step": 15049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.838215291700356e-06,
+      "loss": 2.0376,
+      "step": 15050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.836557484003726e-06,
+      "loss": 2.0706,
+      "step": 15051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.834899936275626e-06,
+      "loss": 2.0952,
+      "step": 15052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.833242648559523e-06,
+      "loss": 2.0005,
+      "step": 15053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.83158562089888e-06,
+      "loss": 2.0181,
+      "step": 15054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.82992885333714e-06,
+      "loss": 2.0774,
+      "step": 15055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.828272345917764e-06,
+      "loss": 2.1111,
+      "step": 15056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.826616098684177e-06,
+      "loss": 2.0608,
+      "step": 15057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.824960111679833e-06,
+      "loss": 2.0569,
+      "step": 15058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.823304384948139e-06,
+      "loss": 2.0811,
+      "step": 15059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.821648918532522e-06,
+      "loss": 2.0642,
+      "step": 15060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.819993712476394e-06,
+      "loss": 2.0608,
+      "step": 15061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.818338766823159e-06,
+      "loss": 2.0735,
+      "step": 15062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.816684081616219e-06,
+      "loss": 2.0654,
+      "step": 15063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.815029656898966e-06,
+      "loss": 2.0662,
+      "step": 15064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.813375492714786e-06,
+      "loss": 2.1135,
+      "step": 15065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.811721589107053e-06,
+      "loss": 2.1255,
+      "step": 15066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8100679461191446e-06,
+      "loss": 2.0732,
+      "step": 15067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.80841456379442e-06,
+      "loss": 2.0703,
+      "step": 15068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.806761442176247e-06,
+      "loss": 2.0142,
+      "step": 15069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8051085813079585e-06,
+      "loss": 2.0234,
+      "step": 15070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.80345598123292e-06,
+      "loss": 2.0227,
+      "step": 15071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.801803641994451e-06,
+      "loss": 2.0356,
+      "step": 15072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.800151563635901e-06,
+      "loss": 2.0474,
+      "step": 15073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.798499746200576e-06,
+      "loss": 2.0105,
+      "step": 15074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.796848189731801e-06,
+      "loss": 1.9905,
+      "step": 15075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.795196894272887e-06,
+      "loss": 2.0234,
+      "step": 15076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.793545859867134e-06,
+      "loss": 2.0874,
+      "step": 15077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.7918950865578395e-06,
+      "loss": 2.0886,
+      "step": 15078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.790244574388297e-06,
+      "loss": 2.051,
+      "step": 15079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.788594323401785e-06,
+      "loss": 2.0696,
+      "step": 15080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.786944333641578e-06,
+      "loss": 2.0713,
+      "step": 15081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.785294605150951e-06,
+      "loss": 2.0449,
+      "step": 15082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.783645137973159e-06,
+      "loss": 2.1125,
+      "step": 15083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.781995932151468e-06,
+      "loss": 2.0808,
+      "step": 15084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.780346987729111e-06,
+      "loss": 2.0486,
+      "step": 15085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.778698304749346e-06,
+      "loss": 2.0322,
+      "step": 15086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.777049883255393e-06,
+      "loss": 2.0359,
+      "step": 15087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.7754017232904936e-06,
+      "loss": 2.0344,
+      "step": 15088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.773753824897859e-06,
+      "loss": 2.0244,
+      "step": 15089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.7721061881207075e-06,
+      "loss": 2.0449,
+      "step": 15090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.770458813002245e-06,
+      "loss": 2.0527,
+      "step": 15091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.7688116995856715e-06,
+      "loss": 2.0542,
+      "step": 15092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.767164847914184e-06,
+      "loss": 2.0693,
+      "step": 15093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.765518258030965e-06,
+      "loss": 2.1138,
+      "step": 15094
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.7638719299792e-06,
+      "loss": 2.0239,
+      "step": 15095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.762225863802055e-06,
+      "loss": 2.0444,
+      "step": 15096
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.760580059542702e-06,
+      "loss": 2.0852,
+      "step": 15097
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.758934517244298e-06,
+      "loss": 2.0811,
+      "step": 15098
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.757289236949993e-06,
+      "loss": 2.0544,
+      "step": 15099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.755644218702937e-06,
+      "loss": 2.0698,
+      "step": 15100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.753999462546271e-06,
+      "loss": 2.043,
+      "step": 15101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.752354968523114e-06,
+      "loss": 2.0811,
+      "step": 15102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.750710736676609e-06,
+      "loss": 2.0601,
+      "step": 15103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.749066767049854e-06,
+      "loss": 2.0659,
+      "step": 15104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.747423059685982e-06,
+      "loss": 2.0745,
+      "step": 15105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.74577961462808e-06,
+      "loss": 2.0891,
+      "step": 15106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7441364319192504e-06,
+      "loss": 2.0646,
+      "step": 15107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.742493511602586e-06,
+      "loss": 2.0515,
+      "step": 15108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7408508537211705e-06,
+      "loss": 2.0366,
+      "step": 15109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7392084583180795e-06,
+      "loss": 2.0396,
+      "step": 15110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.737566325436386e-06,
+      "loss": 2.1106,
+      "step": 15111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7359244551191435e-06,
+      "loss": 2.0708,
+      "step": 15112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.734282847409422e-06,
+      "loss": 2.0461,
+      "step": 15113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.732641502350254e-06,
+      "loss": 2.0581,
+      "step": 15114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.731000419984702e-06,
+      "loss": 2.052,
+      "step": 15115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.729359600355785e-06,
+      "loss": 2.0713,
+      "step": 15116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.727719043506532e-06,
+      "loss": 2.1006,
+      "step": 15117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.726078749479981e-06,
+      "loss": 2.043,
+      "step": 15118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.724438718319126e-06,
+      "loss": 2.0664,
+      "step": 15119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.722798950066994e-06,
+      "loss": 2.0627,
+      "step": 15120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.721159444766573e-06,
+      "loss": 2.0874,
+      "step": 15121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.719520202460859e-06,
+      "loss": 2.0642,
+      "step": 15122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.717881223192843e-06,
+      "loss": 2.1138,
+      "step": 15123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.716242507005504e-06,
+      "loss": 2.0869,
+      "step": 15124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.714604053941813e-06,
+      "loss": 2.0378,
+      "step": 15125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.712965864044743e-06,
+      "loss": 2.0662,
+      "step": 15126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.711327937357242e-06,
+      "loss": 2.0562,
+      "step": 15127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.709690273922277e-06,
+      "loss": 2.0901,
+      "step": 15128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.70805287378278e-06,
+      "loss": 2.061,
+      "step": 15129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.706415736981703e-06,
+      "loss": 2.1494,
+      "step": 15130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.704778863561968e-06,
+      "loss": 2.0513,
+      "step": 15131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7031422535665045e-06,
+      "loss": 2.043,
+      "step": 15132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.70150590703823e-06,
+      "loss": 2.1057,
+      "step": 15133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.699869824020054e-06,
+      "loss": 2.1021,
+      "step": 15134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.698234004554884e-06,
+      "loss": 2.0709,
+      "step": 15135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.696598448685616e-06,
+      "loss": 2.092,
+      "step": 15136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.694963156455141e-06,
+      "loss": 2.033,
+      "step": 15137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.693328127906341e-06,
+      "loss": 2.0312,
+      "step": 15138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.691693363082094e-06,
+      "loss": 2.011,
+      "step": 15139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.69005886202527e-06,
+      "loss": 2.0,
+      "step": 15140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.688424624778737e-06,
+      "loss": 2.0098,
+      "step": 15141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.686790651385336e-06,
+      "loss": 2.0137,
+      "step": 15142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.685156941887934e-06,
+      "loss": 2.0986,
+      "step": 15143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.683523496329356e-06,
+      "loss": 2.1277,
+      "step": 15144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.681890314752456e-06,
+      "loss": 2.0293,
+      "step": 15145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.680257397200046e-06,
+      "loss": 2.0393,
+      "step": 15146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.678624743714955e-06,
+      "loss": 1.9841,
+      "step": 15147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.676992354339997e-06,
+      "loss": 1.9717,
+      "step": 15148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.675360229117979e-06,
+      "loss": 2.0991,
+      "step": 15149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6737283680917e-06,
+      "loss": 2.0627,
+      "step": 15150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.672096771303955e-06,
+      "loss": 2.0239,
+      "step": 15151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.670465438797532e-06,
+      "loss": 2.0276,
+      "step": 15152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.668834370615208e-06,
+      "loss": 2.0735,
+      "step": 15153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.667203566799758e-06,
+      "loss": 2.0315,
+      "step": 15154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.665573027393946e-06,
+      "loss": 2.0667,
+      "step": 15155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.663942752440538e-06,
+      "loss": 2.0928,
+      "step": 15156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.662312741982271e-06,
+      "loss": 2.0994,
+      "step": 15157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.660682996061908e-06,
+      "loss": 2.0974,
+      "step": 15158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.659053514722169e-06,
+      "loss": 2.0188,
+      "step": 15159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.657424298005805e-06,
+      "loss": 2.0339,
+      "step": 15160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.655795345955527e-06,
+      "loss": 2.1062,
+      "step": 15161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.654166658614053e-06,
+      "loss": 2.1392,
+      "step": 15162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.652538236024099e-06,
+      "loss": 2.0366,
+      "step": 15163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.650910078228363e-06,
+      "loss": 2.0216,
+      "step": 15164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.649282185269546e-06,
+      "loss": 1.9595,
+      "step": 15165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.647654557190334e-06,
+      "loss": 1.9961,
+      "step": 15166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.646027194033413e-06,
+      "loss": 2.0588,
+      "step": 15167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.644400095841456e-06,
+      "loss": 2.0359,
+      "step": 15168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.642773262657132e-06,
+      "loss": 2.1333,
+      "step": 15169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.641146694523106e-06,
+      "loss": 2.1509,
+      "step": 15170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.639520391482033e-06,
+      "loss": 2.0686,
+      "step": 15171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.637894353576551e-06,
+      "loss": 2.0522,
+      "step": 15172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.636268580849316e-06,
+      "loss": 2.0588,
+      "step": 15173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.634643073342947e-06,
+      "loss": 2.053,
+      "step": 15174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6330178311000855e-06,
+      "loss": 2.0591,
+      "step": 15175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6313928541633415e-06,
+      "loss": 2.02,
+      "step": 15176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.62976814257533e-06,
+      "loss": 2.0229,
+      "step": 15177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.628143696378659e-06,
+      "loss": 2.0381,
+      "step": 15178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.626519515615927e-06,
+      "loss": 2.0293,
+      "step": 15179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6248956003297265e-06,
+      "loss": 2.0625,
+      "step": 15180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.623271950562646e-06,
+      "loss": 2.0911,
+      "step": 15181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.621648566357252e-06,
+      "loss": 2.1167,
+      "step": 15182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6200254477561345e-06,
+      "loss": 2.0537,
+      "step": 15183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.618402594801839e-06,
+      "loss": 2.0757,
+      "step": 15184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.616780007536933e-06,
+      "loss": 2.084,
+      "step": 15185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.615157686003972e-06,
+      "loss": 2.0806,
+      "step": 15186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.613535630245486e-06,
+      "loss": 2.0496,
+      "step": 15187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.611913840304025e-06,
+      "loss": 2.0593,
+      "step": 15188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.610292316222105e-06,
+      "loss": 2.0547,
+      "step": 15189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.608671058042263e-06,
+      "loss": 2.0491,
+      "step": 15190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.607050065807004e-06,
+      "loss": 2.0903,
+      "step": 15191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.605429339558839e-06,
+      "loss": 2.0999,
+      "step": 15192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.603808879340273e-06,
+      "loss": 2.0757,
+      "step": 15193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.602188685193797e-06,
+      "loss": 2.0732,
+      "step": 15194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.600568757161899e-06,
+      "loss": 2.0356,
+      "step": 15195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.598949095287068e-06,
+      "loss": 2.001,
+      "step": 15196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5973296996117586e-06,
+      "loss": 2.0818,
+      "step": 15197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.59571057017846e-06,
+      "loss": 2.0732,
+      "step": 15198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.594091707029614e-06,
+      "loss": 2.085,
+      "step": 15199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.592473110207688e-06,
+      "loss": 2.0563,
+      "step": 15200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.590854779755116e-06,
+      "loss": 2.0872,
+      "step": 15201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.589236715714343e-06,
+      "loss": 2.1099,
+      "step": 15202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.587618918127799e-06,
+      "loss": 2.0771,
+      "step": 15203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5860013870379086e-06,
+      "loss": 2.0813,
+      "step": 15204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.584384122487093e-06,
+      "loss": 2.0928,
+      "step": 15205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.582767124517757e-06,
+      "loss": 2.0608,
+      "step": 15206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.581150393172309e-06,
+      "loss": 2.0233,
+      "step": 15207
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.579533928493147e-06,
+      "loss": 2.0728,
+      "step": 15208
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.577917730522657e-06,
+      "loss": 2.0833,
+      "step": 15209
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.576301799303222e-06,
+      "loss": 2.0693,
+      "step": 15210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.574686134877224e-06,
+      "loss": 2.0352,
+      "step": 15211
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.573070737287026e-06,
+      "loss": 2.0378,
+      "step": 15212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.571455606574995e-06,
+      "loss": 2.1108,
+      "step": 15213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5698407427834735e-06,
+      "loss": 2.1045,
+      "step": 15214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.568226145954827e-06,
+      "loss": 2.0825,
+      "step": 15215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.566611816131379e-06,
+      "loss": 2.0867,
+      "step": 15216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.564997753355483e-06,
+      "loss": 2.116,
+      "step": 15217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.563383957669449e-06,
+      "loss": 2.0959,
+      "step": 15218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.561770429115603e-06,
+      "loss": 2.1045,
+      "step": 15219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.560157167736258e-06,
+      "loss": 2.0569,
+      "step": 15220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.558544173573721e-06,
+      "loss": 2.1055,
+      "step": 15221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.556931446670289e-06,
+      "loss": 2.0852,
+      "step": 15222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.555318987068256e-06,
+      "loss": 2.0603,
+      "step": 15223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.553706794809904e-06,
+      "loss": 2.0879,
+      "step": 15224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.552094869937513e-06,
+      "loss": 2.0913,
+      "step": 15225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.550483212493352e-06,
+      "loss": 2.1123,
+      "step": 15226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.548871822519687e-06,
+      "loss": 2.0647,
+      "step": 15227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.547260700058781e-06,
+      "loss": 2.0308,
+      "step": 15228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.545649845152865e-06,
+      "loss": 2.0881,
+      "step": 15229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.544039257844203e-06,
+      "loss": 2.0947,
+      "step": 15230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.542428938175012e-06,
+      "loss": 2.0742,
+      "step": 15231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.54081888618754e-06,
+      "loss": 2.0789,
+      "step": 15232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.539209101923994e-06,
+      "loss": 2.0391,
+      "step": 15233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.537599585426595e-06,
+      "loss": 2.0154,
+      "step": 15234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.535990336737548e-06,
+      "loss": 2.071,
+      "step": 15235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5343813558990545e-06,
+      "loss": 2.054,
+      "step": 15236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.53277264295331e-06,
+      "loss": 2.0603,
+      "step": 15237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.531164197942496e-06,
+      "loss": 2.1011,
+      "step": 15238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.529556020908801e-06,
+      "loss": 2.0474,
+      "step": 15239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.52794811189439e-06,
+      "loss": 2.062,
+      "step": 15240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.526340470941433e-06,
+      "loss": 2.0498,
+      "step": 15241
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5247330980920846e-06,
+      "loss": 2.0586,
+      "step": 15242
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.523125993388502e-06,
+      "loss": 2.0779,
+      "step": 15243
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.521519156872818e-06,
+      "loss": 1.9895,
+      "step": 15244
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.519912588587187e-06,
+      "loss": 2.0386,
+      "step": 15245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.518306288573723e-06,
+      "loss": 1.9976,
+      "step": 15246
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.516700256874562e-06,
+      "loss": 2.0364,
+      "step": 15247
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.515094493531813e-06,
+      "loss": 2.0356,
+      "step": 15248
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.513488998587587e-06,
+      "loss": 2.0972,
+      "step": 15249
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5118837720839856e-06,
+      "loss": 2.0708,
+      "step": 15250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.510278814063105e-06,
+      "loss": 2.0159,
+      "step": 15251
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.508674124567034e-06,
+      "loss": 2.0297,
+      "step": 15252
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5070697036378565e-06,
+      "loss": 2.0444,
+      "step": 15253
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.505465551317635e-06,
+      "loss": 2.0396,
+      "step": 15254
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5038616676484485e-06,
+      "loss": 2.0544,
+      "step": 15255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5022580526723524e-06,
+      "loss": 2.0833,
+      "step": 15256
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.500654706431401e-06,
+      "loss": 2.1079,
+      "step": 15257
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.499051628967645e-06,
+      "loss": 2.1372,
+      "step": 15258
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.497448820323108e-06,
+      "loss": 2.1152,
+      "step": 15259
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.495846280539842e-06,
+      "loss": 2.0879,
+      "step": 15260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4942440096598485e-06,
+      "loss": 2.1189,
+      "step": 15261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4926420077251705e-06,
+      "loss": 2.113,
+      "step": 15262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4910402747778e-06,
+      "loss": 2.0408,
+      "step": 15263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.489438810859747e-06,
+      "loss": 2.0649,
+      "step": 15264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.487837616013008e-06,
+      "loss": 1.9631,
+      "step": 15265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.486236690279573e-06,
+      "loss": 1.9795,
+      "step": 15266
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.484636033701424e-06,
+      "loss": 2.0071,
+      "step": 15267
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.483035646320537e-06,
+      "loss": 2.0085,
+      "step": 15268
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.481435528178875e-06,
+      "loss": 2.0835,
+      "step": 15269
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.47983567931841e-06,
+      "loss": 2.0906,
+      "step": 15270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4782360997810805e-06,
+      "loss": 2.1121,
+      "step": 15271
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.476636789608853e-06,
+      "loss": 2.1082,
+      "step": 15272
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.475037748843651e-06,
+      "loss": 2.0823,
+      "step": 15273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4734389775274155e-06,
+      "loss": 2.1067,
+      "step": 15274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.47184047570207e-06,
+      "loss": 2.0542,
+      "step": 15275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.470242243409536e-06,
+      "loss": 2.0586,
+      "step": 15276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.468644280691723e-06,
+      "loss": 2.0701,
+      "step": 15277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4670465875905345e-06,
+      "loss": 2.0776,
+      "step": 15278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.465449164147872e-06,
+      "loss": 2.0906,
+      "step": 15279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4638520104056226e-06,
+      "loss": 2.1011,
+      "step": 15280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4622551264056714e-06,
+      "loss": 2.0627,
+      "step": 15281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.460658512189896e-06,
+      "loss": 2.0781,
+      "step": 15282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.459062167800166e-06,
+      "loss": 2.0959,
+      "step": 15283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.457466093278333e-06,
+      "loss": 2.1028,
+      "step": 15284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.455870288666271e-06,
+      "loss": 2.0793,
+      "step": 15285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.454274754005811e-06,
+      "loss": 2.0215,
+      "step": 15286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.45267948933881e-06,
+      "loss": 2.0886,
+      "step": 15287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.451084494707085e-06,
+      "loss": 2.0813,
+      "step": 15288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.449489770152477e-06,
+      "loss": 2.0969,
+      "step": 15289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.447895315716796e-06,
+      "loss": 2.0593,
+      "step": 15290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.446301131441859e-06,
+      "loss": 2.1057,
+      "step": 15291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.444707217369473e-06,
+      "loss": 2.1013,
+      "step": 15292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.443113573541435e-06,
+      "loss": 2.0688,
+      "step": 15293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.441520199999533e-06,
+      "loss": 2.0195,
+      "step": 15294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.439927096785559e-06,
+      "loss": 2.0601,
+      "step": 15295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.438334263941285e-06,
+      "loss": 2.0649,
+      "step": 15296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.43674170150848e-06,
+      "loss": 2.0859,
+      "step": 15297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.435149409528917e-06,
+      "loss": 2.0964,
+      "step": 15298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.433557388044333e-06,
+      "loss": 2.1138,
+      "step": 15299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.431965637096497e-06,
+      "loss": 2.1116,
+      "step": 15300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.430374156727136e-06,
+      "loss": 2.0415,
+      "step": 15301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.428782946977996e-06,
+      "loss": 2.0535,
+      "step": 15302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.427192007890797e-06,
+      "loss": 2.1221,
+      "step": 15303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.425601339507261e-06,
+      "loss": 2.1143,
+      "step": 15304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.424010941869103e-06,
+      "loss": 1.9836,
+      "step": 15305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.42242081501803e-06,
+      "loss": 1.9951,
+      "step": 15306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.420830958995739e-06,
+      "loss": 2.0217,
+      "step": 15307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.419241373843923e-06,
+      "loss": 2.042,
+      "step": 15308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.417652059604266e-06,
+      "loss": 2.1104,
+      "step": 15309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.416063016318448e-06,
+      "loss": 2.114,
+      "step": 15310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.414474244028138e-06,
+      "loss": 2.1194,
+      "step": 15311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.4128857427750026e-06,
+      "loss": 2.071,
+      "step": 15312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.411297512600701e-06,
+      "loss": 2.0527,
+      "step": 15313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.409709553546867e-06,
+      "loss": 2.0474,
+      "step": 15314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.408121865655163e-06,
+      "loss": 2.0933,
+      "step": 15315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.406534448967208e-06,
+      "loss": 2.0732,
+      "step": 15316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.4049473035246445e-06,
+      "loss": 2.0791,
+      "step": 15317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.403360429369084e-06,
+      "loss": 2.0703,
+      "step": 15318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.40177382654214e-06,
+      "loss": 2.0483,
+      "step": 15319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.4001874950854254e-06,
+      "loss": 2.0403,
+      "step": 15320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.398601435040535e-06,
+      "loss": 2.1001,
+      "step": 15321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.397015646449065e-06,
+      "loss": 2.0938,
+      "step": 15322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.395430129352602e-06,
+      "loss": 2.0784,
+      "step": 15323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.393844883792714e-06,
+      "loss": 2.072,
+      "step": 15324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3922599098109826e-06,
+      "loss": 1.9932,
+      "step": 15325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.390675207448968e-06,
+      "loss": 2.0256,
+      "step": 15326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.389090776748229e-06,
+      "loss": 2.0703,
+      "step": 15327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.387506617750321e-06,
+      "loss": 2.0759,
+      "step": 15328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.385922730496769e-06,
+      "loss": 2.0205,
+      "step": 15329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3843391150291294e-06,
+      "loss": 1.9946,
+      "step": 15330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.382755771388914e-06,
+      "loss": 2.0491,
+      "step": 15331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.381172699617657e-06,
+      "loss": 2.0593,
+      "step": 15332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3795898997568595e-06,
+      "loss": 2.0413,
+      "step": 15333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.378007371848044e-06,
+      "loss": 2.0396,
+      "step": 15334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.376425115932701e-06,
+      "loss": 2.0061,
+      "step": 15335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.374843132052321e-06,
+      "loss": 2.0208,
+      "step": 15336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.373261420248397e-06,
+      "loss": 2.0859,
+      "step": 15337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3716799805624e-06,
+      "loss": 2.0762,
+      "step": 15338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.37009881303581e-06,
+      "loss": 2.0457,
+      "step": 15339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.368517917710089e-06,
+      "loss": 2.0559,
+      "step": 15340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.366937294626682e-06,
+      "loss": 2.0552,
+      "step": 15341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.365356943827058e-06,
+      "loss": 2.0327,
+      "step": 15342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.363776865352643e-06,
+      "loss": 1.9934,
+      "step": 15343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.36219705924489e-06,
+      "loss": 2.0261,
+      "step": 15344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.360617525545213e-06,
+      "loss": 2.0857,
+      "step": 15345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.359038264295038e-06,
+      "loss": 2.0623,
+      "step": 15346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.35745927553578e-06,
+      "loss": 2.0925,
+      "step": 15347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.355880559308845e-06,
+      "loss": 2.0715,
+      "step": 15348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.354302115655636e-06,
+      "loss": 2.092,
+      "step": 15349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.352723944617542e-06,
+      "loss": 2.092,
+      "step": 15350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.351146046235952e-06,
+      "loss": 2.0142,
+      "step": 15351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.349568420552242e-06,
+      "loss": 2.0505,
+      "step": 15352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.347991067607786e-06,
+      "loss": 2.1067,
+      "step": 15353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3464139874439446e-06,
+      "loss": 2.1274,
+      "step": 15354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.344837180102083e-06,
+      "loss": 2.0513,
+      "step": 15355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.343260645623535e-06,
+      "loss": 2.0679,
+      "step": 15356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.341684384049665e-06,
+      "loss": 2.1111,
+      "step": 15357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.340108395421787e-06,
+      "loss": 2.1021,
+      "step": 15358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3385326797812476e-06,
+      "loss": 1.9736,
+      "step": 15359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3369572371693554e-06,
+      "loss": 2.0156,
+      "step": 15360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.335382067627431e-06,
+      "loss": 2.1038,
+      "step": 15361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.333807171196779e-06,
+      "loss": 2.1116,
+      "step": 15362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3322325479187e-06,
+      "loss": 2.043,
+      "step": 15363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.330658197834488e-06,
+      "loss": 2.0857,
+      "step": 15364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.329084120985425e-06,
+      "loss": 2.0522,
+      "step": 15365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.327510317412792e-06,
+      "loss": 2.0562,
+      "step": 15366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3259367871578605e-06,
+      "loss": 2.0764,
+      "step": 15367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.324363530261893e-06,
+      "loss": 2.0745,
+      "step": 15368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.322790546766147e-06,
+      "loss": 2.0974,
+      "step": 15369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.321217836711876e-06,
+      "loss": 2.072,
+      "step": 15370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.31964540014031e-06,
+      "loss": 2.0579,
+      "step": 15371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.318073237092701e-06,
+      "loss": 2.0559,
+      "step": 15372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.316501347610262e-06,
+      "loss": 2.0144,
+      "step": 15373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.31492973173423e-06,
+      "loss": 2.0261,
+      "step": 15374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.313358389505805e-06,
+      "loss": 2.0298,
+      "step": 15375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.311787320966199e-06,
+      "loss": 2.0325,
+      "step": 15376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.310216526156609e-06,
+      "loss": 2.0986,
+      "step": 15377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3086460051182325e-06,
+      "loss": 2.1279,
+      "step": 15378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.30707575789225e-06,
+      "loss": 2.073,
+      "step": 15379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.305505784519839e-06,
+      "loss": 2.1042,
+      "step": 15380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.303936085042174e-06,
+      "loss": 2.1023,
+      "step": 15381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.302366659500416e-06,
+      "loss": 2.1206,
+      "step": 15382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.300797507935721e-06,
+      "loss": 2.0767,
+      "step": 15383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.299228630389242e-06,
+      "loss": 2.1238,
+      "step": 15384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2976600269021195e-06,
+      "loss": 2.012,
+      "step": 15385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.296091697515479e-06,
+      "loss": 2.01,
+      "step": 15386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.294523642270464e-06,
+      "loss": 2.0786,
+      "step": 15387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.29295586120818e-06,
+      "loss": 2.0557,
+      "step": 15388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.291388354369754e-06,
+      "loss": 2.1047,
+      "step": 15389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2898211217962826e-06,
+      "loss": 2.1094,
+      "step": 15390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.288254163528866e-06,
+      "loss": 2.0784,
+      "step": 15391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.286687479608599e-06,
+      "loss": 2.0681,
+      "step": 15392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.285121070076562e-06,
+      "loss": 2.123,
+      "step": 15393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.283554934973836e-06,
+      "loss": 2.1096,
+      "step": 15394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2819890743414885e-06,
+      "loss": 2.0715,
+      "step": 15395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.280423488220584e-06,
+      "loss": 2.0713,
+      "step": 15396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2788581766521776e-06,
+      "loss": 2.0266,
+      "step": 15397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2772931396773156e-06,
+      "loss": 2.021,
+      "step": 15398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2757283773370455e-06,
+      "loss": 2.033,
+      "step": 15399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2741638896723985e-06,
+      "loss": 2.0608,
+      "step": 15400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.272599676724393e-06,
+      "loss": 2.1052,
+      "step": 15401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.271035738534063e-06,
+      "loss": 2.0669,
+      "step": 15402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.269472075142408e-06,
+      "loss": 2.0339,
+      "step": 15403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2679086865904476e-06,
+      "loss": 2.0334,
+      "step": 15404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.266345572919165e-06,
+      "loss": 2.0916,
+      "step": 15405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.26478273416956e-06,
+      "loss": 2.0811,
+      "step": 15406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.263220170382611e-06,
+      "loss": 2.0068,
+      "step": 15407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2616578815992995e-06,
+      "loss": 2.0286,
+      "step": 15408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.260095867860591e-06,
+      "loss": 2.0544,
+      "step": 15409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.258534129207454e-06,
+      "loss": 2.0481,
+      "step": 15410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.256972665680829e-06,
+      "loss": 2.1177,
+      "step": 15411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.255411477321682e-06,
+      "loss": 2.0637,
+      "step": 15412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.253850564170935e-06,
+      "loss": 2.0684,
+      "step": 15413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.25228992626954e-06,
+      "loss": 2.0618,
+      "step": 15414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.250729563658407e-06,
+      "loss": 2.011,
+      "step": 15415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.249169476378462e-06,
+      "loss": 2.0127,
+      "step": 15416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.247609664470615e-06,
+      "loss": 2.0923,
+      "step": 15417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.246050127975771e-06,
+      "loss": 2.1023,
+      "step": 15418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2444908669348246e-06,
+      "loss": 2.0564,
+      "step": 15419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2429318813886705e-06,
+      "loss": 2.0454,
+      "step": 15420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.241373171378188e-06,
+      "loss": 2.0859,
+      "step": 15421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2398147369442515e-06,
+      "loss": 2.0725,
+      "step": 15422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.23825657812773e-06,
+      "loss": 2.1035,
+      "step": 15423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.236698694969486e-06,
+      "loss": 2.0835,
+      "step": 15424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.235141087510376e-06,
+      "loss": 2.0718,
+      "step": 15425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.233583755791236e-06,
+      "loss": 2.0984,
+      "step": 15426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.232026699852917e-06,
+      "loss": 2.0366,
+      "step": 15427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2304699197362385e-06,
+      "loss": 2.0559,
+      "step": 15428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.228913415482043e-06,
+      "loss": 2.0659,
+      "step": 15429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.22735718713113e-06,
+      "loss": 2.0652,
+      "step": 15430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.225801234724319e-06,
+      "loss": 2.0229,
+      "step": 15431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.224245558302411e-06,
+      "loss": 2.0535,
+      "step": 15432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.222690157906202e-06,
+      "loss": 2.0979,
+      "step": 15433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2211350335764815e-06,
+      "loss": 2.1089,
+      "step": 15434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.219580185354031e-06,
+      "loss": 2.0571,
+      "step": 15435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.218025613279622e-06,
+      "loss": 2.0576,
+      "step": 15436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.216471317394023e-06,
+      "loss": 2.1331,
+      "step": 15437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.214917297737993e-06,
+      "loss": 2.0955,
+      "step": 15438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2133635543522866e-06,
+      "loss": 2.0618,
+      "step": 15439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.211810087277653e-06,
+      "loss": 2.0671,
+      "step": 15440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.210256896554811e-06,
+      "loss": 2.0266,
+      "step": 15441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2087039822245154e-06,
+      "loss": 2.0732,
+      "step": 15442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2071513443274735e-06,
+      "loss": 2.1008,
+      "step": 15443
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.205598982904412e-06,
+      "loss": 2.085,
+      "step": 15444
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.20404689799603e-06,
+      "loss": 2.0947,
+      "step": 15445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.202495089643034e-06,
+      "loss": 2.0835,
+      "step": 15446
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.200943557886119e-06,
+      "loss": 2.0908,
+      "step": 15447
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.199392302765969e-06,
+      "loss": 2.0598,
+      "step": 15448
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.197841324323268e-06,
+      "loss": 2.0261,
+      "step": 15449
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.196290622598685e-06,
+      "loss": 2.0366,
+      "step": 15450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.194740197632888e-06,
+      "loss": 2.0073,
+      "step": 15451
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.193190049466533e-06,
+      "loss": 1.991,
+      "step": 15452
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.191640178140273e-06,
+      "loss": 2.0608,
+      "step": 15453
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.19009058369475e-06,
+      "loss": 2.0715,
+      "step": 15454
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.188541266170601e-06,
+      "loss": 2.0396,
+      "step": 15455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.186992225608455e-06,
+      "loss": 2.1074,
+      "step": 15456
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.185443462048936e-06,
+      "loss": 2.0278,
+      "step": 15457
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.18389497553265e-06,
+      "loss": 2.0098,
+      "step": 15458
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.182346766100218e-06,
+      "loss": 2.0671,
+      "step": 15459
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.180798833792222e-06,
+      "loss": 2.0774,
+      "step": 15460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1792511786492756e-06,
+      "loss": 2.014,
+      "step": 15461
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.177703800711951e-06,
+      "loss": 2.0244,
+      "step": 15462
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.176156700020827e-06,
+      "loss": 2.0483,
+      "step": 15463
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.174609876616477e-06,
+      "loss": 2.0519,
+      "step": 15464
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1730633305394645e-06,
+      "loss": 2.0811,
+      "step": 15465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.171517061830345e-06,
+      "loss": 2.1011,
+      "step": 15466
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.169971070529669e-06,
+      "loss": 1.9741,
+      "step": 15467
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1684253566779775e-06,
+      "loss": 1.9919,
+      "step": 15468
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.166879920315805e-06,
+      "loss": 2.0979,
+      "step": 15469
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.165334761483679e-06,
+      "loss": 2.1106,
+      "step": 15470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.163789880222121e-06,
+      "loss": 2.0417,
+      "step": 15471
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1622452765716446e-06,
+      "loss": 2.0654,
+      "step": 15472
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.160700950572745e-06,
+      "loss": 2.1345,
+      "step": 15473
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.159156902265936e-06,
+      "loss": 2.1416,
+      "step": 15474
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.157613131691693e-06,
+      "loss": 2.052,
+      "step": 15475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.156069638890516e-06,
+      "loss": 2.0962,
+      "step": 15476
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.154526423902866e-06,
+      "loss": 2.0825,
+      "step": 15477
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.152983486769221e-06,
+      "loss": 2.062,
+      "step": 15478
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1514408275300405e-06,
+      "loss": 2.1062,
+      "step": 15479
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.149898446225778e-06,
+      "loss": 2.0645,
+      "step": 15480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.148356342896879e-06,
+      "loss": 2.0115,
+      "step": 15481
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.146814517583793e-06,
+      "loss": 2.0442,
+      "step": 15482
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.145272970326937e-06,
+      "loss": 2.0872,
+      "step": 15483
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.143731701166753e-06,
+      "loss": 2.1006,
+      "step": 15484
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.142190710143641e-06,
+      "loss": 2.0781,
+      "step": 15485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.140649997298029e-06,
+      "loss": 2.1199,
+      "step": 15486
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.139109562670312e-06,
+      "loss": 2.0402,
+      "step": 15487
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.137569406300884e-06,
+      "loss": 2.0669,
+      "step": 15488
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.136029528230136e-06,
+      "loss": 2.015,
+      "step": 15489
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.13448992849845e-06,
+      "loss": 2.0554,
+      "step": 15490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.132950607146204e-06,
+      "loss": 2.0393,
+      "step": 15491
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.13141156421376e-06,
+      "loss": 2.0466,
+      "step": 15492
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.12987279974148e-06,
+      "loss": 1.981,
+      "step": 15493
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.128334313769712e-06,
+      "loss": 2.0056,
+      "step": 15494
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.126796106338807e-06,
+      "loss": 2.0713,
+      "step": 15495
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.125258177489098e-06,
+      "loss": 2.072,
+      "step": 15496
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.123720527260923e-06,
+      "loss": 2.0552,
+      "step": 15497
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.122183155694591e-06,
+      "loss": 2.0479,
+      "step": 15498
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.120646062830435e-06,
+      "loss": 2.1118,
+      "step": 15499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.119109248708745e-06,
+      "loss": 2.1138,
+      "step": 15500
+    },
+    {
+      "Bleu_1": 0.773929934710205,
+      "Bleu_2": 0.6133698683210252,
+      "Bleu_3": 0.47152786174665995,
+      "Bleu_4": 0.3583679511881235,
+      "CIDEr": 1.2049194985332106,
+      "METEOR": 0.27813400256498344,
+      "ROUGE_L": 0.5590870114403769,
+      "epoch": 0.78,
+      "step": 15500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.117572713369841e-06,
+      "loss": 2.0967,
+      "step": 15501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.116036456854002e-06,
+      "loss": 2.1016,
+      "step": 15502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.114500479201523e-06,
+      "loss": 1.9844,
+      "step": 15503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.112964780452678e-06,
+      "loss": 1.9849,
+      "step": 15504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.111429360647744e-06,
+      "loss": 2.0786,
+      "step": 15505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.1098942198269814e-06,
+      "loss": 2.0864,
+      "step": 15506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.10835935803065e-06,
+      "loss": 2.1301,
+      "step": 15507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.106824775298998e-06,
+      "loss": 2.1213,
+      "step": 15508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.10529047167227e-06,
+      "loss": 2.0837,
+      "step": 15509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.103756447190699e-06,
+      "loss": 2.106,
+      "step": 15510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.102222701894515e-06,
+      "loss": 2.0417,
+      "step": 15511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.100689235823942e-06,
+      "loss": 2.0007,
+      "step": 15512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.099156049019183e-06,
+      "loss": 2.0803,
+      "step": 15513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.097623141520455e-06,
+      "loss": 2.0818,
+      "step": 15514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.096090513367946e-06,
+      "loss": 2.021,
+      "step": 15515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0945581646018616e-06,
+      "loss": 2.0537,
+      "step": 15516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.093026095262375e-06,
+      "loss": 2.0271,
+      "step": 15517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.091494305389666e-06,
+      "loss": 2.0474,
+      "step": 15518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.089962795023901e-06,
+      "loss": 2.0894,
+      "step": 15519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.088431564205246e-06,
+      "loss": 2.0891,
+      "step": 15520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0869006129738554e-06,
+      "loss": 2.0898,
+      "step": 15521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.085369941369874e-06,
+      "loss": 2.0642,
+      "step": 15522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.083839549433445e-06,
+      "loss": 1.9473,
+      "step": 15523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0823094372047e-06,
+      "loss": 1.9922,
+      "step": 15524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.080779604723764e-06,
+      "loss": 2.0747,
+      "step": 15525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.079250052030755e-06,
+      "loss": 2.0713,
+      "step": 15526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0777207791657895e-06,
+      "loss": 2.0713,
+      "step": 15527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.076191786168956e-06,
+      "loss": 2.074,
+      "step": 15528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.074663073080368e-06,
+      "loss": 2.0325,
+      "step": 15529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0731346399401e-06,
+      "loss": 2.0343,
+      "step": 15530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.071606486788247e-06,
+      "loss": 2.0249,
+      "step": 15531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.070078613664871e-06,
+      "loss": 2.0215,
+      "step": 15532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.068551020610043e-06,
+      "loss": 2.0696,
+      "step": 15533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.067023707663824e-06,
+      "loss": 2.0664,
+      "step": 15534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.065496674866264e-06,
+      "loss": 2.0708,
+      "step": 15535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.063969922257408e-06,
+      "loss": 2.0559,
+      "step": 15536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0624434498772944e-06,
+      "loss": 2.0688,
+      "step": 15537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.060917257765953e-06,
+      "loss": 2.0657,
+      "step": 15538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0593913459634064e-06,
+      "loss": 2.0894,
+      "step": 15539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0578657145096684e-06,
+      "loss": 2.0867,
+      "step": 15540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.056340363444747e-06,
+      "loss": 2.1013,
+      "step": 15541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.054815292808649e-06,
+      "loss": 2.0698,
+      "step": 15542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.053290502641354e-06,
+      "loss": 2.0417,
+      "step": 15543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.051765992982865e-06,
+      "loss": 2.0508,
+      "step": 15544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.050241763873143e-06,
+      "loss": 2.0581,
+      "step": 15545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0487178153521755e-06,
+      "loss": 2.0706,
+      "step": 15546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.047194147459914e-06,
+      "loss": 2.0229,
+      "step": 15547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.045670760236322e-06,
+      "loss": 2.0039,
+      "step": 15548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.044147653721346e-06,
+      "loss": 2.0918,
+      "step": 15549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.042624827954924e-06,
+      "loss": 2.0908,
+      "step": 15550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.041102282976996e-06,
+      "loss": 2.0645,
+      "step": 15551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.039580018827492e-06,
+      "loss": 2.0574,
+      "step": 15552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.038058035546319e-06,
+      "loss": 2.0676,
+      "step": 15553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.036536333173403e-06,
+      "loss": 2.0859,
+      "step": 15554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.035014911748636e-06,
+      "loss": 2.0491,
+      "step": 15555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.03349377131193e-06,
+      "loss": 2.0752,
+      "step": 15556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.031972911903165e-06,
+      "loss": 2.0674,
+      "step": 15557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.030452333562224e-06,
+      "loss": 2.0503,
+      "step": 15558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.028932036328987e-06,
+      "loss": 2.012,
+      "step": 15559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.027412020243315e-06,
+      "loss": 1.9927,
+      "step": 15560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.025892285345079e-06,
+      "loss": 2.0342,
+      "step": 15561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.024372831674123e-06,
+      "loss": 2.0449,
+      "step": 15562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.022853659270299e-06,
+      "loss": 2.0852,
+      "step": 15563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.021334768173442e-06,
+      "loss": 2.092,
+      "step": 15564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0198161584233845e-06,
+      "loss": 2.0793,
+      "step": 15565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0182978300599494e-06,
+      "loss": 2.1165,
+      "step": 15566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.01677978312296e-06,
+      "loss": 2.0071,
+      "step": 15567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0152620176522076e-06,
+      "loss": 2.0217,
+      "step": 15568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.013744533687515e-06,
+      "loss": 2.0208,
+      "step": 15569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.012227331268657e-06,
+      "loss": 2.0791,
+      "step": 15570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0107104104354414e-06,
+      "loss": 2.0474,
+      "step": 15571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.00919377122763e-06,
+      "loss": 2.072,
+      "step": 15572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.007677413685003e-06,
+      "loss": 2.1318,
+      "step": 15573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.00616133784732e-06,
+      "loss": 2.1223,
+      "step": 15574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.004645543754347e-06,
+      "loss": 2.0854,
+      "step": 15575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.0031300314458246e-06,
+      "loss": 2.1118,
+      "step": 15576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.001614800961503e-06,
+      "loss": 2.0657,
+      "step": 15577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.000099852341109e-06,
+      "loss": 2.0862,
+      "step": 15578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.998585185624378e-06,
+      "loss": 2.0369,
+      "step": 15579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.99707080085103e-06,
+      "loss": 2.0674,
+      "step": 15580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.995556698060774e-06,
+      "loss": 2.0276,
+      "step": 15581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.994042877293317e-06,
+      "loss": 2.075,
+      "step": 15582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.992529338588355e-06,
+      "loss": 2.0798,
+      "step": 15583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.991016081985588e-06,
+      "loss": 2.1152,
+      "step": 15584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.989503107524684e-06,
+      "loss": 2.0535,
+      "step": 15585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.987990415245337e-06,
+      "loss": 2.0476,
+      "step": 15586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.986478005187196e-06,
+      "loss": 2.0596,
+      "step": 15587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.984965877389941e-06,
+      "loss": 2.0686,
+      "step": 15588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.983454031893213e-06,
+      "loss": 2.0645,
+      "step": 15589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9819424687366646e-06,
+      "loss": 2.083,
+      "step": 15590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.98043118795993e-06,
+      "loss": 2.0649,
+      "step": 15591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9789201896026455e-06,
+      "loss": 2.0466,
+      "step": 15592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.977409473704433e-06,
+      "loss": 2.0576,
+      "step": 15593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.975899040304909e-06,
+      "loss": 2.0608,
+      "step": 15594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.974388889443683e-06,
+      "loss": 2.0405,
+      "step": 15595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.972879021160357e-06,
+      "loss": 2.051,
+      "step": 15596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.971369435494528e-06,
+      "loss": 2.0515,
+      "step": 15597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9698601324857804e-06,
+      "loss": 2.0559,
+      "step": 15598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.968351112173698e-06,
+      "loss": 2.0242,
+      "step": 15599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.966842374597842e-06,
+      "loss": 2.0134,
+      "step": 15600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.965333919797794e-06,
+      "loss": 2.0703,
+      "step": 15601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.963825747813095e-06,
+      "loss": 2.0896,
+      "step": 15602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.962317858683309e-06,
+      "loss": 2.0425,
+      "step": 15603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.960810252447967e-06,
+      "loss": 2.0115,
+      "step": 15604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.959302929146612e-06,
+      "loss": 2.0417,
+      "step": 15605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.957795888818769e-06,
+      "loss": 2.0464,
+      "step": 15606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9562891315039574e-06,
+      "loss": 2.0605,
+      "step": 15607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.954782657241691e-06,
+      "loss": 2.0571,
+      "step": 15608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9532764660714765e-06,
+      "loss": 2.0496,
+      "step": 15609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.951770558032812e-06,
+      "loss": 2.0349,
+      "step": 15610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.950264933165187e-06,
+      "loss": 2.0586,
+      "step": 15611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.948759591508084e-06,
+      "loss": 2.092,
+      "step": 15612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.947254533100981e-06,
+      "loss": 2.1008,
+      "step": 15613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.94574975798335e-06,
+      "loss": 2.092,
+      "step": 15614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.944245266194639e-06,
+      "loss": 2.0442,
+      "step": 15615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.942741057774319e-06,
+      "loss": 2.0674,
+      "step": 15616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.941237132761817e-06,
+      "loss": 2.05,
+      "step": 15617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.939733491196591e-06,
+      "loss": 2.0696,
+      "step": 15618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.938230133118058e-06,
+      "loss": 2.0269,
+      "step": 15619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.936727058565649e-06,
+      "loss": 1.9817,
+      "step": 15620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.935224267578776e-06,
+      "loss": 2.0195,
+      "step": 15621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.933721760196852e-06,
+      "loss": 1.9954,
+      "step": 15622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.932219536459276e-06,
+      "loss": 2.1038,
+      "step": 15623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.930717596405446e-06,
+      "loss": 2.1272,
+      "step": 15624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.929215940074738e-06,
+      "loss": 2.0715,
+      "step": 15625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.927714567506548e-06,
+      "loss": 2.1072,
+      "step": 15626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9262134787402275e-06,
+      "loss": 2.0566,
+      "step": 15627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.924712673815161e-06,
+      "loss": 2.0737,
+      "step": 15628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.923212152770694e-06,
+      "loss": 2.0706,
+      "step": 15629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.921711915646172e-06,
+      "loss": 2.0969,
+      "step": 15630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.920211962480952e-06,
+      "loss": 2.1018,
+      "step": 15631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.918712293314349e-06,
+      "loss": 2.0964,
+      "step": 15632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.917212908185711e-06,
+      "loss": 2.1084,
+      "step": 15633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.915713807134344e-06,
+      "loss": 2.1072,
+      "step": 15634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.91421499019956e-06,
+      "loss": 2.1052,
+      "step": 15635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.912716457420672e-06,
+      "loss": 2.0796,
+      "step": 15636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.911218208836968e-06,
+      "loss": 2.0825,
+      "step": 15637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9097202444877435e-06,
+      "loss": 2.1147,
+      "step": 15638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.908222564412284e-06,
+      "loss": 2.1155,
+      "step": 15639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.906725168649854e-06,
+      "loss": 2.1172,
+      "step": 15640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.90522805723973e-06,
+      "loss": 2.0876,
+      "step": 15641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.903731230221166e-06,
+      "loss": 2.1038,
+      "step": 15642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9022346876334235e-06,
+      "loss": 2.0515,
+      "step": 15643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.900738429515738e-06,
+      "loss": 2.0383,
+      "step": 15644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.899242455907351e-06,
+      "loss": 2.0234,
+      "step": 15645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.897746766847491e-06,
+      "loss": 2.0212,
+      "step": 15646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8962513623753825e-06,
+      "loss": 2.1257,
+      "step": 15647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.894756242530243e-06,
+      "loss": 2.1345,
+      "step": 15648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.893261407351276e-06,
+      "loss": 2.0959,
+      "step": 15649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.891766856877683e-06,
+      "loss": 2.113,
+      "step": 15650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8902725911486575e-06,
+      "loss": 2.0305,
+      "step": 15651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.888778610203386e-06,
+      "loss": 2.0156,
+      "step": 15652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.887284914081044e-06,
+      "loss": 2.104,
+      "step": 15653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8857915028208086e-06,
+      "loss": 2.0781,
+      "step": 15654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.884298376461828e-06,
+      "loss": 2.0879,
+      "step": 15655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.882805535043276e-06,
+      "loss": 2.073,
+      "step": 15656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.881312978604282e-06,
+      "loss": 2.0825,
+      "step": 15657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.879820707184005e-06,
+      "loss": 2.0796,
+      "step": 15658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8783287208215656e-06,
+      "loss": 2.0618,
+      "step": 15659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.876837019556093e-06,
+      "loss": 2.0603,
+      "step": 15660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.875345603426705e-06,
+      "loss": 2.1047,
+      "step": 15661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.873854472472514e-06,
+      "loss": 2.1001,
+      "step": 15662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.872363626732619e-06,
+      "loss": 2.0955,
+      "step": 15663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.870873066246122e-06,
+      "loss": 2.1069,
+      "step": 15664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.869382791052108e-06,
+      "loss": 2.0679,
+      "step": 15665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.867892801189653e-06,
+      "loss": 2.0676,
+      "step": 15666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.866403096697838e-06,
+      "loss": 2.0442,
+      "step": 15667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.864913677615726e-06,
+      "loss": 2.0303,
+      "step": 15668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.86342454398238e-06,
+      "loss": 2.0686,
+      "step": 15669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.861935695836835e-06,
+      "loss": 2.1016,
+      "step": 15670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.860447133218154e-06,
+      "loss": 1.9817,
+      "step": 15671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8589588561653555e-06,
+      "loss": 2.0125,
+      "step": 15672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.857470864717485e-06,
+      "loss": 2.0806,
+      "step": 15673
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8559831589135495e-06,
+      "loss": 2.0588,
+      "step": 15674
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.854495738792568e-06,
+      "loss": 2.1111,
+      "step": 15675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8530086043935445e-06,
+      "loss": 2.0984,
+      "step": 15676
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.851521755755481e-06,
+      "loss": 2.0559,
+      "step": 15677
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8500351929173645e-06,
+      "loss": 2.0387,
+      "step": 15678
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8485489159181786e-06,
+      "loss": 2.0063,
+      "step": 15679
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.847062924796901e-06,
+      "loss": 2.0022,
+      "step": 15680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8455772195925015e-06,
+      "loss": 2.0352,
+      "step": 15681
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.844091800343936e-06,
+      "loss": 2.0518,
+      "step": 15682
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8426066670901656e-06,
+      "loss": 2.0173,
+      "step": 15683
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.841121819870131e-06,
+      "loss": 2.04,
+      "step": 15684
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.839637258722766e-06,
+      "loss": 2.0332,
+      "step": 15685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.838152983687012e-06,
+      "loss": 2.0334,
+      "step": 15686
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.836668994801779e-06,
+      "loss": 2.0627,
+      "step": 15687
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8351852921060006e-06,
+      "loss": 2.0442,
+      "step": 15688
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.833701875638573e-06,
+      "loss": 2.0847,
+      "step": 15689
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.832218745438397e-06,
+      "loss": 2.0864,
+      "step": 15690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.830735901544369e-06,
+      "loss": 2.051,
+      "step": 15691
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.829253343995375e-06,
+      "loss": 2.04,
+      "step": 15692
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.827771072830292e-06,
+      "loss": 2.0864,
+      "step": 15693
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.826289088087996e-06,
+      "loss": 2.0798,
+      "step": 15694
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.824807389807338e-06,
+      "loss": 2.1138,
+      "step": 15695
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.82332597802719e-06,
+      "loss": 2.0811,
+      "step": 15696
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.821844852786384e-06,
+      "loss": 2.0933,
+      "step": 15697
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8203640141237724e-06,
+      "loss": 2.085,
+      "step": 15698
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.818883462078185e-06,
+      "loss": 2.0598,
+      "step": 15699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.817403196688447e-06,
+      "loss": 2.0505,
+      "step": 15700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.815923217993383e-06,
+      "loss": 2.0388,
+      "step": 15701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.814443526031789e-06,
+      "loss": 2.051,
+      "step": 15702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.812964120842485e-06,
+      "loss": 2.061,
+      "step": 15703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.8114850024642495e-06,
+      "loss": 2.0576,
+      "step": 15704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.810006170935889e-06,
+      "loss": 2.0845,
+      "step": 15705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.808527626296169e-06,
+      "loss": 2.0762,
+      "step": 15706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.807049368583871e-06,
+      "loss": 2.0681,
+      "step": 15707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.805571397837758e-06,
+      "loss": 2.0645,
+      "step": 15708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.8040937140965865e-06,
+      "loss": 2.062,
+      "step": 15709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.802616317399112e-06,
+      "loss": 2.0706,
+      "step": 15710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.801139207784076e-06,
+      "loss": 2.0176,
+      "step": 15711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7996623852902056e-06,
+      "loss": 2.0186,
+      "step": 15712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7981858499562425e-06,
+      "loss": 2.0623,
+      "step": 15713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.796709601820891e-06,
+      "loss": 2.0591,
+      "step": 15714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.795233640922883e-06,
+      "loss": 1.9966,
+      "step": 15715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.79375796730091e-06,
+      "loss": 2.0203,
+      "step": 15716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.792282580993675e-06,
+      "loss": 2.0764,
+      "step": 15717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.790807482039865e-06,
+      "loss": 2.0928,
+      "step": 15718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7893326704781655e-06,
+      "loss": 2.0474,
+      "step": 15719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.787858146347251e-06,
+      "loss": 2.0674,
+      "step": 15720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.786383909685789e-06,
+      "loss": 2.0339,
+      "step": 15721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.784909960532439e-06,
+      "loss": 2.051,
+      "step": 15722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.783436298925857e-06,
+      "loss": 2.115,
+      "step": 15723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7819629249046844e-06,
+      "loss": 2.1147,
+      "step": 15724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.780489838507559e-06,
+      "loss": 2.1169,
+      "step": 15725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.779017039773116e-06,
+      "loss": 2.0847,
+      "step": 15726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.777544528739967e-06,
+      "loss": 2.1221,
+      "step": 15727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.77607230544674e-06,
+      "loss": 2.1042,
+      "step": 15728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.774600369932029e-06,
+      "loss": 2.0444,
+      "step": 15729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.773128722234448e-06,
+      "loss": 2.0684,
+      "step": 15730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7716573623925774e-06,
+      "loss": 2.082,
+      "step": 15731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.770186290445006e-06,
+      "loss": 2.1226,
+      "step": 15732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.768715506430312e-06,
+      "loss": 2.052,
+      "step": 15733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7672450103870655e-06,
+      "loss": 2.0869,
+      "step": 15734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.765774802353825e-06,
+      "loss": 2.075,
+      "step": 15735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.76430488236915e-06,
+      "loss": 2.1055,
+      "step": 15736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.762835250471582e-06,
+      "loss": 2.0776,
+      "step": 15737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.761365906699665e-06,
+      "loss": 2.0703,
+      "step": 15738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.759896851091927e-06,
+      "loss": 2.0959,
+      "step": 15739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.758428083686896e-06,
+      "loss": 2.0933,
+      "step": 15740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.756959604523092e-06,
+      "loss": 2.0469,
+      "step": 15741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.755491413639011e-06,
+      "loss": 2.0718,
+      "step": 15742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7540235110731675e-06,
+      "loss": 2.0769,
+      "step": 15743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7525558968640474e-06,
+      "loss": 2.0833,
+      "step": 15744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7510885710501455e-06,
+      "loss": 2.0015,
+      "step": 15745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.749621533669933e-06,
+      "loss": 2.0115,
+      "step": 15746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7481547847618824e-06,
+      "loss": 2.0564,
+      "step": 15747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7466883243644616e-06,
+      "loss": 2.0579,
+      "step": 15748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.745222152516123e-06,
+      "loss": 2.0801,
+      "step": 15749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.743756269255317e-06,
+      "loss": 2.0752,
+      "step": 15750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.742290674620484e-06,
+      "loss": 2.0713,
+      "step": 15751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.740825368650059e-06,
+      "loss": 2.0607,
+      "step": 15752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7393603513824665e-06,
+      "loss": 2.051,
+      "step": 15753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7378956228561246e-06,
+      "loss": 2.043,
+      "step": 15754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.736431183109445e-06,
+      "loss": 2.0349,
+      "step": 15755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.734967032180833e-06,
+      "loss": 2.0515,
+      "step": 15756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.733503170108676e-06,
+      "loss": 2.0601,
+      "step": 15757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.732039596931376e-06,
+      "loss": 2.0735,
+      "step": 15758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.730576312687295e-06,
+      "loss": 2.0732,
+      "step": 15759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.729113317414827e-06,
+      "loss": 2.125,
+      "step": 15760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.727650611152322e-06,
+      "loss": 2.0796,
+      "step": 15761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.726188193938143e-06,
+      "loss": 2.1128,
+      "step": 15762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.724726065810639e-06,
+      "loss": 2.0312,
+      "step": 15763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.723264226808155e-06,
+      "loss": 2.0547,
+      "step": 15764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.721802676969021e-06,
+      "loss": 2.0537,
+      "step": 15765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.720341416331574e-06,
+      "loss": 2.0854,
+      "step": 15766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.718880444934121e-06,
+      "loss": 2.0952,
+      "step": 15767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.717419762814982e-06,
+      "loss": 2.0967,
+      "step": 15768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7159593700124624e-06,
+      "loss": 2.0295,
+      "step": 15769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7144992665648575e-06,
+      "loss": 2.0691,
+      "step": 15770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.713039452510461e-06,
+      "loss": 2.0435,
+      "step": 15771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.711579927887543e-06,
+      "loss": 2.01,
+      "step": 15772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.710120692734393e-06,
+      "loss": 2.1206,
+      "step": 15773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.708661747089262e-06,
+      "loss": 2.0957,
+      "step": 15774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.707203090990426e-06,
+      "loss": 2.0208,
+      "step": 15775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.705744724476124e-06,
+      "loss": 2.0452,
+      "step": 15776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.704286647584604e-06,
+      "loss": 2.054,
+      "step": 15777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.702828860354102e-06,
+      "loss": 2.0464,
+      "step": 15778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.7013713628228464e-06,
+      "loss": 2.0847,
+      "step": 15779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.699914155029062e-06,
+      "loss": 2.071,
+      "step": 15780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.698457237010959e-06,
+      "loss": 2.1169,
+      "step": 15781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.697000608806739e-06,
+      "loss": 2.1272,
+      "step": 15782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.695544270454613e-06,
+      "loss": 2.0867,
+      "step": 15783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.694088221992757e-06,
+      "loss": 2.0293,
+      "step": 15784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.692632463459369e-06,
+      "loss": 2.0479,
+      "step": 15785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.691176994892612e-06,
+      "loss": 2.0615,
+      "step": 15786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6897218163306585e-06,
+      "loss": 2.0764,
+      "step": 15787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.688266927811669e-06,
+      "loss": 2.0894,
+      "step": 15788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.686812329373797e-06,
+      "loss": 2.0459,
+      "step": 15789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.685358021055188e-06,
+      "loss": 2.0857,
+      "step": 15790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6839040028939764e-06,
+      "loss": 1.9739,
+      "step": 15791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.682450274928295e-06,
+      "loss": 1.9897,
+      "step": 15792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6809968371962686e-06,
+      "loss": 2.0425,
+      "step": 15793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.679543689736005e-06,
+      "loss": 2.0459,
+      "step": 15794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.678090832585617e-06,
+      "loss": 2.1152,
+      "step": 15795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.676638265783206e-06,
+      "loss": 2.1011,
+      "step": 15796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.675185989366852e-06,
+      "loss": 2.0952,
+      "step": 15797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.673734003374656e-06,
+      "loss": 2.071,
+      "step": 15798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.672282307844676e-06,
+      "loss": 1.999,
+      "step": 15799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.670830902815002e-06,
+      "loss": 2.0066,
+      "step": 15800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.669379788323679e-06,
+      "loss": 2.092,
+      "step": 15801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.667928964408765e-06,
+      "loss": 2.074,
+      "step": 15802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.666478431108308e-06,
+      "loss": 2.0757,
+      "step": 15803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.665028188460347e-06,
+      "loss": 2.1011,
+      "step": 15804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.663578236502912e-06,
+      "loss": 2.1208,
+      "step": 15805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.662128575274026e-06,
+      "loss": 2.1272,
+      "step": 15806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6606792048117045e-06,
+      "loss": 2.0732,
+      "step": 15807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.659230125153958e-06,
+      "loss": 2.0615,
+      "step": 15808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.657781336338781e-06,
+      "loss": 2.0767,
+      "step": 15809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.656332838404173e-06,
+      "loss": 2.0955,
+      "step": 15810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.65488463138812e-06,
+      "loss": 2.0386,
+      "step": 15811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.653436715328589e-06,
+      "loss": 2.0657,
+      "step": 15812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.651989090263562e-06,
+      "loss": 2.0598,
+      "step": 15813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.65054175623099e-06,
+      "loss": 2.0447,
+      "step": 15814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6490947132688405e-06,
+      "loss": 2.0715,
+      "step": 15815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.647647961415051e-06,
+      "loss": 2.072,
+      "step": 15816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.646201500707564e-06,
+      "loss": 2.0344,
+      "step": 15817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.644755331184311e-06,
+      "loss": 2.0349,
+      "step": 15818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.643309452883215e-06,
+      "loss": 2.0554,
+      "step": 15819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.641863865842196e-06,
+      "loss": 2.0527,
+      "step": 15820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.640418570099158e-06,
+      "loss": 2.0669,
+      "step": 15821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6389735656920044e-06,
+      "loss": 2.0754,
+      "step": 15822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.637528852658629e-06,
+      "loss": 2.0708,
+      "step": 15823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.636084431036919e-06,
+      "loss": 2.062,
+      "step": 15824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.634640300864752e-06,
+      "loss": 2.0879,
+      "step": 15825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.633196462179995e-06,
+      "loss": 2.0903,
+      "step": 15826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6317529150205175e-06,
+      "loss": 2.0837,
+      "step": 15827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.630309659424174e-06,
+      "loss": 2.0801,
+      "step": 15828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.628866695428802e-06,
+      "loss": 2.1521,
+      "step": 15829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.627424023072256e-06,
+      "loss": 2.1272,
+      "step": 15830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.625981642392353e-06,
+      "loss": 2.0471,
+      "step": 15831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.624539553426936e-06,
+      "loss": 2.0305,
+      "step": 15832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.623097756213808e-06,
+      "loss": 2.0073,
+      "step": 15833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.621656250790782e-06,
+      "loss": 2.0237,
+      "step": 15834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.620215037195661e-06,
+      "loss": 2.0527,
+      "step": 15835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.61877411546624e-06,
+      "loss": 2.0217,
+      "step": 15836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.617333485640304e-06,
+      "loss": 2.0945,
+      "step": 15837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.61589314775563e-06,
+      "loss": 2.0573,
+      "step": 15838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.614453101849995e-06,
+      "loss": 2.0671,
+      "step": 15839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.613013347961156e-06,
+      "loss": 2.0701,
+      "step": 15840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.611573886126873e-06,
+      "loss": 2.0906,
+      "step": 15841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.610134716384893e-06,
+      "loss": 2.0637,
+      "step": 15842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.608695838772961e-06,
+      "loss": 2.092,
+      "step": 15843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.607257253328795e-06,
+      "loss": 2.0889,
+      "step": 15844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.605818960090141e-06,
+      "loss": 2.0063,
+      "step": 15845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.604380959094698e-06,
+      "loss": 2.0076,
+      "step": 15846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.602943250380191e-06,
+      "loss": 2.0979,
+      "step": 15847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.6015058339843135e-06,
+      "loss": 2.0493,
+      "step": 15848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.600068709944759e-06,
+      "loss": 2.0471,
+      "step": 15849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.598631878299221e-06,
+      "loss": 2.0522,
+      "step": 15850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.597195339085372e-06,
+      "loss": 2.0457,
+      "step": 15851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.595759092340889e-06,
+      "loss": 2.0562,
+      "step": 15852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.5943231381034355e-06,
+      "loss": 2.1121,
+      "step": 15853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.592887476410659e-06,
+      "loss": 2.1089,
+      "step": 15854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.591452107300223e-06,
+      "loss": 2.1006,
+      "step": 15855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.590017030809752e-06,
+      "loss": 2.0781,
+      "step": 15856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.588582246976895e-06,
+      "loss": 2.0371,
+      "step": 15857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.587147755839268e-06,
+      "loss": 2.0413,
+      "step": 15858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.585713557434488e-06,
+      "loss": 2.0461,
+      "step": 15859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.58427965180017e-06,
+      "loss": 2.0681,
+      "step": 15860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.582846038973915e-06,
+      "loss": 2.051,
+      "step": 15861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.581412718993316e-06,
+      "loss": 2.0789,
+      "step": 15862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.579979691895964e-06,
+      "loss": 2.0151,
+      "step": 15863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.578546957719434e-06,
+      "loss": 2.0161,
+      "step": 15864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.577114516501299e-06,
+      "loss": 1.9905,
+      "step": 15865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.575682368279126e-06,
+      "loss": 2.0442,
+      "step": 15866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.574250513090468e-06,
+      "loss": 2.0857,
+      "step": 15867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.572818950972879e-06,
+      "loss": 2.0798,
+      "step": 15868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.571387681963891e-06,
+      "loss": 2.1292,
+      "step": 15869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.569956706101049e-06,
+      "loss": 2.1089,
+      "step": 15870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.568526023421863e-06,
+      "loss": 2.0991,
+      "step": 15871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.56709563396387e-06,
+      "loss": 2.1013,
+      "step": 15872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.565665537764564e-06,
+      "loss": 2.021,
+      "step": 15873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.564235734861457e-06,
+      "loss": 2.0415,
+      "step": 15874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.56280622529204e-06,
+      "loss": 2.0601,
+      "step": 15875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.561377009093801e-06,
+      "loss": 2.0276,
+      "step": 15876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.559948086304221e-06,
+      "loss": 2.083,
+      "step": 15877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.55851945696077e-06,
+      "loss": 2.1082,
+      "step": 15878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.557091121100912e-06,
+      "loss": 2.0801,
+      "step": 15879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.5556630787621056e-06,
+      "loss": 2.0938,
+      "step": 15880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.554235329981798e-06,
+      "loss": 2.075,
+      "step": 15881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.552807874797429e-06,
+      "loss": 2.0662,
+      "step": 15882
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.551380713246438e-06,
+      "loss": 2.0535,
+      "step": 15883
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.549953845366241e-06,
+      "loss": 2.0701,
+      "step": 15884
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.5485272711942646e-06,
+      "loss": 2.0452,
+      "step": 15885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.547100990767912e-06,
+      "loss": 2.0693,
+      "step": 15886
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.545675004124594e-06,
+      "loss": 1.9917,
+      "step": 15887
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.544249311301697e-06,
+      "loss": 1.998,
+      "step": 15888
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.542823912336614e-06,
+      "loss": 2.0942,
+      "step": 15889
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.54139880726672e-06,
+      "loss": 2.0549,
+      "step": 15890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.539973996129391e-06,
+      "loss": 2.04,
+      "step": 15891
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.538549478961987e-06,
+      "loss": 2.0197,
+      "step": 15892
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.537125255801867e-06,
+      "loss": 2.082,
+      "step": 15893
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.535701326686381e-06,
+      "loss": 2.0732,
+      "step": 15894
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.534277691652865e-06,
+      "loss": 2.0847,
+      "step": 15895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.532854350738657e-06,
+      "loss": 2.0505,
+      "step": 15896
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.531431303981079e-06,
+      "loss": 2.0803,
+      "step": 15897
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.530008551417455e-06,
+      "loss": 2.0657,
+      "step": 15898
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.528586093085082e-06,
+      "loss": 2.0515,
+      "step": 15899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.527163929021279e-06,
+      "loss": 2.0554,
+      "step": 15900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.525742059263323e-06,
+      "loss": 2.0364,
+      "step": 15901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.524320483848519e-06,
+      "loss": 2.0156,
+      "step": 15902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.522899202814136e-06,
+      "loss": 2.1287,
+      "step": 15903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.521478216197441e-06,
+      "loss": 2.0894,
+      "step": 15904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5200575240357065e-06,
+      "loss": 2.0417,
+      "step": 15905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.518637126366187e-06,
+      "loss": 2.0671,
+      "step": 15906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5172170232261275e-06,
+      "loss": 2.0439,
+      "step": 15907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.51579721465277e-06,
+      "loss": 2.0447,
+      "step": 15908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.514377700683347e-06,
+      "loss": 2.0896,
+      "step": 15909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.512958481355086e-06,
+      "loss": 2.0928,
+      "step": 15910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5115395567052005e-06,
+      "loss": 2.0898,
+      "step": 15911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.510120926770905e-06,
+      "loss": 2.1211,
+      "step": 15912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.508702591589399e-06,
+      "loss": 2.0652,
+      "step": 15913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.507284551197872e-06,
+      "loss": 2.0645,
+      "step": 15914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.505866805633521e-06,
+      "loss": 2.0598,
+      "step": 15915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.50444935493351e-06,
+      "loss": 2.0649,
+      "step": 15916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.503032199135027e-06,
+      "loss": 2.0659,
+      "step": 15917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5016153382752245e-06,
+      "loss": 2.042,
+      "step": 15918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.500198772391257e-06,
+      "loss": 2.0132,
+      "step": 15919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.498782501520278e-06,
+      "loss": 2.0078,
+      "step": 15920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4973665256994255e-06,
+      "loss": 2.0781,
+      "step": 15921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.495950844965832e-06,
+      "loss": 2.1143,
+      "step": 15922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.494535459356624e-06,
+      "loss": 2.0684,
+      "step": 15923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.493120368908908e-06,
+      "loss": 2.0649,
+      "step": 15924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4917055736598085e-06,
+      "loss": 2.0701,
+      "step": 15925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4902910736464125e-06,
+      "loss": 2.0618,
+      "step": 15926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.48887686890583e-06,
+      "loss": 2.0398,
+      "step": 15927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.487462959475132e-06,
+      "loss": 2.0776,
+      "step": 15928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4860493453914e-06,
+      "loss": 2.0872,
+      "step": 15929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.484636026691708e-06,
+      "loss": 2.1128,
+      "step": 15930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4832230034131185e-06,
+      "loss": 2.0884,
+      "step": 15931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.481810275592685e-06,
+      "loss": 2.0891,
+      "step": 15932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4803978432674534e-06,
+      "loss": 2.0908,
+      "step": 15933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.478985706474463e-06,
+      "loss": 2.095,
+      "step": 15934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.477573865250749e-06,
+      "loss": 2.1121,
+      "step": 15935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4761623196333314e-06,
+      "loss": 2.1067,
+      "step": 15936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.47475106965923e-06,
+      "loss": 2.0198,
+      "step": 15937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.473340115365455e-06,
+      "loss": 2.0068,
+      "step": 15938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.471929456788993e-06,
+      "loss": 2.1208,
+      "step": 15939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.470519093966856e-06,
+      "loss": 2.0933,
+      "step": 15940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.469109026936013e-06,
+      "loss": 2.0688,
+      "step": 15941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.467699255733458e-06,
+      "loss": 2.1018,
+      "step": 15942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.466289780396142e-06,
+      "loss": 2.0537,
+      "step": 15943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.464880600961044e-06,
+      "loss": 2.0642,
+      "step": 15944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.463471717465106e-06,
+      "loss": 2.0781,
+      "step": 15945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.462063129945278e-06,
+      "loss": 2.0659,
+      "step": 15946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.460654838438499e-06,
+      "loss": 2.0244,
+      "step": 15947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.459246842981702e-06,
+      "loss": 2.0491,
+      "step": 15948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.457839143611806e-06,
+      "loss": 2.0515,
+      "step": 15949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.456431740365729e-06,
+      "loss": 2.0754,
+      "step": 15950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.455024633280378e-06,
+      "loss": 2.0706,
+      "step": 15951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.453617822392654e-06,
+      "loss": 2.1008,
+      "step": 15952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.452211307739444e-06,
+      "loss": 2.1208,
+      "step": 15953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4508050893576384e-06,
+      "loss": 2.1326,
+      "step": 15954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4493991672841145e-06,
+      "loss": 2.0776,
+      "step": 15955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.447993541555729e-06,
+      "loss": 2.0457,
+      "step": 15956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.44658821220936e-06,
+      "loss": 2.0564,
+      "step": 15957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.445183179281843e-06,
+      "loss": 2.0918,
+      "step": 15958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4437784428100405e-06,
+      "loss": 2.0708,
+      "step": 15959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.442374002830777e-06,
+      "loss": 2.0662,
+      "step": 15960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.440969859380888e-06,
+      "loss": 1.998,
+      "step": 15961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4395660124971946e-06,
+      "loss": 2.051,
+      "step": 15962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.438162462216512e-06,
+      "loss": 2.0923,
+      "step": 15963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.436759208575645e-06,
+      "loss": 2.1045,
+      "step": 15964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.435356251611392e-06,
+      "loss": 2.0566,
+      "step": 15965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4339535913605465e-06,
+      "loss": 2.0588,
+      "step": 15966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.432551227859891e-06,
+      "loss": 2.0325,
+      "step": 15967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.431149161146198e-06,
+      "loss": 2.0188,
+      "step": 15968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.429747391256237e-06,
+      "loss": 2.0415,
+      "step": 15969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.428345918226775e-06,
+      "loss": 2.0415,
+      "step": 15970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.426944742094546e-06,
+      "loss": 2.0039,
+      "step": 15971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4255438628963145e-06,
+      "loss": 1.9895,
+      "step": 15972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.424143280668799e-06,
+      "loss": 2.0188,
+      "step": 15973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.422742995448744e-06,
+      "loss": 2.0352,
+      "step": 15974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.421343007272859e-06,
+      "loss": 2.0642,
+      "step": 15975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4199433161778615e-06,
+      "loss": 2.075,
+      "step": 15976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.418543922200456e-06,
+      "loss": 2.0725,
+      "step": 15977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.417144825377342e-06,
+      "loss": 2.0767,
+      "step": 15978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.415746025745206e-06,
+      "loss": 2.0996,
+      "step": 15979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.414347523340733e-06,
+      "loss": 2.0798,
+      "step": 15980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.412949318200593e-06,
+      "loss": 2.0876,
+      "step": 15981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.411551410361457e-06,
+      "loss": 2.0754,
+      "step": 15982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.41015379985998e-06,
+      "loss": 2.0342,
+      "step": 15983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.408756486732815e-06,
+      "loss": 2.0178,
+      "step": 15984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.407359471016607e-06,
+      "loss": 2.0537,
+      "step": 15985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.405962752747982e-06,
+      "loss": 2.0876,
+      "step": 15986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.40456633196358e-06,
+      "loss": 2.0601,
+      "step": 15987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.403170208700006e-06,
+      "loss": 2.0757,
+      "step": 15988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.401774382993887e-06,
+      "loss": 2.0825,
+      "step": 15989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.400378854881818e-06,
+      "loss": 2.0852,
+      "step": 15990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.398983624400394e-06,
+      "loss": 2.072,
+      "step": 15991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3975886915862076e-06,
+      "loss": 2.1086,
+      "step": 15992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.396194056475838e-06,
+      "loss": 2.0479,
+      "step": 15993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3947997191058564e-06,
+      "loss": 2.0637,
+      "step": 15994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.393405679512834e-06,
+      "loss": 2.0613,
+      "step": 15995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.392011937733315e-06,
+      "loss": 2.0439,
+      "step": 15996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.390618493803862e-06,
+      "loss": 2.0574,
+      "step": 15997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.389225347761003e-06,
+      "loss": 2.0493,
+      "step": 15998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.38783249964129e-06,
+      "loss": 2.0583,
+      "step": 15999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3864399494812325e-06,
+      "loss": 2.063,
+      "step": 16000
+    },
+    {
+      "Bleu_1": 0.7746688037438593,
+      "Bleu_2": 0.6139648497372396,
+      "Bleu_3": 0.4710759889538972,
+      "Bleu_4": 0.3569679655024206,
+      "CIDEr": 1.2039701943159902,
+      "METEOR": 0.27851607349722923,
+      "ROUGE_L": 0.5605462926508457,
+      "epoch": 0.8,
+      "step": 16000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.385047697317356e-06,
+      "loss": 2.042,
+      "step": 16001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.383655743186166e-06,
+      "loss": 2.0322,
+      "step": 16002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.382264087124166e-06,
+      "loss": 2.1104,
+      "step": 16003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.380872729167861e-06,
+      "loss": 2.1045,
+      "step": 16004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.379481669353723e-06,
+      "loss": 2.0356,
+      "step": 16005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.378090907718239e-06,
+      "loss": 2.0376,
+      "step": 16006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.376700444297877e-06,
+      "loss": 2.0684,
+      "step": 16007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.375310279129101e-06,
+      "loss": 2.0747,
+      "step": 16008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.373920412248369e-06,
+      "loss": 2.0605,
+      "step": 16009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.372530843692129e-06,
+      "loss": 2.0496,
+      "step": 16010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3711415734968105e-06,
+      "loss": 2.0742,
+      "step": 16011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.369752601698864e-06,
+      "loss": 2.0981,
+      "step": 16012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.368363928334693e-06,
+      "loss": 2.0232,
+      "step": 16013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.366975553440734e-06,
+      "loss": 1.9843,
+      "step": 16014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.365587477053376e-06,
+      "loss": 2.0051,
+      "step": 16015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3641996992090345e-06,
+      "loss": 2.0112,
+      "step": 16016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3628122199440965e-06,
+      "loss": 2.0386,
+      "step": 16017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.361425039294946e-06,
+      "loss": 2.0281,
+      "step": 16018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3600381572979625e-06,
+      "loss": 2.0537,
+      "step": 16019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.358651573989515e-06,
+      "loss": 2.0356,
+      "step": 16020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.357265289405965e-06,
+      "loss": 2.093,
+      "step": 16021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.355879303583665e-06,
+      "loss": 2.094,
+      "step": 16022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.354493616558961e-06,
+      "loss": 2.0874,
+      "step": 16023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.353108228368195e-06,
+      "loss": 2.105,
+      "step": 16024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.351723139047695e-06,
+      "loss": 2.0801,
+      "step": 16025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.350338348633774e-06,
+      "loss": 2.0974,
+      "step": 16026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.348953857162765e-06,
+      "loss": 2.0273,
+      "step": 16027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3475696646709566e-06,
+      "loss": 1.9951,
+      "step": 16028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.346185771194662e-06,
+      "loss": 2.0933,
+      "step": 16029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.344802176770161e-06,
+      "loss": 2.085,
+      "step": 16030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3434188814337435e-06,
+      "loss": 2.0842,
+      "step": 16031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.342035885221683e-06,
+      "loss": 2.1008,
+      "step": 16032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.340653188170246e-06,
+      "loss": 2.0642,
+      "step": 16033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3392707903156945e-06,
+      "loss": 2.0691,
+      "step": 16034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3378886916942775e-06,
+      "loss": 2.0659,
+      "step": 16035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.336506892342239e-06,
+      "loss": 2.0474,
+      "step": 16036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.335125392295819e-06,
+      "loss": 2.0244,
+      "step": 16037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.333744191591242e-06,
+      "loss": 2.0088,
+      "step": 16038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.332363290264729e-06,
+      "loss": 2.0518,
+      "step": 16039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.330982688352496e-06,
+      "loss": 2.0386,
+      "step": 16040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.329602385890739e-06,
+      "loss": 2.0457,
+      "step": 16041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.328222382915667e-06,
+      "loss": 2.0374,
+      "step": 16042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.326842679463453e-06,
+      "loss": 2.0674,
+      "step": 16043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.325463275570297e-06,
+      "loss": 2.05,
+      "step": 16044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.32408417127236e-06,
+      "loss": 2.0247,
+      "step": 16045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.322705366605809e-06,
+      "loss": 2.0327,
+      "step": 16046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.321326861606803e-06,
+      "loss": 2.1033,
+      "step": 16047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.319948656311493e-06,
+      "loss": 2.1143,
+      "step": 16048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.318570750756018e-06,
+      "loss": 2.0645,
+      "step": 16049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.317193144976513e-06,
+      "loss": 2.0688,
+      "step": 16050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.315815839009104e-06,
+      "loss": 2.1023,
+      "step": 16051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.314438832889911e-06,
+      "loss": 2.0964,
+      "step": 16052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.313062126655042e-06,
+      "loss": 2.1289,
+      "step": 16053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.311685720340601e-06,
+      "loss": 2.1338,
+      "step": 16054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.310309613982687e-06,
+      "loss": 2.0872,
+      "step": 16055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.308933807617372e-06,
+      "loss": 2.0625,
+      "step": 16056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.307558301280755e-06,
+      "loss": 2.0527,
+      "step": 16057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3061830950088885e-06,
+      "loss": 2.054,
+      "step": 16058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.304808188837851e-06,
+      "loss": 2.0784,
+      "step": 16059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3034335828036846e-06,
+      "loss": 2.092,
+      "step": 16060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.302059276942449e-06,
+      "loss": 2.0947,
+      "step": 16061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3006852712901755e-06,
+      "loss": 2.0742,
+      "step": 16062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2993115658829e-06,
+      "loss": 2.021,
+      "step": 16063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.297938160756642e-06,
+      "loss": 2.0002,
+      "step": 16064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.29656505594742e-06,
+      "loss": 2.0737,
+      "step": 16065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2951922514912435e-06,
+      "loss": 2.0754,
+      "step": 16066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2938197474241145e-06,
+      "loss": 2.0662,
+      "step": 16067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.292447543782015e-06,
+      "loss": 2.0459,
+      "step": 16068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.291075640600944e-06,
+      "loss": 2.0588,
+      "step": 16069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.289704037916863e-06,
+      "loss": 2.0884,
+      "step": 16070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.288332735765756e-06,
+      "loss": 2.0566,
+      "step": 16071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2869617341835725e-06,
+      "loss": 2.0745,
+      "step": 16072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2855910332062645e-06,
+      "loss": 1.96,
+      "step": 16073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.284220632869788e-06,
+      "loss": 1.9604,
+      "step": 16074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2828505332100685e-06,
+      "loss": 2.0217,
+      "step": 16075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.281480734263047e-06,
+      "loss": 2.0078,
+      "step": 16076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.280111236064636e-06,
+      "loss": 2.053,
+      "step": 16077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2787420386507505e-06,
+      "loss": 2.0491,
+      "step": 16078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2773731420572985e-06,
+      "loss": 2.0605,
+      "step": 16079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.276004546320174e-06,
+      "loss": 2.0535,
+      "step": 16080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.274636251475269e-06,
+      "loss": 2.0681,
+      "step": 16081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.273268257558471e-06,
+      "loss": 2.0547,
+      "step": 16082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.271900564605641e-06,
+      "loss": 2.0474,
+      "step": 16083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.270533172652659e-06,
+      "loss": 2.0764,
+      "step": 16084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.269166081735371e-06,
+      "loss": 2.0779,
+      "step": 16085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.267799291889641e-06,
+      "loss": 2.0757,
+      "step": 16086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2664328031513e-06,
+      "loss": 2.1072,
+      "step": 16087
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.265066615556186e-06,
+      "loss": 2.1497,
+      "step": 16088
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.263700729140125e-06,
+      "loss": 2.0146,
+      "step": 16089
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.262335143938938e-06,
+      "loss": 2.0002,
+      "step": 16090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.260969859988435e-06,
+      "loss": 2.1189,
+      "step": 16091
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.259604877324419e-06,
+      "loss": 2.1143,
+      "step": 16092
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2582401959826865e-06,
+      "loss": 2.0527,
+      "step": 16093
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.25687581599902e-06,
+      "loss": 2.051,
+      "step": 16094
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.255511737409204e-06,
+      "loss": 2.0186,
+      "step": 16095
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.254147960249007e-06,
+      "loss": 2.0171,
+      "step": 16096
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.252784484554197e-06,
+      "loss": 2.0762,
+      "step": 16097
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2514213103605185e-06,
+      "loss": 2.098,
+      "step": 16098
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2500584377037345e-06,
+      "loss": 2.064,
+      "step": 16099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.248695866619569e-06,
+      "loss": 2.1135,
+      "step": 16100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2473335971437675e-06,
+      "loss": 2.0859,
+      "step": 16101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.245971629312045e-06,
+      "loss": 2.0742,
+      "step": 16102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.244609963160121e-06,
+      "loss": 2.0381,
+      "step": 16103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.243248598723702e-06,
+      "loss": 2.0469,
+      "step": 16104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2418875360384905e-06,
+      "loss": 2.0798,
+      "step": 16105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2405267751401794e-06,
+      "loss": 2.0676,
+      "step": 16106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.239166316064448e-06,
+      "loss": 2.0869,
+      "step": 16107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.237806158846979e-06,
+      "loss": 2.0723,
+      "step": 16108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2364463035234364e-06,
+      "loss": 2.0493,
+      "step": 16109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.235086750129482e-06,
+      "loss": 2.0786,
+      "step": 16110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.233727498700769e-06,
+      "loss": 2.0122,
+      "step": 16111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.232368549272947e-06,
+      "loss": 2.0317,
+      "step": 16112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.23100990188164e-06,
+      "loss": 2.0295,
+      "step": 16113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.229651556562494e-06,
+      "loss": 2.0364,
+      "step": 16114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2282935133511116e-06,
+      "loss": 2.0718,
+      "step": 16115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.226935772283124e-06,
+      "loss": 2.0896,
+      "step": 16116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.225578333394124e-06,
+      "loss": 2.0347,
+      "step": 16117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.224221196719711e-06,
+      "loss": 2.0493,
+      "step": 16118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.222864362295476e-06,
+      "loss": 2.0649,
+      "step": 16119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.221507830157001e-06,
+      "loss": 2.0735,
+      "step": 16120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.22015160033986e-06,
+      "loss": 2.0837,
+      "step": 16121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2187956728796155e-06,
+      "loss": 2.0674,
+      "step": 16122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.217440047811828e-06,
+      "loss": 2.0295,
+      "step": 16123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.216084725172046e-06,
+      "loss": 2.075,
+      "step": 16124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.21472970499581e-06,
+      "loss": 2.0088,
+      "step": 16125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.213374987318657e-06,
+      "loss": 1.978,
+      "step": 16126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.2120205721761136e-06,
+      "loss": 2.0789,
+      "step": 16127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.21066645960369e-06,
+      "loss": 2.0696,
+      "step": 16128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.209312649636906e-06,
+      "loss": 2.0203,
+      "step": 16129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.207959142311255e-06,
+      "loss": 2.0337,
+      "step": 16130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.20660593766224e-06,
+      "loss": 2.0969,
+      "step": 16131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.205253035725341e-06,
+      "loss": 2.1204,
+      "step": 16132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.203900436536038e-06,
+      "loss": 2.1001,
+      "step": 16133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.202548140129799e-06,
+      "loss": 2.1067,
+      "step": 16134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.201196146542091e-06,
+      "loss": 2.0662,
+      "step": 16135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.199844455808365e-06,
+      "loss": 2.0991,
+      "step": 16136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1984930679640735e-06,
+      "loss": 2.0493,
+      "step": 16137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.197141983044641e-06,
+      "loss": 2.0784,
+      "step": 16138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.195791201085514e-06,
+      "loss": 2.0762,
+      "step": 16139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.194440722122105e-06,
+      "loss": 2.0554,
+      "step": 16140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.193090546189832e-06,
+      "loss": 2.1333,
+      "step": 16141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.19174067332411e-06,
+      "loss": 2.1355,
+      "step": 16142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1903911035603175e-06,
+      "loss": 2.0164,
+      "step": 16143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1890418369338705e-06,
+      "loss": 1.9817,
+      "step": 16144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1876928734801275e-06,
+      "loss": 2.0227,
+      "step": 16145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.186344213234486e-06,
+      "loss": 2.0327,
+      "step": 16146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1849958562322945e-06,
+      "loss": 2.0405,
+      "step": 16147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.183647802508924e-06,
+      "loss": 2.0386,
+      "step": 16148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.182300052099721e-06,
+      "loss": 2.0129,
+      "step": 16149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.180952605040028e-06,
+      "loss": 2.0154,
+      "step": 16150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.179605461365182e-06,
+      "loss": 2.0842,
+      "step": 16151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.178258621110513e-06,
+      "loss": 2.103,
+      "step": 16152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.17691208431133e-06,
+      "loss": 2.0605,
+      "step": 16153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.175565851002961e-06,
+      "loss": 2.0654,
+      "step": 16154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.17421992122069e-06,
+      "loss": 2.0627,
+      "step": 16155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.17287429499983e-06,
+      "loss": 2.072,
+      "step": 16156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.171528972375657e-06,
+      "loss": 2.0681,
+      "step": 16157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.170183953383454e-06,
+      "loss": 2.0496,
+      "step": 16158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.168839238058495e-06,
+      "loss": 2.0386,
+      "step": 16159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.167494826436038e-06,
+      "loss": 2.0376,
+      "step": 16160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.166150718551347e-06,
+      "loss": 2.0623,
+      "step": 16161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.164806914439661e-06,
+      "loss": 2.0925,
+      "step": 16162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.163463414136225e-06,
+      "loss": 2.0913,
+      "step": 16163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1621202176762706e-06,
+      "loss": 2.0786,
+      "step": 16164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.160777325095017e-06,
+      "loss": 2.0864,
+      "step": 16165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.159434736427685e-06,
+      "loss": 2.0784,
+      "step": 16166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.158092451709483e-06,
+      "loss": 2.0945,
+      "step": 16167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.156750470975604e-06,
+      "loss": 2.105,
+      "step": 16168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.155408794261249e-06,
+      "loss": 2.0244,
+      "step": 16169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1540674216015925e-06,
+      "loss": 2.0317,
+      "step": 16170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.152726353031821e-06,
+      "loss": 2.073,
+      "step": 16171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1513855885870925e-06,
+      "loss": 2.0771,
+      "step": 16172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.150045128302572e-06,
+      "loss": 2.0664,
+      "step": 16173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.148704972213412e-06,
+      "loss": 2.0903,
+      "step": 16174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.147365120354752e-06,
+      "loss": 2.0176,
+      "step": 16175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.146025572761735e-06,
+      "loss": 2.0183,
+      "step": 16176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.144686329469482e-06,
+      "loss": 2.0908,
+      "step": 16177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.143347390513119e-06,
+      "loss": 2.0879,
+      "step": 16178
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.142008755927754e-06,
+      "loss": 2.0325,
+      "step": 16179
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.140670425748496e-06,
+      "loss": 2.031,
+      "step": 16180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1393324000104356e-06,
+      "loss": 2.0972,
+      "step": 16181
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.137994678748663e-06,
+      "loss": 2.0786,
+      "step": 16182
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1366572619982576e-06,
+      "loss": 1.9817,
+      "step": 16183
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.135320149794298e-06,
+      "loss": 1.9729,
+      "step": 16184
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1339833421718355e-06,
+      "loss": 2.0278,
+      "step": 16185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.132646839165941e-06,
+      "loss": 2.0288,
+      "step": 16186
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.13131064081165e-06,
+      "loss": 2.0747,
+      "step": 16187
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.129974747144015e-06,
+      "loss": 2.0461,
+      "step": 16188
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.128639158198058e-06,
+      "loss": 2.1267,
+      "step": 16189
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.127303874008806e-06,
+      "loss": 2.1062,
+      "step": 16190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.125968894611277e-06,
+      "loss": 2.0596,
+      "step": 16191
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1246342200404775e-06,
+      "loss": 2.0767,
+      "step": 16192
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.12329985033141e-06,
+      "loss": 2.0354,
+      "step": 16193
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1219657855190656e-06,
+      "loss": 2.0383,
+      "step": 16194
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.120632025638428e-06,
+      "loss": 2.0796,
+      "step": 16195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.119298570724473e-06,
+      "loss": 2.0703,
+      "step": 16196
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.117965420812173e-06,
+      "loss": 2.0747,
+      "step": 16197
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.116632575936485e-06,
+      "loss": 2.0669,
+      "step": 16198
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.115300036132362e-06,
+      "loss": 2.0759,
+      "step": 16199
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.113967801434744e-06,
+      "loss": 2.0474,
+      "step": 16200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.112635871878579e-06,
+      "loss": 2.0305,
+      "step": 16201
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.111304247498777e-06,
+      "loss": 2.052,
+      "step": 16202
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.10997292833028e-06,
+      "loss": 2.0576,
+      "step": 16203
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.108641914407982e-06,
+      "loss": 2.021,
+      "step": 16204
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1073112057667956e-06,
+      "loss": 2.0811,
+      "step": 16205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1059808024416165e-06,
+      "loss": 2.0603,
+      "step": 16206
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1046507044673315e-06,
+      "loss": 2.031,
+      "step": 16207
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.103320911878823e-06,
+      "loss": 2.0297,
+      "step": 16208
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.101991424710966e-06,
+      "loss": 2.0662,
+      "step": 16209
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.1006622429986125e-06,
+      "loss": 2.0842,
+      "step": 16210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.099333366776632e-06,
+      "loss": 2.1045,
+      "step": 16211
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.098004796079866e-06,
+      "loss": 2.0615,
+      "step": 16212
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.096676530943159e-06,
+      "loss": 2.0549,
+      "step": 16213
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.095348571401343e-06,
+      "loss": 2.0667,
+      "step": 16214
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.094020917489233e-06,
+      "loss": 2.0435,
+      "step": 16215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.09269356924166e-06,
+      "loss": 2.0442,
+      "step": 16216
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.091366526693415e-06,
+      "loss": 2.1172,
+      "step": 16217
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.090039789879316e-06,
+      "loss": 2.0991,
+      "step": 16218
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.088713358834144e-06,
+      "loss": 2.0522,
+      "step": 16219
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.087387233592684e-06,
+      "loss": 2.0564,
+      "step": 16220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.086061414189715e-06,
+      "loss": 2.0696,
+      "step": 16221
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.084735900660004e-06,
+      "loss": 2.0598,
+      "step": 16222
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.083410693038311e-06,
+      "loss": 2.0835,
+      "step": 16223
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.082085791359392e-06,
+      "loss": 2.0764,
+      "step": 16224
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.080761195657978e-06,
+      "loss": 2.0439,
+      "step": 16225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.079436905968821e-06,
+      "loss": 2.0942,
+      "step": 16226
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.078112922326636e-06,
+      "loss": 2.1113,
+      "step": 16227
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.076789244766156e-06,
+      "loss": 2.0969,
+      "step": 16228
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.075465873322084e-06,
+      "loss": 2.0903,
+      "step": 16229
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0741428080291235e-06,
+      "loss": 2.0544,
+      "step": 16230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.072820048921973e-06,
+      "loss": 2.1431,
+      "step": 16231
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0714975960353205e-06,
+      "loss": 2.1421,
+      "step": 16232
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.070175449403843e-06,
+      "loss": 2.0808,
+      "step": 16233
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.068853609062216e-06,
+      "loss": 2.0674,
+      "step": 16234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.067532075045102e-06,
+      "loss": 2.0303,
+      "step": 16235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.066210847387155e-06,
+      "loss": 1.9948,
+      "step": 16236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.064889926123027e-06,
+      "loss": 2.0935,
+      "step": 16237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.063569311287351e-06,
+      "loss": 2.0845,
+      "step": 16238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.062249002914766e-06,
+      "loss": 2.0881,
+      "step": 16239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.060929001039887e-06,
+      "loss": 2.1038,
+      "step": 16240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.059609305697341e-06,
+      "loss": 2.0344,
+      "step": 16241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.058289916921721e-06,
+      "loss": 2.0183,
+      "step": 16242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.056970834747641e-06,
+      "loss": 2.1277,
+      "step": 16243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.055652059209682e-06,
+      "loss": 2.1167,
+      "step": 16244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.054333590342433e-06,
+      "loss": 2.0691,
+      "step": 16245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.053015428180463e-06,
+      "loss": 2.0137,
+      "step": 16246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0516975727583475e-06,
+      "loss": 2.0554,
+      "step": 16247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.050380024110642e-06,
+      "loss": 2.0559,
+      "step": 16248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.049062782271895e-06,
+      "loss": 2.0618,
+      "step": 16249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.047745847276652e-06,
+      "loss": 2.0767,
+      "step": 16250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.046429219159451e-06,
+      "loss": 2.0776,
+      "step": 16251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.045112897954814e-06,
+      "loss": 2.0496,
+      "step": 16252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.043796883697264e-06,
+      "loss": 2.1233,
+      "step": 16253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.04248117642131e-06,
+      "loss": 2.1082,
+      "step": 16254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.041165776161453e-06,
+      "loss": 2.0828,
+      "step": 16255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0398506829521935e-06,
+      "loss": 2.0884,
+      "step": 16256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0385358968280104e-06,
+      "loss": 2.0754,
+      "step": 16257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0372214178233916e-06,
+      "loss": 2.0542,
+      "step": 16258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.035907245972798e-06,
+      "loss": 2.0659,
+      "step": 16259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.034593381310699e-06,
+      "loss": 2.0813,
+      "step": 16260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.03327982387155e-06,
+      "loss": 2.0498,
+      "step": 16261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.031966573689791e-06,
+      "loss": 2.0684,
+      "step": 16262
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.030653630799866e-06,
+      "loss": 2.1067,
+      "step": 16263
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.029340995236204e-06,
+      "loss": 2.0955,
+      "step": 16264
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.028028667033227e-06,
+      "loss": 2.0747,
+      "step": 16265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.026716646225349e-06,
+      "loss": 2.1064,
+      "step": 16266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.025404932846976e-06,
+      "loss": 2.0833,
+      "step": 16267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.024093526932507e-06,
+      "loss": 2.0591,
+      "step": 16268
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.022782428516335e-06,
+      "loss": 2.0508,
+      "step": 16269
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0214716376328334e-06,
+      "loss": 2.0518,
+      "step": 16270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.020161154316388e-06,
+      "loss": 2.0471,
+      "step": 16271
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0188509786013515e-06,
+      "loss": 2.0498,
+      "step": 16272
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.017541110522097e-06,
+      "loss": 2.0667,
+      "step": 16273
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0162315501129595e-06,
+      "loss": 2.0537,
+      "step": 16274
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.014922297408289e-06,
+      "loss": 2.1033,
+      "step": 16275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.013613352442415e-06,
+      "loss": 2.1372,
+      "step": 16276
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.012304715249667e-06,
+      "loss": 2.1008,
+      "step": 16277
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0109963858643625e-06,
+      "loss": 2.1001,
+      "step": 16278
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.009688364320809e-06,
+      "loss": 2.0754,
+      "step": 16279
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.008380650653304e-06,
+      "loss": 2.0669,
+      "step": 16280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0070732448961465e-06,
+      "loss": 2.093,
+      "step": 16281
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0057661470836224e-06,
+      "loss": 2.0729,
+      "step": 16282
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.0044593572500045e-06,
+      "loss": 1.9988,
+      "step": 16283
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.003152875429568e-06,
+      "loss": 2.0222,
+      "step": 16284
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.001846701656561e-06,
+      "loss": 2.0955,
+      "step": 16285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.000540835965253e-06,
+      "loss": 2.095,
+      "step": 16286
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.999235278389875e-06,
+      "loss": 2.1052,
+      "step": 16287
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.997930028964677e-06,
+      "loss": 2.1311,
+      "step": 16288
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.996625087723875e-06,
+      "loss": 2.0354,
+      "step": 16289
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.995320454701696e-06,
+      "loss": 2.0059,
+      "step": 16290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.994016129932349e-06,
+      "loss": 2.0444,
+      "step": 16291
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.992712113450044e-06,
+      "loss": 2.0862,
+      "step": 16292
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.991408405288972e-06,
+      "loss": 2.032,
+      "step": 16293
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.990105005483327e-06,
+      "loss": 2.0366,
+      "step": 16294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.988801914067279e-06,
+      "loss": 1.9944,
+      "step": 16295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.987499131075012e-06,
+      "loss": 1.9814,
+      "step": 16296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.9861966565406795e-06,
+      "loss": 2.0911,
+      "step": 16297
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.984894490498448e-06,
+      "loss": 2.0715,
+      "step": 16298
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.983592632982457e-06,
+      "loss": 2.0947,
+      "step": 16299
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.982291084026849e-06,
+      "loss": 2.0933,
+      "step": 16300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.980989843665753e-06,
+      "loss": 2.062,
+      "step": 16301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.979688911933296e-06,
+      "loss": 2.0793,
+      "step": 16302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.978388288863596e-06,
+      "loss": 2.043,
+      "step": 16303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9770879744907525e-06,
+      "loss": 2.0757,
+      "step": 16304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.97578796884887e-06,
+      "loss": 2.0806,
+      "step": 16305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.974488271972038e-06,
+      "loss": 2.0981,
+      "step": 16306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.973188883894342e-06,
+      "loss": 2.073,
+      "step": 16307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.971889804649855e-06,
+      "loss": 2.0481,
+      "step": 16308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.970591034272643e-06,
+      "loss": 1.9927,
+      "step": 16309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.969292572796766e-06,
+      "loss": 2.0127,
+      "step": 16310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.967994420256277e-06,
+      "loss": 2.041,
+      "step": 16311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.966696576685209e-06,
+      "loss": 2.0198,
+      "step": 16312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.96539904211761e-06,
+      "loss": 2.031,
+      "step": 16313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.964101816587494e-06,
+      "loss": 2.0586,
+      "step": 16314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.962804900128891e-06,
+      "loss": 2.0808,
+      "step": 16315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.961508292775801e-06,
+      "loss": 2.071,
+      "step": 16316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9602119945622316e-06,
+      "loss": 2.0762,
+      "step": 16317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.958916005522174e-06,
+      "loss": 2.074,
+      "step": 16318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.957620325689616e-06,
+      "loss": 1.9824,
+      "step": 16319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.956324955098533e-06,
+      "loss": 2.0017,
+      "step": 16320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.955029893782897e-06,
+      "loss": 2.0845,
+      "step": 16321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.953735141776669e-06,
+      "loss": 2.0367,
+      "step": 16322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.952440699113801e-06,
+      "loss": 2.0605,
+      "step": 16323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.95114656582824e-06,
+      "loss": 2.0449,
+      "step": 16324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.949852741953923e-06,
+      "loss": 2.1113,
+      "step": 16325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9485592275247795e-06,
+      "loss": 2.1162,
+      "step": 16326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.947266022574722e-06,
+      "loss": 2.0786,
+      "step": 16327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.94597312713768e-06,
+      "loss": 2.1108,
+      "step": 16328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9446805412475404e-06,
+      "loss": 2.0522,
+      "step": 16329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.943388264938214e-06,
+      "loss": 2.1021,
+      "step": 16330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.942096298243582e-06,
+      "loss": 2.0471,
+      "step": 16331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.940804641197525e-06,
+      "loss": 2.0276,
+      "step": 16332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.939513293833917e-06,
+      "loss": 2.1216,
+      "step": 16333
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.938222256186621e-06,
+      "loss": 2.0886,
+      "step": 16334
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.936931528289493e-06,
+      "loss": 2.1118,
+      "step": 16335
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.935641110176381e-06,
+      "loss": 2.1243,
+      "step": 16336
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.934351001881125e-06,
+      "loss": 2.0845,
+      "step": 16337
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.933061203437558e-06,
+      "loss": 2.0894,
+      "step": 16338
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9317717148795e-06,
+      "loss": 2.0923,
+      "step": 16339
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.930482536240768e-06,
+      "loss": 2.0793,
+      "step": 16340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.929193667555175e-06,
+      "loss": 2.0261,
+      "step": 16341
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.927905108856506e-06,
+      "loss": 2.0471,
+      "step": 16342
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.926616860178568e-06,
+      "loss": 2.0457,
+      "step": 16343
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.925328921555128e-06,
+      "loss": 2.0479,
+      "step": 16344
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.92404129301998e-06,
+      "loss": 2.0557,
+      "step": 16345
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.922753974606873e-06,
+      "loss": 2.0314,
+      "step": 16346
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9214669663495725e-06,
+      "loss": 2.1008,
+      "step": 16347
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.920180268281828e-06,
+      "loss": 2.0676,
+      "step": 16348
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.918893880437384e-06,
+      "loss": 2.0747,
+      "step": 16349
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.91760780284997e-06,
+      "loss": 2.0627,
+      "step": 16350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9163220355533155e-06,
+      "loss": 2.0908,
+      "step": 16351
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.915036578581136e-06,
+      "loss": 2.1106,
+      "step": 16352
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.913751431967145e-06,
+      "loss": 2.0813,
+      "step": 16353
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.91246659574504e-06,
+      "loss": 2.1233,
+      "step": 16354
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.911182069948515e-06,
+      "loss": 2.0955,
+      "step": 16355
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.90989785461126e-06,
+      "loss": 2.0737,
+      "step": 16356
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9086139497669416e-06,
+      "loss": 2.0491,
+      "step": 16357
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.907330355449241e-06,
+      "loss": 2.043,
+      "step": 16358
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.906047071691805e-06,
+      "loss": 2.0984,
+      "step": 16359
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.904764098528305e-06,
+      "loss": 2.1289,
+      "step": 16360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.903481435992367e-06,
+      "loss": 2.0869,
+      "step": 16361
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.902199084117636e-06,
+      "loss": 2.0632,
+      "step": 16362
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.900917042937739e-06,
+      "loss": 1.9995,
+      "step": 16363
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.899635312486298e-06,
+      "loss": 1.9971,
+      "step": 16364
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.898353892796922e-06,
+      "loss": 2.0667,
+      "step": 16365
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.89707278390322e-06,
+      "loss": 2.064,
+      "step": 16366
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.895791985838776e-06,
+      "loss": 2.0549,
+      "step": 16367
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.894511498637191e-06,
+      "loss": 2.0266,
+      "step": 16368
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.893231322332034e-06,
+      "loss": 2.0479,
+      "step": 16369
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.891951456956886e-06,
+      "loss": 2.0391,
+      "step": 16370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.890671902545299e-06,
+      "loss": 2.1028,
+      "step": 16371
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8893926591308345e-06,
+      "loss": 2.1284,
+      "step": 16372
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.888113726747038e-06,
+      "loss": 2.1064,
+      "step": 16373
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8868351054274505e-06,
+      "loss": 2.094,
+      "step": 16374
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.885556795205595e-06,
+      "loss": 2.0422,
+      "step": 16375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.884278796115001e-06,
+      "loss": 2.0303,
+      "step": 16376
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.883001108189179e-06,
+      "loss": 2.0537,
+      "step": 16377
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8817237314616374e-06,
+      "loss": 2.0509,
+      "step": 16378
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.880446665965872e-06,
+      "loss": 2.0974,
+      "step": 16379
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.879169911735371e-06,
+      "loss": 2.1089,
+      "step": 16380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8778934688036236e-06,
+      "loss": 2.0588,
+      "step": 16381
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8766173372040875e-06,
+      "loss": 2.0405,
+      "step": 16382
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.875341516970247e-06,
+      "loss": 2.0908,
+      "step": 16383
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.874066008135541e-06,
+      "loss": 2.0769,
+      "step": 16384
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8727908107334305e-06,
+      "loss": 2.0718,
+      "step": 16385
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.871515924797353e-06,
+      "loss": 2.0967,
+      "step": 16386
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.870241350360737e-06,
+      "loss": 2.0662,
+      "step": 16387
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.868967087457011e-06,
+      "loss": 2.0779,
+      "step": 16388
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.867693136119589e-06,
+      "loss": 2.0552,
+      "step": 16389
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.866419496381882e-06,
+      "loss": 2.0732,
+      "step": 16390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.865146168277284e-06,
+      "loss": 2.0291,
+      "step": 16391
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.863873151839192e-06,
+      "loss": 2.0437,
+      "step": 16392
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.862600447100988e-06,
+      "loss": 2.0991,
+      "step": 16393
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.861328054096044e-06,
+      "loss": 2.1011,
+      "step": 16394
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8600559728577315e-06,
+      "loss": 2.0669,
+      "step": 16395
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8587842034194115e-06,
+      "loss": 2.0742,
+      "step": 16396
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.857512745814424e-06,
+      "loss": 2.0793,
+      "step": 16397
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8562416000761234e-06,
+      "loss": 2.0466,
+      "step": 16398
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.854970766237833e-06,
+      "loss": 2.075,
+      "step": 16399
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.853700244332889e-06,
+      "loss": 2.0776,
+      "step": 16400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.852430034394603e-06,
+      "loss": 2.0432,
+      "step": 16401
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.851160136456287e-06,
+      "loss": 2.0596,
+      "step": 16402
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.849890550551243e-06,
+      "loss": 2.0793,
+      "step": 16403
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.848621276712763e-06,
+      "loss": 2.0964,
+      "step": 16404
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.847352314974134e-06,
+      "loss": 2.0483,
+      "step": 16405
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.846083665368631e-06,
+      "loss": 2.0991,
+      "step": 16406
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.844815327929526e-06,
+      "loss": 2.0574,
+      "step": 16407
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.843547302690075e-06,
+      "loss": 2.0547,
+      "step": 16408
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.842279589683535e-06,
+      "loss": 2.0757,
+      "step": 16409
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8410121889431465e-06,
+      "loss": 2.0569,
+      "step": 16410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.839745100502153e-06,
+      "loss": 2.0403,
+      "step": 16411
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.838478324393771e-06,
+      "loss": 2.0586,
+      "step": 16412
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.83721186065123e-06,
+      "loss": 2.0706,
+      "step": 16413
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.835945709307732e-06,
+      "loss": 2.0605,
+      "step": 16414
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.834679870396493e-06,
+      "loss": 2.0552,
+      "step": 16415
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.833414343950699e-06,
+      "loss": 2.0251,
+      "step": 16416
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8321491300035375e-06,
+      "loss": 2.0869,
+      "step": 16417
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.830884228588191e-06,
+      "loss": 2.1062,
+      "step": 16418
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.829619639737826e-06,
+      "loss": 2.0774,
+      "step": 16419
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.828355363485608e-06,
+      "loss": 2.0544,
+      "step": 16420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.82709139986469e-06,
+      "loss": 2.103,
+      "step": 16421
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.82582774890822e-06,
+      "loss": 2.0867,
+      "step": 16422
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.824564410649331e-06,
+      "loss": 2.0701,
+      "step": 16423
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.823301385121157e-06,
+      "loss": 2.0752,
+      "step": 16424
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.822038672356816e-06,
+      "loss": 2.1113,
+      "step": 16425
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8207762723894245e-06,
+      "loss": 2.0903,
+      "step": 16426
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.819514185252087e-06,
+      "loss": 2.125,
+      "step": 16427
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8182524109779e-06,
+      "loss": 2.0972,
+      "step": 16428
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.816990949599947e-06,
+      "loss": 2.1067,
+      "step": 16429
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.815729801151319e-06,
+      "loss": 2.1125,
+      "step": 16430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8144689656650742e-06,
+      "loss": 2.0991,
+      "step": 16431
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8132084431742927e-06,
+      "loss": 2.1196,
+      "step": 16432
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8119482337120153e-06,
+      "loss": 2.0725,
+      "step": 16433
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8106883373112983e-06,
+      "loss": 2.1179,
+      "step": 16434
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.809428754005178e-06,
+      "loss": 2.0686,
+      "step": 16435
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.808169483826685e-06,
+      "loss": 2.061,
+      "step": 16436
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.806910526808846e-06,
+      "loss": 2.0457,
+      "step": 16437
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.805651882984675e-06,
+      "loss": 2.074,
+      "step": 16438
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8043935523871697e-06,
+      "loss": 2.0991,
+      "step": 16439
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.803135535049343e-06,
+      "loss": 2.0825,
+      "step": 16440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8018778310041696e-06,
+      "loss": 2.0527,
+      "step": 16441
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.800620440284647e-06,
+      "loss": 2.0488,
+      "step": 16442
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7993633629237346e-06,
+      "loss": 2.0791,
+      "step": 16443
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.798106598954405e-06,
+      "loss": 2.0867,
+      "step": 16444
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.796850148409615e-06,
+      "loss": 2.0681,
+      "step": 16445
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.795594011322309e-06,
+      "loss": 2.0432,
+      "step": 16446
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7943381877254392e-06,
+      "loss": 2.0798,
+      "step": 16447
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.793082677651927e-06,
+      "loss": 2.1096,
+      "step": 16448
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.791827481134699e-06,
+      "loss": 2.052,
+      "step": 16449
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7905725982066736e-06,
+      "loss": 2.064,
+      "step": 16450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.789318028900757e-06,
+      "loss": 2.0859,
+      "step": 16451
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.78806377324985e-06,
+      "loss": 2.0803,
+      "step": 16452
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7868098312868444e-06,
+      "loss": 2.0869,
+      "step": 16453
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7855562030446177e-06,
+      "loss": 2.0842,
+      "step": 16454
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.784302888556056e-06,
+      "loss": 2.0361,
+      "step": 16455
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.783049887854011e-06,
+      "loss": 2.0728,
+      "step": 16456
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7817972009713595e-06,
+      "loss": 2.0691,
+      "step": 16457
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.780544827940936e-06,
+      "loss": 2.0618,
+      "step": 16458
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.77929276879559e-06,
+      "loss": 2.0718,
+      "step": 16459
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7780410235681528e-06,
+      "loss": 2.0806,
+      "step": 16460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.776789592291451e-06,
+      "loss": 2.0735,
+      "step": 16461
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7755384749983013e-06,
+      "loss": 2.0642,
+      "step": 16462
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.774287671721515e-06,
+      "loss": 2.146,
+      "step": 16463
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7730371824938898e-06,
+      "loss": 2.1543,
+      "step": 16464
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7717870073482203e-06,
+      "loss": 2.1072,
+      "step": 16465
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7705371463172918e-06,
+      "loss": 2.1157,
+      "step": 16466
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.769287599433878e-06,
+      "loss": 2.0583,
+      "step": 16467
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7680383667307504e-06,
+      "loss": 2.0601,
+      "step": 16468
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7667894482406606e-06,
+      "loss": 2.0322,
+      "step": 16469
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7655408439963736e-06,
+      "loss": 2.0378,
+      "step": 16470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7642925540306143e-06,
+      "loss": 2.0664,
+      "step": 16471
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7630445783761393e-06,
+      "loss": 2.1006,
+      "step": 16472
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7617969170656563e-06,
+      "loss": 2.0383,
+      "step": 16473
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.760549570131893e-06,
+      "loss": 2.0371,
+      "step": 16474
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.75930253760756e-06,
+      "loss": 2.0264,
+      "step": 16475
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7580558195253544e-06,
+      "loss": 2.04,
+      "step": 16476
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.756809415917975e-06,
+      "loss": 2.0491,
+      "step": 16477
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.755563326818104e-06,
+      "loss": 2.1016,
+      "step": 16478
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.75431755225842e-06,
+      "loss": 2.0708,
+      "step": 16479
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7530720922715923e-06,
+      "loss": 2.0466,
+      "step": 16480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.751826946890281e-06,
+      "loss": 2.0657,
+      "step": 16481
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.750582116147137e-06,
+      "loss": 2.0828,
+      "step": 16482
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7493376000748123e-06,
+      "loss": 2.0488,
+      "step": 16483
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7480933987059297e-06,
+      "loss": 2.0537,
+      "step": 16484
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7468495120731306e-06,
+      "loss": 2.0828,
+      "step": 16485
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7456059402090208e-06,
+      "loss": 2.0564,
+      "step": 16486
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.744362683146226e-06,
+      "loss": 2.0422,
+      "step": 16487
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.743119740917339e-06,
+      "loss": 2.0586,
+      "step": 16488
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.741877113554956e-06,
+      "loss": 2.0378,
+      "step": 16489
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.740634801091667e-06,
+      "loss": 2.0645,
+      "step": 16490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7393928035600467e-06,
+      "loss": 2.0713,
+      "step": 16491
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.738151120992666e-06,
+      "loss": 2.0576,
+      "step": 16492
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.736909753422089e-06,
+      "loss": 2.0557,
+      "step": 16493
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7356687008808648e-06,
+      "loss": 2.0374,
+      "step": 16494
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.734427963401542e-06,
+      "loss": 2.1055,
+      "step": 16495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7331875410166566e-06,
+      "loss": 2.1333,
+      "step": 16496
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.731947433758737e-06,
+      "loss": 2.0759,
+      "step": 16497
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.7307076416603064e-06,
+      "loss": 2.0542,
+      "step": 16498
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.729468164753868e-06,
+      "loss": 2.0591,
+      "step": 16499
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.728229003071938e-06,
+      "loss": 2.0544,
+      "step": 16500
+    },
+    {
+      "Bleu_1": 0.7746447926410122,
+      "Bleu_2": 0.6140951008554155,
+      "Bleu_3": 0.4721838218448938,
+      "Bleu_4": 0.35920011839840793,
+      "CIDEr": 1.2062031019316448,
+      "METEOR": 0.27864590893331254,
+      "ROUGE_L": 0.5600376426755593,
+      "epoch": 0.82,
+      "step": 16500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.726990156647e-06,
+      "loss": 2.0583,
+      "step": 16501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.725751625511553e-06,
+      "loss": 2.0579,
+      "step": 16502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7245134096980664e-06,
+      "loss": 2.0884,
+      "step": 16503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7232755092390157e-06,
+      "loss": 2.0543,
+      "step": 16504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7220379241668617e-06,
+      "loss": 2.0964,
+      "step": 16505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.72080065451406e-06,
+      "loss": 2.0916,
+      "step": 16506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.719563700313056e-06,
+      "loss": 2.0483,
+      "step": 16507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7183270615962894e-06,
+      "loss": 2.0642,
+      "step": 16508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7170907383961823e-06,
+      "loss": 2.0718,
+      "step": 16509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.715854730745167e-06,
+      "loss": 2.0713,
+      "step": 16510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7146190386756424e-06,
+      "loss": 2.0881,
+      "step": 16511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7133836622200296e-06,
+      "loss": 2.1135,
+      "step": 16512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7121486014107127e-06,
+      "loss": 2.0586,
+      "step": 16513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.710913856280084e-06,
+      "loss": 2.0344,
+      "step": 16514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7096794268605223e-06,
+      "loss": 2.0383,
+      "step": 16515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7084453131843956e-06,
+      "loss": 2.019,
+      "step": 16516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7072115152840773e-06,
+      "loss": 2.0352,
+      "step": 16517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7059780331919123e-06,
+      "loss": 2.0396,
+      "step": 16518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.704744866940253e-06,
+      "loss": 2.1042,
+      "step": 16519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7035120165614344e-06,
+      "loss": 2.074,
+      "step": 16520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.702279482087788e-06,
+      "loss": 2.0671,
+      "step": 16521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7010472635516364e-06,
+      "loss": 2.054,
+      "step": 16522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6998153609852945e-06,
+      "loss": 2.0776,
+      "step": 16523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6985837744210595e-06,
+      "loss": 2.0859,
+      "step": 16524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6973525038912393e-06,
+      "loss": 2.0645,
+      "step": 16525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.696121549428112e-06,
+      "loss": 2.0662,
+      "step": 16526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6948909110639685e-06,
+      "loss": 2.0986,
+      "step": 16527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.693660588831071e-06,
+      "loss": 2.0923,
+      "step": 16528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6924305827616903e-06,
+      "loss": 2.0505,
+      "step": 16529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6912008928880782e-06,
+      "loss": 2.0696,
+      "step": 16530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6899715192424843e-06,
+      "loss": 2.043,
+      "step": 16531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6887424618571454e-06,
+      "loss": 2.0281,
+      "step": 16532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6875137207642936e-06,
+      "loss": 2.0952,
+      "step": 16533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6862852959961496e-06,
+      "loss": 2.0847,
+      "step": 16534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6850571875849282e-06,
+      "loss": 2.0583,
+      "step": 16535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6838293955628367e-06,
+      "loss": 2.0447,
+      "step": 16536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.682601919962071e-06,
+      "loss": 2.0574,
+      "step": 16537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6813747608148236e-06,
+      "loss": 2.0781,
+      "step": 16538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6801479181532656e-06,
+      "loss": 2.0361,
+      "step": 16539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.678921392009581e-06,
+      "loss": 2.0217,
+      "step": 16540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6776951824159246e-06,
+      "loss": 1.9856,
+      "step": 16541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6764692894044618e-06,
+      "loss": 2.0391,
+      "step": 16542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6752437130073337e-06,
+      "loss": 2.0591,
+      "step": 16543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6740184532566796e-06,
+      "loss": 2.0703,
+      "step": 16544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.672793510184632e-06,
+      "loss": 1.9927,
+      "step": 16545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.671568883823314e-06,
+      "loss": 2.019,
+      "step": 16546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6703445742048396e-06,
+      "loss": 2.062,
+      "step": 16547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6691205813613148e-06,
+      "loss": 2.0774,
+      "step": 16548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.667896905324839e-06,
+      "loss": 2.094,
+      "step": 16549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6666735461274996e-06,
+      "loss": 2.1111,
+      "step": 16550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.665450503801376e-06,
+      "loss": 2.0791,
+      "step": 16551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6642277783785464e-06,
+      "loss": 2.0623,
+      "step": 16552
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.663005369891073e-06,
+      "loss": 2.0554,
+      "step": 16553
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.661783278371009e-06,
+      "loss": 2.041,
+      "step": 16554
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.660561503850409e-06,
+      "loss": 2.0669,
+      "step": 16555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6593400463613024e-06,
+      "loss": 2.0586,
+      "step": 16556
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.658118905935733e-06,
+      "loss": 2.0771,
+      "step": 16557
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6568980826057113e-06,
+      "loss": 2.0808,
+      "step": 16558
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.655677576403262e-06,
+      "loss": 2.063,
+      "step": 16559
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6544573873603854e-06,
+      "loss": 2.0654,
+      "step": 16560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.653237515509083e-06,
+      "loss": 2.1074,
+      "step": 16561
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6520179608813415e-06,
+      "loss": 2.0684,
+      "step": 16562
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.650798723509147e-06,
+      "loss": 2.053,
+      "step": 16563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.649579803424465e-06,
+      "loss": 2.0554,
+      "step": 16564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6483612006592686e-06,
+      "loss": 2.0659,
+      "step": 16565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6471429152455075e-06,
+      "loss": 2.0619,
+      "step": 16566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6459249472151346e-06,
+      "loss": 2.0137,
+      "step": 16567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.644707296600087e-06,
+      "loss": 1.9789,
+      "step": 16568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.643489963432298e-06,
+      "loss": 2.1167,
+      "step": 16569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.642272947743692e-06,
+      "loss": 2.1062,
+      "step": 16570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.641056249566175e-06,
+      "loss": 2.0813,
+      "step": 16571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6398398689316685e-06,
+      "loss": 2.0947,
+      "step": 16572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6386238058720535e-06,
+      "loss": 2.071,
+      "step": 16573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6374080604192357e-06,
+      "loss": 2.0774,
+      "step": 16574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.636192632605088e-06,
+      "loss": 2.0669,
+      "step": 16575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.634977522461484e-06,
+      "loss": 2.0876,
+      "step": 16576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6337627300202877e-06,
+      "loss": 2.0662,
+      "step": 16577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.63254825531336e-06,
+      "loss": 2.0808,
+      "step": 16578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6313340983725455e-06,
+      "loss": 2.061,
+      "step": 16579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.630120259229688e-06,
+      "loss": 2.0769,
+      "step": 16580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6289067379166123e-06,
+      "loss": 2.1062,
+      "step": 16581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6276935344651507e-06,
+      "loss": 2.127,
+      "step": 16582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.626480648907106e-06,
+      "loss": 2.0667,
+      "step": 16583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6252680812742985e-06,
+      "loss": 2.0562,
+      "step": 16584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6240558315985166e-06,
+      "loss": 2.0469,
+      "step": 16585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.622843899911549e-06,
+      "loss": 2.03,
+      "step": 16586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6216322862451874e-06,
+      "loss": 2.0823,
+      "step": 16587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.620420990631192e-06,
+      "loss": 2.0854,
+      "step": 16588
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6192100131013422e-06,
+      "loss": 2.061,
+      "step": 16589
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.617999353687383e-06,
+      "loss": 2.0825,
+      "step": 16590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6167890124210666e-06,
+      "loss": 2.0901,
+      "step": 16591
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6155789893341317e-06,
+      "loss": 2.0801,
+      "step": 16592
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.614369284458311e-06,
+      "loss": 2.0876,
+      "step": 16593
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6131598978253274e-06,
+      "loss": 2.1135,
+      "step": 16594
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6119508294668987e-06,
+      "loss": 2.0879,
+      "step": 16595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.610742079414721e-06,
+      "loss": 2.0747,
+      "step": 16596
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.609533647700505e-06,
+      "loss": 2.0928,
+      "step": 16597
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.608325534355929e-06,
+      "loss": 2.071,
+      "step": 16598
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6071177394126853e-06,
+      "loss": 2.0779,
+      "step": 16599
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6059102629024397e-06,
+      "loss": 2.0952,
+      "step": 16600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.604703104856861e-06,
+      "loss": 2.0244,
+      "step": 16601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.603496265307601e-06,
+      "loss": 1.9916,
+      "step": 16602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.602289744286312e-06,
+      "loss": 2.0623,
+      "step": 16603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.60108354182463e-06,
+      "loss": 2.0806,
+      "step": 16604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5998776579541875e-06,
+      "loss": 2.092,
+      "step": 16605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5986720927066084e-06,
+      "loss": 2.0706,
+      "step": 16606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5974668461135076e-06,
+      "loss": 2.0833,
+      "step": 16607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5962619182064904e-06,
+      "loss": 2.083,
+      "step": 16608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.595057309017154e-06,
+      "loss": 2.0164,
+      "step": 16609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.593853018577093e-06,
+      "loss": 2.0098,
+      "step": 16610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5926490469178753e-06,
+      "loss": 2.0598,
+      "step": 16611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.591445394071091e-06,
+      "loss": 2.0762,
+      "step": 16612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.590242060068289e-06,
+      "loss": 2.0552,
+      "step": 16613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5890390449410397e-06,
+      "loss": 2.0674,
+      "step": 16614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5878363487208783e-06,
+      "loss": 2.0859,
+      "step": 16615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5866339714393486e-06,
+      "loss": 2.0803,
+      "step": 16616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.585431913127983e-06,
+      "loss": 2.0923,
+      "step": 16617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5842301738183037e-06,
+      "loss": 2.0967,
+      "step": 16618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5830287535418244e-06,
+      "loss": 2.0095,
+      "step": 16619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5818276523300516e-06,
+      "loss": 2.0342,
+      "step": 16620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5806268702144814e-06,
+      "loss": 2.0334,
+      "step": 16621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.579426407226602e-06,
+      "loss": 2.0242,
+      "step": 16622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5782262633978984e-06,
+      "loss": 2.1218,
+      "step": 16623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5770264387598404e-06,
+      "loss": 2.1265,
+      "step": 16624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.575826933343893e-06,
+      "loss": 2.0239,
+      "step": 16625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.574627747181506e-06,
+      "loss": 2.0554,
+      "step": 16626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.573428880304137e-06,
+      "loss": 2.0737,
+      "step": 16627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.572230332743213e-06,
+      "loss": 2.0813,
+      "step": 16628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5710321045301793e-06,
+      "loss": 2.043,
+      "step": 16629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5698341956964446e-06,
+      "loss": 1.9968,
+      "step": 16630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.568636606273428e-06,
+      "loss": 2.0835,
+      "step": 16631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5674393362925346e-06,
+      "loss": 2.0781,
+      "step": 16632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5662423857851633e-06,
+      "loss": 2.0735,
+      "step": 16633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.565045754782698e-06,
+      "loss": 2.095,
+      "step": 16634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5638494433165218e-06,
+      "loss": 2.0972,
+      "step": 16635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5626534514180082e-06,
+      "loss": 2.1267,
+      "step": 16636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5614577791185198e-06,
+      "loss": 2.0747,
+      "step": 16637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.560262426449409e-06,
+      "loss": 2.0625,
+      "step": 16638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.559067393442028e-06,
+      "loss": 2.0605,
+      "step": 16639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5578726801277112e-06,
+      "loss": 2.0286,
+      "step": 16640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5566782865377846e-06,
+      "loss": 2.0601,
+      "step": 16641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.55548421270358e-06,
+      "loss": 2.0818,
+      "step": 16642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5542904586563983e-06,
+      "loss": 2.0691,
+      "step": 16643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5530970244275574e-06,
+      "loss": 2.1023,
+      "step": 16644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5519039100483446e-06,
+      "loss": 2.041,
+      "step": 16645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5507111155500504e-06,
+      "loss": 2.04,
+      "step": 16646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5495186409639548e-06,
+      "loss": 2.0891,
+      "step": 16647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.548326486321329e-06,
+      "loss": 2.0239,
+      "step": 16648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5471346516534346e-06,
+      "loss": 2.043,
+      "step": 16649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5459431369915325e-06,
+      "loss": 2.0278,
+      "step": 16650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.544751942366855e-06,
+      "loss": 2.0442,
+      "step": 16651
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.543561067810656e-06,
+      "loss": 2.0403,
+      "step": 16652
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5423705133541507e-06,
+      "loss": 2.033,
+      "step": 16653
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5411802790285685e-06,
+      "loss": 2.0354,
+      "step": 16654
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5399903648651232e-06,
+      "loss": 2.0898,
+      "step": 16655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5388007708950114e-06,
+      "loss": 2.083,
+      "step": 16656
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.537611497149437e-06,
+      "loss": 2.0637,
+      "step": 16657
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.536422543659579e-06,
+      "loss": 2.0603,
+      "step": 16658
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5352339104566245e-06,
+      "loss": 2.0681,
+      "step": 16659
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.534045597571739e-06,
+      "loss": 2.0991,
+      "step": 16660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5328576050360857e-06,
+      "loss": 2.0586,
+      "step": 16661
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.531669932880818e-06,
+      "loss": 2.0879,
+      "step": 16662
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5304825811370806e-06,
+      "loss": 2.0745,
+      "step": 16663
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5292955498360123e-06,
+      "loss": 2.1074,
+      "step": 16664
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5281088390087453e-06,
+      "loss": 2.0173,
+      "step": 16665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5269224486863853e-06,
+      "loss": 2.03,
+      "step": 16666
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5257363789000636e-06,
+      "loss": 2.1038,
+      "step": 16667
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.524550629680866e-06,
+      "loss": 2.1152,
+      "step": 16668
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5233652010599033e-06,
+      "loss": 2.0474,
+      "step": 16669
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.522180093068245e-06,
+      "loss": 2.0369,
+      "step": 16670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.520995305736987e-06,
+      "loss": 2.0791,
+      "step": 16671
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.519810839097186e-06,
+      "loss": 2.0757,
+      "step": 16672
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5186266931799062e-06,
+      "loss": 2.0471,
+      "step": 16673
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5174428680162017e-06,
+      "loss": 2.0422,
+      "step": 16674
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.516259363637118e-06,
+      "loss": 2.0598,
+      "step": 16675
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5150761800736885e-06,
+      "loss": 2.0288,
+      "step": 16676
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.513893317356945e-06,
+      "loss": 2.0916,
+      "step": 16677
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5127107755179023e-06,
+      "loss": 2.0726,
+      "step": 16678
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.511528554587571e-06,
+      "loss": 2.1226,
+      "step": 16679
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5103466545969576e-06,
+      "loss": 2.1104,
+      "step": 16680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.509165075577054e-06,
+      "loss": 2.1292,
+      "step": 16681
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.507983817558848e-06,
+      "loss": 2.1033,
+      "step": 16682
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.506802880573307e-06,
+      "loss": 2.0969,
+      "step": 16683
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5056222646514135e-06,
+      "loss": 2.0884,
+      "step": 16684
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.504441969824116e-06,
+      "loss": 2.0237,
+      "step": 16685
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.503261996122378e-06,
+      "loss": 2.0317,
+      "step": 16686
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5020823435771326e-06,
+      "loss": 2.043,
+      "step": 16687
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.500903012219319e-06,
+      "loss": 2.0215,
+      "step": 16688
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4997240020798627e-06,
+      "loss": 2.0994,
+      "step": 16689
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4985453131896833e-06,
+      "loss": 2.0735,
+      "step": 16690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.497366945579693e-06,
+      "loss": 2.031,
+      "step": 16691
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4961888992807856e-06,
+      "loss": 2.0432,
+      "step": 16692
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4950111743238593e-06,
+      "loss": 2.0798,
+      "step": 16693
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4938337707398e-06,
+      "loss": 2.0413,
+      "step": 16694
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4926566885594806e-06,
+      "loss": 2.082,
+      "step": 16695
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4914799278137705e-06,
+      "loss": 2.0676,
+      "step": 16696
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4903034885335293e-06,
+      "loss": 2.0737,
+      "step": 16697
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4891273707495998e-06,
+      "loss": 2.1138,
+      "step": 16698
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4879515744928383e-06,
+      "loss": 2.0981,
+      "step": 16699
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4867760997940667e-06,
+      "loss": 2.0723,
+      "step": 16700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4856009466841176e-06,
+      "loss": 2.0718,
+      "step": 16701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4844261151938062e-06,
+      "loss": 2.0469,
+      "step": 16702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4832516053539393e-06,
+      "loss": 2.1191,
+      "step": 16703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4820774171953175e-06,
+      "loss": 2.1008,
+      "step": 16704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4809035507487326e-06,
+      "loss": 2.0669,
+      "step": 16705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.47973000604497e-06,
+      "loss": 2.0793,
+      "step": 16706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4785567831148035e-06,
+      "loss": 2.1006,
+      "step": 16707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4773838819889968e-06,
+      "loss": 2.1313,
+      "step": 16708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.476211302698312e-06,
+      "loss": 2.0947,
+      "step": 16709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4750390452734966e-06,
+      "loss": 2.0891,
+      "step": 16710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.473867109745292e-06,
+      "loss": 2.0547,
+      "step": 16711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4726954961444335e-06,
+      "loss": 2.0295,
+      "step": 16712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4715242045016356e-06,
+      "loss": 2.0779,
+      "step": 16713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.470353234847628e-06,
+      "loss": 2.0669,
+      "step": 16714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.469182587213105e-06,
+      "loss": 1.9978,
+      "step": 16715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4680122616287784e-06,
+      "loss": 2.0393,
+      "step": 16716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.466842258125327e-06,
+      "loss": 2.0703,
+      "step": 16717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4656725767334384e-06,
+      "loss": 2.0854,
+      "step": 16718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4645032174837843e-06,
+      "loss": 2.103,
+      "step": 16719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4633341804070318e-06,
+      "loss": 2.0918,
+      "step": 16720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4621654655338362e-06,
+      "loss": 2.1064,
+      "step": 16721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4609970728948488e-06,
+      "loss": 2.0969,
+      "step": 16722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4598290025207007e-06,
+      "loss": 2.0225,
+      "step": 16723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4586612544420316e-06,
+      "loss": 1.9729,
+      "step": 16724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.457493828689463e-06,
+      "loss": 2.0146,
+      "step": 16725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4563267252936076e-06,
+      "loss": 2.054,
+      "step": 16726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.455159944285074e-06,
+      "loss": 2.0935,
+      "step": 16727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4539934856944525e-06,
+      "loss": 2.1094,
+      "step": 16728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.452827349552341e-06,
+      "loss": 2.0742,
+      "step": 16729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.451661535889313e-06,
+      "loss": 2.0876,
+      "step": 16730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4504960447359487e-06,
+      "loss": 2.0884,
+      "step": 16731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.449330876122802e-06,
+      "loss": 2.1162,
+      "step": 16732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4481660300804356e-06,
+      "loss": 2.0317,
+      "step": 16733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4470015066393917e-06,
+      "loss": 2.0298,
+      "step": 16734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4458373058302107e-06,
+      "loss": 2.1018,
+      "step": 16735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.444673427683423e-06,
+      "loss": 2.0647,
+      "step": 16736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.443509872229551e-06,
+      "loss": 2.1014,
+      "step": 16737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4423466394991006e-06,
+      "loss": 2.085,
+      "step": 16738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4411837295225866e-06,
+      "loss": 2.1079,
+      "step": 16739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.440021142330494e-06,
+      "loss": 2.1077,
+      "step": 16740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4388588779533208e-06,
+      "loss": 2.0781,
+      "step": 16741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.43769693642154e-06,
+      "loss": 2.0801,
+      "step": 16742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4365353177656235e-06,
+      "loss": 2.0503,
+      "step": 16743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4353740220160303e-06,
+      "loss": 2.0271,
+      "step": 16744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.434213049203221e-06,
+      "loss": 2.0647,
+      "step": 16745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4330523993576347e-06,
+      "loss": 2.0615,
+      "step": 16746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4318920725097103e-06,
+      "loss": 2.0574,
+      "step": 16747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.430732068689875e-06,
+      "loss": 2.0659,
+      "step": 16748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.429572387928552e-06,
+      "loss": 2.0859,
+      "step": 16749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.428413030256147e-06,
+      "loss": 2.0896,
+      "step": 16750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4272539957030673e-06,
+      "loss": 2.1309,
+      "step": 16751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4260952842997077e-06,
+      "loss": 2.0913,
+      "step": 16752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4249368960764477e-06,
+      "loss": 2.0276,
+      "step": 16753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.423778831063676e-06,
+      "loss": 2.0305,
+      "step": 16754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4226210892917446e-06,
+      "loss": 2.0471,
+      "step": 16755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4214636707910334e-06,
+      "loss": 2.0396,
+      "step": 16756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.420306575591883e-06,
+      "loss": 2.0574,
+      "step": 16757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.419149803724635e-06,
+      "loss": 2.0569,
+      "step": 16758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.417993355219631e-06,
+      "loss": 2.0808,
+      "step": 16759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4168372301071927e-06,
+      "loss": 2.0886,
+      "step": 16760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.415681428417641e-06,
+      "loss": 2.0649,
+      "step": 16761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.414525950181286e-06,
+      "loss": 2.0876,
+      "step": 16762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4133707954284253e-06,
+      "loss": 2.0881,
+      "step": 16763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.412215964189356e-06,
+      "loss": 2.0774,
+      "step": 16764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4110614564943574e-06,
+      "loss": 2.0173,
+      "step": 16765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4099072723737087e-06,
+      "loss": 2.0256,
+      "step": 16766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4087534118576776e-06,
+      "loss": 2.0537,
+      "step": 16767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4075998749765167e-06,
+      "loss": 2.0518,
+      "step": 16768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4064466617604876e-06,
+      "loss": 2.0388,
+      "step": 16769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.405293772239816e-06,
+      "loss": 2.0442,
+      "step": 16770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.404141206444752e-06,
+      "loss": 2.0425,
+      "step": 16771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.402988964405507e-06,
+      "loss": 2.0269,
+      "step": 16772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4018370461523047e-06,
+      "loss": 2.0654,
+      "step": 16773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4006854517153496e-06,
+      "loss": 2.0463,
+      "step": 16774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3995341811248425e-06,
+      "loss": 2.094,
+      "step": 16775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3983832344109727e-06,
+      "loss": 2.0889,
+      "step": 16776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3972326116039227e-06,
+      "loss": 2.0613,
+      "step": 16777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.396082312733868e-06,
+      "loss": 2.0579,
+      "step": 16778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3949323378309734e-06,
+      "loss": 2.0251,
+      "step": 16779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.393782686925394e-06,
+      "loss": 2.0303,
+      "step": 16780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3926333600472782e-06,
+      "loss": 2.0781,
+      "step": 16781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.391484357226771e-06,
+      "loss": 2.0576,
+      "step": 16782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3903356784939943e-06,
+      "loss": 2.0974,
+      "step": 16783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3891873238790783e-06,
+      "loss": 2.0928,
+      "step": 16784
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3880392934121308e-06,
+      "loss": 2.0679,
+      "step": 16785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3868915871232693e-06,
+      "loss": 2.0623,
+      "step": 16786
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3857442050425746e-06,
+      "loss": 2.0776,
+      "step": 16787
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3845971472001514e-06,
+      "loss": 2.0903,
+      "step": 16788
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3834504136260686e-06,
+      "loss": 1.988,
+      "step": 16789
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.382304004350404e-06,
+      "loss": 1.9805,
+      "step": 16790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3811579194032167e-06,
+      "loss": 2.0498,
+      "step": 16791
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3800121588145646e-06,
+      "loss": 2.061,
+      "step": 16792
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3788667226144946e-06,
+      "loss": 2.0906,
+      "step": 16793
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3777216108330405e-06,
+      "loss": 2.0837,
+      "step": 16794
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.376576823500236e-06,
+      "loss": 2.1145,
+      "step": 16795
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3754323606460985e-06,
+      "loss": 2.105,
+      "step": 16796
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.374288222300641e-06,
+      "loss": 2.1001,
+      "step": 16797
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3731444084938677e-06,
+      "loss": 2.075,
+      "step": 16798
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3720009192557782e-06,
+      "loss": 2.0596,
+      "step": 16799
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3708577546163482e-06,
+      "loss": 2.0498,
+      "step": 16800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3697149146055674e-06,
+      "loss": 2.0918,
+      "step": 16801
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3685723992533947e-06,
+      "loss": 2.084,
+      "step": 16802
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.367430208589803e-06,
+      "loss": 2.0598,
+      "step": 16803
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3662883426447366e-06,
+      "loss": 2.0447,
+      "step": 16804
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.365146801448142e-06,
+      "loss": 2.0701,
+      "step": 16805
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3640055850299546e-06,
+      "loss": 2.0967,
+      "step": 16806
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3628646934201006e-06,
+      "loss": 2.0725,
+      "step": 16807
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3617241266485e-06,
+      "loss": 2.0959,
+      "step": 16808
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3605838847450646e-06,
+      "loss": 2.0295,
+      "step": 16809
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.35944396773969e-06,
+      "loss": 2.0425,
+      "step": 16810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.358304375662276e-06,
+      "loss": 2.061,
+      "step": 16811
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3571651085427e-06,
+      "loss": 2.0208,
+      "step": 16812
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3560261664108465e-06,
+      "loss": 2.0496,
+      "step": 16813
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.354887549296576e-06,
+      "loss": 2.0581,
+      "step": 16814
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.35374925722975e-06,
+      "loss": 2.1099,
+      "step": 16815
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.352611290240219e-06,
+      "loss": 2.0669,
+      "step": 16816
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.351473648357824e-06,
+      "loss": 2.051,
+      "step": 16817
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3503363316124e-06,
+      "loss": 2.1025,
+      "step": 16818
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.349199340033769e-06,
+      "loss": 2.0613,
+      "step": 16819
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3480626736517493e-06,
+      "loss": 2.073,
+      "step": 16820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3469263324961473e-06,
+      "loss": 2.0166,
+      "step": 16821
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.345790316596766e-06,
+      "loss": 1.9983,
+      "step": 16822
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.344654625983392e-06,
+      "loss": 2.072,
+      "step": 16823
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.343519260685811e-06,
+      "loss": 2.0869,
+      "step": 16824
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.342384220733789e-06,
+      "loss": 2.0715,
+      "step": 16825
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3412495061571035e-06,
+      "loss": 2.0776,
+      "step": 16826
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3401151169854976e-06,
+      "loss": 2.0825,
+      "step": 16827
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3389810532487307e-06,
+      "loss": 2.0601,
+      "step": 16828
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3378473149765358e-06,
+      "loss": 2.0815,
+      "step": 16829
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3367139021986457e-06,
+      "loss": 2.1021,
+      "step": 16830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3355808149447835e-06,
+      "loss": 2.1016,
+      "step": 16831
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3344480532446586e-06,
+      "loss": 2.1121,
+      "step": 16832
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.333315617127984e-06,
+      "loss": 2.072,
+      "step": 16833
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3321835066244506e-06,
+      "loss": 2.0432,
+      "step": 16834
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.33105172176375e-06,
+      "loss": 2.0278,
+      "step": 16835
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.32992026257556e-06,
+      "loss": 2.0117,
+      "step": 16836
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.32878912908955e-06,
+      "loss": 2.0791,
+      "step": 16837
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3276583213353867e-06,
+      "loss": 2.0776,
+      "step": 16838
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.326527839342728e-06,
+      "loss": 2.1377,
+      "step": 16839
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3253976831412065e-06,
+      "loss": 2.0876,
+      "step": 16840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.324267852760472e-06,
+      "loss": 2.0022,
+      "step": 16841
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3231383482301407e-06,
+      "loss": 2.0227,
+      "step": 16842
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3220091695798477e-06,
+      "loss": 2.0737,
+      "step": 16843
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.320880316839192e-06,
+      "loss": 2.0813,
+      "step": 16844
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.319751790037784e-06,
+      "loss": 2.1067,
+      "step": 16845
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.318623589205211e-06,
+      "loss": 2.0862,
+      "step": 16846
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3174957143710626e-06,
+      "loss": 2.0691,
+      "step": 16847
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.316368165564919e-06,
+      "loss": 2.0793,
+      "step": 16848
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3152409428163434e-06,
+      "loss": 2.0513,
+      "step": 16849
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3141140461548995e-06,
+      "loss": 2.0728,
+      "step": 16850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.312987475610138e-06,
+      "loss": 2.0447,
+      "step": 16851
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3118612312116e-06,
+      "loss": 2.0264,
+      "step": 16852
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.310735312988823e-06,
+      "loss": 2.0244,
+      "step": 16853
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.309609720971334e-06,
+      "loss": 2.0261,
+      "step": 16854
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.308484455188642e-06,
+      "loss": 2.0884,
+      "step": 16855
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.307359515670266e-06,
+      "loss": 2.0652,
+      "step": 16856
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.306234902445699e-06,
+      "loss": 2.0662,
+      "step": 16857
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3051106155444385e-06,
+      "loss": 2.0459,
+      "step": 16858
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3039866549959622e-06,
+      "loss": 2.0625,
+      "step": 16859
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3028630208297495e-06,
+      "loss": 2.0537,
+      "step": 16860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3017397130752606e-06,
+      "loss": 2.0212,
+      "step": 16861
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.300616731761958e-06,
+      "loss": 1.9972,
+      "step": 16862
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2994940769192887e-06,
+      "loss": 2.0359,
+      "step": 16863
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2983717485766935e-06,
+      "loss": 1.9833,
+      "step": 16864
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2972497467636046e-06,
+      "loss": 2.0693,
+      "step": 16865
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.296128071509443e-06,
+      "loss": 2.0134,
+      "step": 16866
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.295006722843625e-06,
+      "loss": 2.0947,
+      "step": 16867
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.293885700795556e-06,
+      "loss": 2.0464,
+      "step": 16868
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.292765005394639e-06,
+      "loss": 2.0444,
+      "step": 16869
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2916446366702496e-06,
+      "loss": 2.0288,
+      "step": 16870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2905245946517844e-06,
+      "loss": 2.0894,
+      "step": 16871
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2894048793685993e-06,
+      "loss": 2.1313,
+      "step": 16872
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.288285490850073e-06,
+      "loss": 2.0813,
+      "step": 16873
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2871664291255503e-06,
+      "loss": 2.0571,
+      "step": 16874
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2860476942243777e-06,
+      "loss": 2.03,
+      "step": 16875
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.284929286175897e-06,
+      "loss": 2.0173,
+      "step": 16876
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.283811205009434e-06,
+      "loss": 2.0312,
+      "step": 16877
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2826934507543088e-06,
+      "loss": 2.0544,
+      "step": 16878
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2815760234398384e-06,
+      "loss": 2.0869,
+      "step": 16879
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2804589230953157e-06,
+      "loss": 2.1042,
+      "step": 16880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2793421497500477e-06,
+      "loss": 2.0991,
+      "step": 16881
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2782257034333086e-06,
+      "loss": 2.0586,
+      "step": 16882
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.277109584174387e-06,
+      "loss": 2.0386,
+      "step": 16883
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2759937920025445e-06,
+      "loss": 2.0173,
+      "step": 16884
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2748783269470444e-06,
+      "loss": 2.0842,
+      "step": 16885
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.273763189037135e-06,
+      "loss": 2.0759,
+      "step": 16886
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2726483783020644e-06,
+      "loss": 2.1077,
+      "step": 16887
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2715338947710645e-06,
+      "loss": 2.1284,
+      "step": 16888
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2704197384733613e-06,
+      "loss": 2.093,
+      "step": 16889
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.269305909438172e-06,
+      "loss": 2.0989,
+      "step": 16890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2681924076947068e-06,
+      "loss": 2.115,
+      "step": 16891
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2670792332721665e-06,
+      "loss": 2.1094,
+      "step": 16892
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.26596638619974e-06,
+      "loss": 2.0732,
+      "step": 16893
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.264853866506616e-06,
+      "loss": 2.0557,
+      "step": 16894
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.26374167422196e-06,
+      "loss": 2.0776,
+      "step": 16895
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.262629809374948e-06,
+      "loss": 2.0635,
+      "step": 16896
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.261518271994726e-06,
+      "loss": 2.092,
+      "step": 16897
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2604070621104566e-06,
+      "loss": 2.094,
+      "step": 16898
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2592961797512688e-06,
+      "loss": 2.0413,
+      "step": 16899
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2581856249462984e-06,
+      "loss": 2.0132,
+      "step": 16900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.257075397724668e-06,
+      "loss": 2.0347,
+      "step": 16901
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.255965498115492e-06,
+      "loss": 2.0403,
+      "step": 16902
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2548559261478745e-06,
+      "loss": 2.0813,
+      "step": 16903
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2537466818509176e-06,
+      "loss": 2.0769,
+      "step": 16904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2526377652537064e-06,
+      "loss": 2.0813,
+      "step": 16905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2515291763853196e-06,
+      "loss": 2.0798,
+      "step": 16906
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2504209152748317e-06,
+      "loss": 2.0637,
+      "step": 16907
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2493129819513048e-06,
+      "loss": 2.0693,
+      "step": 16908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2482053764437903e-06,
+      "loss": 2.0925,
+      "step": 16909
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2470980987813386e-06,
+      "loss": 2.0889,
+      "step": 16910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.245991148992988e-06,
+      "loss": 2.0764,
+      "step": 16911
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.244884527107756e-06,
+      "loss": 2.0447,
+      "step": 16912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2437782331546766e-06,
+      "loss": 2.0693,
+      "step": 16913
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.242672267162749e-06,
+      "loss": 2.0847,
+      "step": 16914
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2415666291609845e-06,
+      "loss": 2.1055,
+      "step": 16915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.240461319178374e-06,
+      "loss": 2.1072,
+      "step": 16916
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2393563372439e-06,
+      "loss": 2.0676,
+      "step": 16917
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.238251683386545e-06,
+      "loss": 2.0964,
+      "step": 16918
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.237147357635274e-06,
+      "loss": 1.9788,
+      "step": 16919
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.236043360019047e-06,
+      "loss": 1.9807,
+      "step": 16920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2349396905668145e-06,
+      "loss": 2.0559,
+      "step": 16921
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.233836349307521e-06,
+      "loss": 2.0845,
+      "step": 16922
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.232733336270098e-06,
+      "loss": 2.0715,
+      "step": 16923
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2316306514834724e-06,
+      "loss": 2.0659,
+      "step": 16924
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.230528294976561e-06,
+      "loss": 2.0513,
+      "step": 16925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2294262667782727e-06,
+      "loss": 2.0681,
+      "step": 16926
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2283245669174998e-06,
+      "loss": 2.0708,
+      "step": 16927
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.227223195423143e-06,
+      "loss": 2.0747,
+      "step": 16928
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2261221523240755e-06,
+      "loss": 2.0383,
+      "step": 16929
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.225021437649182e-06,
+      "loss": 2.0251,
+      "step": 16930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.223921051427317e-06,
+      "loss": 2.0222,
+      "step": 16931
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2228209936873402e-06,
+      "loss": 2.0376,
+      "step": 16932
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2217212644581005e-06,
+      "loss": 2.0334,
+      "step": 16933
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2206218637684367e-06,
+      "loss": 2.0645,
+      "step": 16934
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.21952279164718e-06,
+      "loss": 2.0854,
+      "step": 16935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.218424048123151e-06,
+      "loss": 2.0793,
+      "step": 16936
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2173256332251615e-06,
+      "loss": 2.1287,
+      "step": 16937
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.216227546982019e-06,
+      "loss": 2.0969,
+      "step": 16938
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2151297894225186e-06,
+      "loss": 2.0415,
+      "step": 16939
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2140323605754468e-06,
+      "loss": 2.0718,
+      "step": 16940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.212935260469586e-06,
+      "loss": 2.0337,
+      "step": 16941
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2118384891337e-06,
+      "loss": 2.0051,
+      "step": 16942
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.210742046596559e-06,
+      "loss": 2.0906,
+      "step": 16943
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2096459328869027e-06,
+      "loss": 2.0835,
+      "step": 16944
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2085501480334905e-06,
+      "loss": 2.1106,
+      "step": 16945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2074546920650493e-06,
+      "loss": 2.1323,
+      "step": 16946
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2063595650103076e-06,
+      "loss": 2.0789,
+      "step": 16947
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2052647668979844e-06,
+      "loss": 2.0803,
+      "step": 16948
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2041702977567883e-06,
+      "loss": 2.0146,
+      "step": 16949
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.203076157615423e-06,
+      "loss": 2.0569,
+      "step": 16950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2019823465025816e-06,
+      "loss": 2.0715,
+      "step": 16951
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.2008888644469403e-06,
+      "loss": 2.0859,
+      "step": 16952
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.199795711477186e-06,
+      "loss": 2.0691,
+      "step": 16953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.198702887621972e-06,
+      "loss": 2.0598,
+      "step": 16954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1976103929099727e-06,
+      "loss": 2.0398,
+      "step": 16955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1965182273698247e-06,
+      "loss": 2.0698,
+      "step": 16956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1954263910301735e-06,
+      "loss": 2.0579,
+      "step": 16957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.194334883919648e-06,
+      "loss": 2.0547,
+      "step": 16958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1932437060668734e-06,
+      "loss": 2.0891,
+      "step": 16959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1921528575004704e-06,
+      "loss": 2.083,
+      "step": 16960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1910623382490376e-06,
+      "loss": 2.093,
+      "step": 16961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1899721483411744e-06,
+      "loss": 2.0518,
+      "step": 16962
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.18888228780547e-06,
+      "loss": 2.0891,
+      "step": 16963
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1877927566705077e-06,
+      "loss": 2.0901,
+      "step": 16964
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1867035549648543e-06,
+      "loss": 2.012,
+      "step": 16965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1856146827170783e-06,
+      "loss": 1.979,
+      "step": 16966
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1845261399557274e-06,
+      "loss": 2.1084,
+      "step": 16967
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1834379267093567e-06,
+      "loss": 2.1353,
+      "step": 16968
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.18235004300649e-06,
+      "loss": 2.1338,
+      "step": 16969
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1812624888756714e-06,
+      "loss": 2.1174,
+      "step": 16970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.180175264345408e-06,
+      "loss": 2.0916,
+      "step": 16971
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.179088369444217e-06,
+      "loss": 2.062,
+      "step": 16972
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1780018042005987e-06,
+      "loss": 2.0938,
+      "step": 16973
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1769155686430506e-06,
+      "loss": 2.1023,
+      "step": 16974
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.175829662800053e-06,
+      "loss": 2.074,
+      "step": 16975
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.174744086700087e-06,
+      "loss": 2.0759,
+      "step": 16976
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.173658840371616e-06,
+      "loss": 2.0149,
+      "step": 16977
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.172573923843104e-06,
+      "loss": 2.0088,
+      "step": 16978
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.171489337142998e-06,
+      "loss": 2.0759,
+      "step": 16979
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.170405080299744e-06,
+      "loss": 2.094,
+      "step": 16980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1693211533417743e-06,
+      "loss": 2.05,
+      "step": 16981
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1682375562975075e-06,
+      "loss": 2.0371,
+      "step": 16982
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.167154289195369e-06,
+      "loss": 2.0635,
+      "step": 16983
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.166071352063757e-06,
+      "loss": 2.0439,
+      "step": 16984
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1649887449310794e-06,
+      "loss": 2.0513,
+      "step": 16985
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1639064678257205e-06,
+      "loss": 2.0471,
+      "step": 16986
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.162824520776063e-06,
+      "loss": 2.0784,
+      "step": 16987
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1617429038104805e-06,
+      "loss": 2.0791,
+      "step": 16988
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1606616169573338e-06,
+      "loss": 2.0781,
+      "step": 16989
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1595806602449835e-06,
+      "loss": 2.0349,
+      "step": 16990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1585000337017735e-06,
+      "loss": 2.0667,
+      "step": 16991
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1574197373560426e-06,
+      "loss": 2.0691,
+      "step": 16992
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1563397712361203e-06,
+      "loss": 2.0876,
+      "step": 16993
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1552601353703287e-06,
+      "loss": 2.1077,
+      "step": 16994
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1541808297869767e-06,
+      "loss": 2.104,
+      "step": 16995
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.153101854514372e-06,
+      "loss": 2.072,
+      "step": 16996
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.152023209580803e-06,
+      "loss": 2.1033,
+      "step": 16997
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1509448950145665e-06,
+      "loss": 2.1096,
+      "step": 16998
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.149866910843927e-06,
+      "loss": 2.1152,
+      "step": 16999
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1487892570971675e-06,
+      "loss": 2.0957,
+      "step": 17000
+    },
+    {
+      "Bleu_1": 0.7745378452608926,
+      "Bleu_2": 0.6136642595545805,
+      "Bleu_3": 0.4715093724396915,
+      "Bleu_4": 0.35795627087375753,
+      "CIDEr": 1.2059338398317352,
+      "METEOR": 0.27889984485431907,
+      "ROUGE_L": 0.5598927372794362,
+      "epoch": 0.85,
+      "step": 17000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1477119338025357e-06,
+      "loss": 2.074,
+      "step": 17001
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.14663494098829e-06,
+      "loss": 2.0165,
+      "step": 17002
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1455582786826714e-06,
+      "loss": 2.0364,
+      "step": 17003
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.144481946913914e-06,
+      "loss": 2.0603,
+      "step": 17004
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1434059457102436e-06,
+      "loss": 2.0999,
+      "step": 17005
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1423302750998777e-06,
+      "loss": 2.0901,
+      "step": 17006
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1412549351110236e-06,
+      "loss": 2.0403,
+      "step": 17007
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1401799257718803e-06,
+      "loss": 2.0305,
+      "step": 17008
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1391052471106422e-06,
+      "loss": 2.0579,
+      "step": 17009
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1380308991554867e-06,
+      "loss": 2.0522,
+      "step": 17010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.136956881934593e-06,
+      "loss": 2.0642,
+      "step": 17011
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1358831954761164e-06,
+      "loss": 2.1084,
+      "step": 17012
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1348098398082264e-06,
+      "loss": 2.042,
+      "step": 17013
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1337368149590557e-06,
+      "loss": 2.0249,
+      "step": 17014
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.132664120956758e-06,
+      "loss": 2.0608,
+      "step": 17015
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1315917578294526e-06,
+      "loss": 2.0486,
+      "step": 17016
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.130519725605264e-06,
+      "loss": 2.1025,
+      "step": 17017
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.129448024312308e-06,
+      "loss": 2.0872,
+      "step": 17018
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1283766539786834e-06,
+      "loss": 2.0312,
+      "step": 17019
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1273056146324896e-06,
+      "loss": 2.0325,
+      "step": 17020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1262349063018156e-06,
+      "loss": 2.0549,
+      "step": 17021
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1251645290147296e-06,
+      "loss": 2.0745,
+      "step": 17022
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1240944827993136e-06,
+      "loss": 2.0349,
+      "step": 17023
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1230247676836167e-06,
+      "loss": 2.0515,
+      "step": 17024
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1219553836957016e-06,
+      "loss": 2.0637,
+      "step": 17025
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.120886330863604e-06,
+      "loss": 2.0537,
+      "step": 17026
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1198176092153624e-06,
+      "loss": 2.0432,
+      "step": 17027
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1187492187790017e-06,
+      "loss": 2.0676,
+      "step": 17028
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.117681159582533e-06,
+      "loss": 2.0977,
+      "step": 17029
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.11661343165398e-06,
+      "loss": 2.0681,
+      "step": 17030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1155460350213265e-06,
+      "loss": 2.1228,
+      "step": 17031
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.114478969712577e-06,
+      "loss": 2.1313,
+      "step": 17032
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1134122357557047e-06,
+      "loss": 2.0264,
+      "step": 17033
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.112345833178688e-06,
+      "loss": 2.0242,
+      "step": 17034
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1112797620094874e-06,
+      "loss": 2.0723,
+      "step": 17035
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1102140222760655e-06,
+      "loss": 2.0557,
+      "step": 17036
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1091486140063667e-06,
+      "loss": 2.0635,
+      "step": 17037
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1080835372283324e-06,
+      "loss": 2.0623,
+      "step": 17038
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1070187919698876e-06,
+      "loss": 2.0693,
+      "step": 17039
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.105954378258963e-06,
+      "loss": 2.0842,
+      "step": 17040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.104890296123458e-06,
+      "loss": 2.02,
+      "step": 17041
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1038265455912932e-06,
+      "loss": 2.0225,
+      "step": 17042
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1027631266903543e-06,
+      "loss": 2.0559,
+      "step": 17043
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1017000394485274e-06,
+      "loss": 2.052,
+      "step": 17044
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.1006372838936946e-06,
+      "loss": 2.0833,
+      "step": 17045
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0995748600537246e-06,
+      "loss": 2.0918,
+      "step": 17046
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.098512767956479e-06,
+      "loss": 2.0452,
+      "step": 17047
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0974510076298078e-06,
+      "loss": 2.0471,
+      "step": 17048
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0963895791015556e-06,
+      "loss": 2.0505,
+      "step": 17049
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0953284823995565e-06,
+      "loss": 2.064,
+      "step": 17050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.094267717551638e-06,
+      "loss": 2.1016,
+      "step": 17051
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.093207284585617e-06,
+      "loss": 2.0864,
+      "step": 17052
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0921471835293034e-06,
+      "loss": 1.9795,
+      "step": 17053
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0910874144104893e-06,
+      "loss": 1.9844,
+      "step": 17054
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0900279772569773e-06,
+      "loss": 2.0715,
+      "step": 17055
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0889688720965416e-06,
+      "loss": 2.073,
+      "step": 17056
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0879100989569614e-06,
+      "loss": 2.0371,
+      "step": 17057
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.086851657866e-06,
+      "loss": 2.0427,
+      "step": 17058
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.08579354885141e-06,
+      "loss": 2.0635,
+      "step": 17059
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.084735771940944e-06,
+      "loss": 2.0055,
+      "step": 17060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0836783271623386e-06,
+      "loss": 2.0898,
+      "step": 17061
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0826212145433246e-06,
+      "loss": 2.0942,
+      "step": 17062
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.081564434111625e-06,
+      "loss": 2.0398,
+      "step": 17063
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0805079858949492e-06,
+      "loss": 2.0276,
+      "step": 17064
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.079451869921003e-06,
+      "loss": 2.043,
+      "step": 17065
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0783960862174855e-06,
+      "loss": 2.0566,
+      "step": 17066
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.077340634812078e-06,
+      "loss": 2.0547,
+      "step": 17067
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0762855157324643e-06,
+      "loss": 2.019,
+      "step": 17068
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.075230729006304e-06,
+      "loss": 2.0281,
+      "step": 17069
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0741762746612675e-06,
+      "loss": 2.0461,
+      "step": 17070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.073122152725e-06,
+      "loss": 2.0381,
+      "step": 17071
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0720683632251506e-06,
+      "loss": 2.051,
+      "step": 17072
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0710149061893495e-06,
+      "loss": 2.0811,
+      "step": 17073
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.069961781645224e-06,
+      "loss": 2.0635,
+      "step": 17074
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0689089896203892e-06,
+      "loss": 2.0737,
+      "step": 17075
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0678565301424544e-06,
+      "loss": 2.041,
+      "step": 17076
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.066804403239019e-06,
+      "loss": 2.0732,
+      "step": 17077
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0657526089376765e-06,
+      "loss": 2.1267,
+      "step": 17078
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.064701147266004e-06,
+      "loss": 2.0439,
+      "step": 17079
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0636500182515787e-06,
+      "loss": 2.0664,
+      "step": 17080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.062599221921963e-06,
+      "loss": 2.0806,
+      "step": 17081
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.061548758304715e-06,
+      "loss": 2.0754,
+      "step": 17082
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.060498627427384e-06,
+      "loss": 2.1055,
+      "step": 17083
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0594488293174983e-06,
+      "loss": 2.1213,
+      "step": 17084
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.058399364002601e-06,
+      "loss": 2.0608,
+      "step": 17085
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0573502315102015e-06,
+      "loss": 2.0549,
+      "step": 17086
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.056301431867823e-06,
+      "loss": 2.052,
+      "step": 17087
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0552529651029583e-06,
+      "loss": 2.0181,
+      "step": 17088
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0542048312431105e-06,
+      "loss": 2.092,
+      "step": 17089
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0531570303157608e-06,
+      "loss": 2.0991,
+      "step": 17090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0521095623483873e-06,
+      "loss": 2.064,
+      "step": 17091
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0510624273684603e-06,
+      "loss": 2.0879,
+      "step": 17092
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0500156254034416e-06,
+      "loss": 2.0952,
+      "step": 17093
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0489691564807737e-06,
+      "loss": 2.1077,
+      "step": 17094
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0479230206279124e-06,
+      "loss": 2.0916,
+      "step": 17095
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.046877217872276e-06,
+      "loss": 2.0876,
+      "step": 17096
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0458317482413024e-06,
+      "loss": 2.0571,
+      "step": 17097
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0447866117624045e-06,
+      "loss": 2.0437,
+      "step": 17098
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.043741808462984e-06,
+      "loss": 2.0117,
+      "step": 17099
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0426973383704477e-06,
+      "loss": 1.9951,
+      "step": 17100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0416532015121763e-06,
+      "loss": 2.021,
+      "step": 17101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.040609397915563e-06,
+      "loss": 2.0183,
+      "step": 17102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0395659276079713e-06,
+      "loss": 2.0601,
+      "step": 17103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.038522790616767e-06,
+      "loss": 2.0339,
+      "step": 17104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.037479986969306e-06,
+      "loss": 2.103,
+      "step": 17105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.036437516692935e-06,
+      "loss": 2.0942,
+      "step": 17106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0353953798149885e-06,
+      "loss": 2.0469,
+      "step": 17107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0343535763628023e-06,
+      "loss": 2.0396,
+      "step": 17108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0333121063636858e-06,
+      "loss": 2.0527,
+      "step": 17109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.032270969844963e-06,
+      "loss": 2.0886,
+      "step": 17110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.031230166833925e-06,
+      "loss": 2.0476,
+      "step": 17111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.030189697357875e-06,
+      "loss": 2.0403,
+      "step": 17112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0291495614440922e-06,
+      "loss": 2.0845,
+      "step": 17113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0281097591198524e-06,
+      "loss": 2.0504,
+      "step": 17114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.027070290412427e-06,
+      "loss": 2.0984,
+      "step": 17115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.026031155349072e-06,
+      "loss": 2.0886,
+      "step": 17116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0249923539570396e-06,
+      "loss": 2.1064,
+      "step": 17117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.023953886263571e-06,
+      "loss": 2.0623,
+      "step": 17118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0229157522958956e-06,
+      "loss": 2.0383,
+      "step": 17119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.021877952081242e-06,
+      "loss": 2.0549,
+      "step": 17120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0208404856468225e-06,
+      "loss": 2.0415,
+      "step": 17121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0198033530198444e-06,
+      "loss": 2.0593,
+      "step": 17122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.018766554227506e-06,
+      "loss": 2.0984,
+      "step": 17123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.017730089296991e-06,
+      "loss": 2.0996,
+      "step": 17124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0166939582554877e-06,
+      "loss": 2.0454,
+      "step": 17125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.015658161130159e-06,
+      "loss": 2.0322,
+      "step": 17126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.014622697948178e-06,
+      "loss": 2.0457,
+      "step": 17127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0135875687366883e-06,
+      "loss": 2.0454,
+      "step": 17128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.012552773522838e-06,
+      "loss": 2.0146,
+      "step": 17129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.011518312333766e-06,
+      "loss": 2.0044,
+      "step": 17130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.010484185196598e-06,
+      "loss": 2.0715,
+      "step": 17131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0094503921384515e-06,
+      "loss": 2.0457,
+      "step": 17132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0084169331864393e-06,
+      "loss": 2.0776,
+      "step": 17133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0073838083676597e-06,
+      "loss": 2.0867,
+      "step": 17134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0063510177092095e-06,
+      "loss": 2.0679,
+      "step": 17135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.005318561238167e-06,
+      "loss": 2.105,
+      "step": 17136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.004286438981613e-06,
+      "loss": 2.053,
+      "step": 17137
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.003254650966609e-06,
+      "loss": 2.0779,
+      "step": 17138
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.0022231972202137e-06,
+      "loss": 2.0229,
+      "step": 17139
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.001192077769479e-06,
+      "loss": 2.022,
+      "step": 17140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.000161292641437e-06,
+      "loss": 2.0806,
+      "step": 17141
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9991308418631284e-06,
+      "loss": 2.1113,
+      "step": 17142
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9981007254615696e-06,
+      "loss": 1.9714,
+      "step": 17143
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9970709434637732e-06,
+      "loss": 1.9502,
+      "step": 17144
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9960414958967488e-06,
+      "loss": 2.0793,
+      "step": 17145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9950123827874903e-06,
+      "loss": 2.0908,
+      "step": 17146
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.993983604162984e-06,
+      "loss": 2.1118,
+      "step": 17147
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9929551600502104e-06,
+      "loss": 2.075,
+      "step": 17148
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.991927050476135e-06,
+      "loss": 2.052,
+      "step": 17149
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.990899275467724e-06,
+      "loss": 2.0835,
+      "step": 17150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9898718350519294e-06,
+      "loss": 2.0889,
+      "step": 17151
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9888447292556905e-06,
+      "loss": 2.0789,
+      "step": 17152
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.987817958105943e-06,
+      "loss": 2.125,
+      "step": 17153
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.986791521629617e-06,
+      "loss": 2.1143,
+      "step": 17154
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9857654198536266e-06,
+      "loss": 2.0439,
+      "step": 17155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9847396528048772e-06,
+      "loss": 2.0442,
+      "step": 17156
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9837142205102736e-06,
+      "loss": 2.0359,
+      "step": 17157
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.982689122996701e-06,
+      "loss": 2.0059,
+      "step": 17158
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9816643602910486e-06,
+      "loss": 2.0974,
+      "step": 17159
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9806399324201846e-06,
+      "loss": 2.0767,
+      "step": 17160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9796158394109734e-06,
+      "loss": 2.0725,
+      "step": 17161
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.978592081290271e-06,
+      "loss": 2.0754,
+      "step": 17162
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9775686580849257e-06,
+      "loss": 2.0667,
+      "step": 17163
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9765455698217764e-06,
+      "loss": 2.0642,
+      "step": 17164
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9755228165276527e-06,
+      "loss": 2.0391,
+      "step": 17165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.974500398229367e-06,
+      "loss": 2.0454,
+      "step": 17166
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9734783149537404e-06,
+      "loss": 2.0281,
+      "step": 17167
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.972456566727572e-06,
+      "loss": 2.0623,
+      "step": 17168
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9714351535776574e-06,
+      "loss": 2.1194,
+      "step": 17169
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9704140755307846e-06,
+      "loss": 2.106,
+      "step": 17170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9693933326137205e-06,
+      "loss": 2.0662,
+      "step": 17171
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9683729248532475e-06,
+      "loss": 2.0215,
+      "step": 17172
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.967352852276107e-06,
+      "loss": 2.1116,
+      "step": 17173
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9663331149090666e-06,
+      "loss": 2.1289,
+      "step": 17174
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.965313712778856e-06,
+      "loss": 2.062,
+      "step": 17175
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9642946459122122e-06,
+      "loss": 2.0632,
+      "step": 17176
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.963275914335859e-06,
+      "loss": 2.0972,
+      "step": 17177
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.962257518076508e-06,
+      "loss": 2.0598,
+      "step": 17178
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.961239457160872e-06,
+      "loss": 2.1099,
+      "step": 17179
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9602217316156454e-06,
+      "loss": 2.0911,
+      "step": 17180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.959204341467511e-06,
+      "loss": 2.0398,
+      "step": 17181
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9581872867431598e-06,
+      "loss": 2.0295,
+      "step": 17182
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.957170567469251e-06,
+      "loss": 2.0588,
+      "step": 17183
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9561541836724604e-06,
+      "loss": 2.0457,
+      "step": 17184
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9551381353794295e-06,
+      "loss": 2.0349,
+      "step": 17185
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.954122422616807e-06,
+      "loss": 2.0439,
+      "step": 17186
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9531070454112296e-06,
+      "loss": 2.0459,
+      "step": 17187
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.952092003789325e-06,
+      "loss": 2.0247,
+      "step": 17188
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.95107729777771e-06,
+      "loss": 2.064,
+      "step": 17189
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.950062927402995e-06,
+      "loss": 2.0781,
+      "step": 17190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9490488926917802e-06,
+      "loss": 2.0774,
+      "step": 17191
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.948035193670658e-06,
+      "loss": 2.1035,
+      "step": 17192
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9470218303662085e-06,
+      "loss": 2.0725,
+      "step": 17193
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.946008802805012e-06,
+      "loss": 2.094,
+      "step": 17194
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.944996111013631e-06,
+      "loss": 2.01,
+      "step": 17195
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9439837550186188e-06,
+      "loss": 2.0142,
+      "step": 17196
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9429717348465296e-06,
+      "loss": 2.1228,
+      "step": 17197
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9419600505238955e-06,
+      "loss": 2.127,
+      "step": 17198
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.940948702077253e-06,
+      "loss": 2.0691,
+      "step": 17199
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.93993768953312e-06,
+      "loss": 2.0203,
+      "step": 17200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9389270129180097e-06,
+      "loss": 2.0701,
+      "step": 17201
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.937916672258426e-06,
+      "loss": 2.0842,
+      "step": 17202
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.936906667580863e-06,
+      "loss": 2.0652,
+      "step": 17203
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9358969989118077e-06,
+      "loss": 2.053,
+      "step": 17204
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.934887666277741e-06,
+      "loss": 2.0984,
+      "step": 17205
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9338786697051257e-06,
+      "loss": 2.0767,
+      "step": 17206
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9328700092204245e-06,
+      "loss": 2.0276,
+      "step": 17207
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.93186168485009e-06,
+      "loss": 1.9979,
+      "step": 17208
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.930853696620559e-06,
+      "loss": 2.0674,
+      "step": 17209
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9298460445582735e-06,
+      "loss": 2.0302,
+      "step": 17210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9288387286896446e-06,
+      "loss": 2.0496,
+      "step": 17211
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.927831749041104e-06,
+      "loss": 2.052,
+      "step": 17212
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.926825105639044e-06,
+      "loss": 2.0474,
+      "step": 17213
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9258187985098734e-06,
+      "loss": 2.0211,
+      "step": 17214
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9248128276799764e-06,
+      "loss": 2.1221,
+      "step": 17215
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9238071931757348e-06,
+      "loss": 2.1436,
+      "step": 17216
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9228018950235195e-06,
+      "loss": 2.0718,
+      "step": 17217
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.921796933249693e-06,
+      "loss": 2.114,
+      "step": 17218
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9207923078806087e-06,
+      "loss": 2.0015,
+      "step": 17219
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.919788018942614e-06,
+      "loss": 1.9785,
+      "step": 17220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.918784066462043e-06,
+      "loss": 2.054,
+      "step": 17221
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.917780450465226e-06,
+      "loss": 2.0684,
+      "step": 17222
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9167771709784774e-06,
+      "loss": 2.0781,
+      "step": 17223
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9157742280281124e-06,
+      "loss": 2.0791,
+      "step": 17224
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9147716216404284e-06,
+      "loss": 2.0967,
+      "step": 17225
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9137693518417164e-06,
+      "loss": 2.0828,
+      "step": 17226
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.912767418658265e-06,
+      "loss": 2.0603,
+      "step": 17227
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9117658221163397e-06,
+      "loss": 2.0662,
+      "step": 17228
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.910764562242218e-06,
+      "loss": 2.019,
+      "step": 17229
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.909763639062149e-06,
+      "loss": 2.0483,
+      "step": 17230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.90876305260238e-06,
+      "loss": 2.0552,
+      "step": 17231
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.907762802889155e-06,
+      "loss": 2.0188,
+      "step": 17232
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9067628899486996e-06,
+      "loss": 2.0596,
+      "step": 17233
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9057633138072398e-06,
+      "loss": 2.0583,
+      "step": 17234
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.904764074490989e-06,
+      "loss": 2.0459,
+      "step": 17235
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9037651720261423e-06,
+      "loss": 2.0481,
+      "step": 17236
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9027666064389026e-06,
+      "loss": 2.0637,
+      "step": 17237
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.901768377755456e-06,
+      "loss": 2.0684,
+      "step": 17238
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9007704860019784e-06,
+      "loss": 2.0742,
+      "step": 17239
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8997729312046405e-06,
+      "loss": 2.0801,
+      "step": 17240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.898775713389595e-06,
+      "loss": 2.0652,
+      "step": 17241
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.897778832583e-06,
+      "loss": 2.093,
+      "step": 17242
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.896782288810992e-06,
+      "loss": 2.0173,
+      "step": 17243
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8957860820997123e-06,
+      "loss": 2.0178,
+      "step": 17244
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.894790212475279e-06,
+      "loss": 2.0513,
+      "step": 17245
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8937946799638083e-06,
+      "loss": 2.053,
+      "step": 17246
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.892799484591406e-06,
+      "loss": 2.0874,
+      "step": 17247
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8918046263841706e-06,
+      "loss": 2.1018,
+      "step": 17248
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8908101053681955e-06,
+      "loss": 2.0493,
+      "step": 17249
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.889815921569558e-06,
+      "loss": 2.0823,
+      "step": 17250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.888822075014322e-06,
+      "loss": 2.0559,
+      "step": 17251
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.887828565728563e-06,
+      "loss": 2.0925,
+      "step": 17252
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.886835393738324e-06,
+      "loss": 2.0613,
+      "step": 17253
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.885842559069659e-06,
+      "loss": 2.0718,
+      "step": 17254
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8848500617485956e-06,
+      "loss": 2.0632,
+      "step": 17255
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8838579018011645e-06,
+      "loss": 2.0291,
+      "step": 17256
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8828660792533834e-06,
+      "loss": 2.0562,
+      "step": 17257
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.881874594131263e-06,
+      "loss": 2.0627,
+      "step": 17258
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8808834464608026e-06,
+      "loss": 2.0354,
+      "step": 17259
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.879892636267993e-06,
+      "loss": 2.0344,
+      "step": 17260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8789021635788183e-06,
+      "loss": 2.0237,
+      "step": 17261
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8779120284192546e-06,
+      "loss": 2.0586,
+      "step": 17262
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8769222308152628e-06,
+      "loss": 2.1077,
+      "step": 17263
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.875932770792802e-06,
+      "loss": 2.113,
+      "step": 17264
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8749436483778213e-06,
+      "loss": 2.0493,
+      "step": 17265
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.873954863596252e-06,
+      "loss": 2.0645,
+      "step": 17266
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.872966416474033e-06,
+      "loss": 2.0623,
+      "step": 17267
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8719783070370776e-06,
+      "loss": 2.0593,
+      "step": 17268
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8709905353113043e-06,
+      "loss": 2.0256,
+      "step": 17269
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8700031013226126e-06,
+      "loss": 2.0459,
+      "step": 17270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8690160050968967e-06,
+      "loss": 2.072,
+      "step": 17271
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8680292466600427e-06,
+      "loss": 2.0667,
+      "step": 17272
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.867042826037925e-06,
+      "loss": 2.0601,
+      "step": 17273
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8660567432564162e-06,
+      "loss": 2.0723,
+      "step": 17274
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8650709983413703e-06,
+      "loss": 2.0151,
+      "step": 17275
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8640855913186404e-06,
+      "loss": 1.9888,
+      "step": 17276
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.863100522214067e-06,
+      "loss": 2.0022,
+      "step": 17277
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.862115791053482e-06,
+      "loss": 2.04,
+      "step": 17278
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8611313978627084e-06,
+      "loss": 2.0615,
+      "step": 17279
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.860147342667562e-06,
+      "loss": 2.0527,
+      "step": 17280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8591636254938475e-06,
+      "loss": 2.05,
+      "step": 17281
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.858180246367365e-06,
+      "loss": 2.0786,
+      "step": 17282
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8571972053138937e-06,
+      "loss": 2.0991,
+      "step": 17283
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.856214502359225e-06,
+      "loss": 2.1162,
+      "step": 17284
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8552321375291163e-06,
+      "loss": 2.1467,
+      "step": 17285
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.854250110849342e-06,
+      "loss": 2.147,
+      "step": 17286
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8532684223456444e-06,
+      "loss": 2.0732,
+      "step": 17287
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8522870720437703e-06,
+      "loss": 2.0486,
+      "step": 17288
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8513060599694557e-06,
+      "loss": 2.0647,
+      "step": 17289
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8503253861484265e-06,
+      "loss": 2.0635,
+      "step": 17290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8493450506063965e-06,
+      "loss": 2.0327,
+      "step": 17291
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.848365053369078e-06,
+      "loss": 2.0486,
+      "step": 17292
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.847385394462168e-06,
+      "loss": 2.0532,
+      "step": 17293
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8464060739113578e-06,
+      "loss": 2.041,
+      "step": 17294
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8454270917423287e-06,
+      "loss": 2.1167,
+      "step": 17295
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.844448447980753e-06,
+      "loss": 2.1187,
+      "step": 17296
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8434701426522974e-06,
+      "loss": 2.0618,
+      "step": 17297
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.8424921757826093e-06,
+      "loss": 2.0879,
+      "step": 17298
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.841514547397347e-06,
+      "loss": 2.0974,
+      "step": 17299
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.840537257522133e-06,
+      "loss": 2.1047,
+      "step": 17300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.839560306182608e-06,
+      "loss": 2.0149,
+      "step": 17301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8385836934043857e-06,
+      "loss": 1.9952,
+      "step": 17302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8376074192130737e-06,
+      "loss": 2.0938,
+      "step": 17303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.836631483634281e-06,
+      "loss": 2.0989,
+      "step": 17304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8356558866935943e-06,
+      "loss": 2.0105,
+      "step": 17305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8346806284166015e-06,
+      "loss": 1.9685,
+      "step": 17306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8337057088288756e-06,
+      "loss": 2.0681,
+      "step": 17307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.832731127955982e-06,
+      "loss": 2.0669,
+      "step": 17308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.831756885823481e-06,
+      "loss": 2.0667,
+      "step": 17309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.830782982456917e-06,
+      "loss": 2.0403,
+      "step": 17310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8298094178818313e-06,
+      "loss": 2.0806,
+      "step": 17311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.828836192123758e-06,
+      "loss": 2.0842,
+      "step": 17312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8278633052082107e-06,
+      "loss": 2.0852,
+      "step": 17313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.826890757160712e-06,
+      "loss": 2.1016,
+      "step": 17314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8259185480067566e-06,
+      "loss": 2.0532,
+      "step": 17315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.824946677771849e-06,
+      "loss": 2.032,
+      "step": 17316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8239751464814673e-06,
+      "loss": 2.0872,
+      "step": 17317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8230039541610926e-06,
+      "loss": 2.0479,
+      "step": 17318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8220331008361893e-06,
+      "loss": 2.072,
+      "step": 17319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8210625865322226e-06,
+      "loss": 2.0764,
+      "step": 17320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8200924112746403e-06,
+      "loss": 2.083,
+      "step": 17321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8191225750888865e-06,
+      "loss": 2.0789,
+      "step": 17322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8181530780003865e-06,
+      "loss": 2.03,
+      "step": 17323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.817183920034575e-06,
+      "loss": 2.0181,
+      "step": 17324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8162151012168564e-06,
+      "loss": 2.126,
+      "step": 17325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8152466215726487e-06,
+      "loss": 2.1189,
+      "step": 17326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.814278481127338e-06,
+      "loss": 1.9871,
+      "step": 17327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.813310679906318e-06,
+      "loss": 2.0264,
+      "step": 17328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.812343217934966e-06,
+      "loss": 2.0742,
+      "step": 17329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.811376095238656e-06,
+      "loss": 2.0889,
+      "step": 17330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8104093118427463e-06,
+      "loss": 2.0781,
+      "step": 17331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.809442867772593e-06,
+      "loss": 2.0735,
+      "step": 17332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.808476763053535e-06,
+      "loss": 2.0833,
+      "step": 17333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8075109977109124e-06,
+      "loss": 2.094,
+      "step": 17334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.806545571770048e-06,
+      "loss": 2.0847,
+      "step": 17335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8055804852562606e-06,
+      "loss": 2.0759,
+      "step": 17336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8046157381948624e-06,
+      "loss": 2.1035,
+      "step": 17337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8036513306111417e-06,
+      "loss": 2.0869,
+      "step": 17338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8026872625303985e-06,
+      "loss": 2.071,
+      "step": 17339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.801723533977909e-06,
+      "loss": 2.0837,
+      "step": 17340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.8007601449789534e-06,
+      "loss": 2.1138,
+      "step": 17341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7997970955587876e-06,
+      "loss": 2.1123,
+      "step": 17342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7988343857426686e-06,
+      "loss": 2.0742,
+      "step": 17343
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7978720155558422e-06,
+      "loss": 2.0579,
+      "step": 17344
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.796909985023549e-06,
+      "loss": 2.0581,
+      "step": 17345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.795948294171012e-06,
+      "loss": 2.0835,
+      "step": 17346
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7949869430234536e-06,
+      "loss": 2.0771,
+      "step": 17347
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.794025931606083e-06,
+      "loss": 2.0637,
+      "step": 17348
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.793065259944101e-06,
+      "loss": 2.0344,
+      "step": 17349
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7921049280627025e-06,
+      "loss": 2.0281,
+      "step": 17350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.791144935987071e-06,
+      "loss": 2.0737,
+      "step": 17351
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7901852837423796e-06,
+      "loss": 2.074,
+      "step": 17352
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7892259713537898e-06,
+      "loss": 2.0151,
+      "step": 17353
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7882669988464697e-06,
+      "loss": 2.0413,
+      "step": 17354
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7873083662455528e-06,
+      "loss": 2.105,
+      "step": 17355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7863500735761922e-06,
+      "loss": 2.0884,
+      "step": 17356
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7853921208635093e-06,
+      "loss": 2.0166,
+      "step": 17357
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.784434508132629e-06,
+      "loss": 1.9973,
+      "step": 17358
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7834772354086597e-06,
+      "loss": 2.0483,
+      "step": 17359
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7825203027167095e-06,
+      "loss": 2.0488,
+      "step": 17360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7815637100818687e-06,
+      "loss": 2.0481,
+      "step": 17361
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.780607457529225e-06,
+      "loss": 2.0457,
+      "step": 17362
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7796515450838556e-06,
+      "loss": 2.1111,
+      "step": 17363
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7786959727708266e-06,
+      "loss": 2.0994,
+      "step": 17364
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7777407406151967e-06,
+      "loss": 2.0901,
+      "step": 17365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7767858486420172e-06,
+      "loss": 2.0791,
+      "step": 17366
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7758312968763316e-06,
+      "loss": 2.0383,
+      "step": 17367
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.774877085343165e-06,
+      "loss": 2.0361,
+      "step": 17368
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.773923214067547e-06,
+      "loss": 2.0874,
+      "step": 17369
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7729696830744845e-06,
+      "loss": 2.0979,
+      "step": 17370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7720164923889923e-06,
+      "loss": 2.0073,
+      "step": 17371
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.771063642036059e-06,
+      "loss": 2.0613,
+      "step": 17372
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7701111320406763e-06,
+      "loss": 2.0613,
+      "step": 17373
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.769158962427819e-06,
+      "loss": 2.0654,
+      "step": 17374
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.768207133222459e-06,
+      "loss": 2.0488,
+      "step": 17375
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.767255644449558e-06,
+      "loss": 2.0676,
+      "step": 17376
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.766304496134064e-06,
+      "loss": 2.0098,
+      "step": 17377
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7653536883009224e-06,
+      "loss": 2.011,
+      "step": 17378
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.764403220975068e-06,
+      "loss": 2.0796,
+      "step": 17379
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.763453094181423e-06,
+      "loss": 2.0769,
+      "step": 17380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.762503307944905e-06,
+      "loss": 2.0349,
+      "step": 17381
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7615538622904252e-06,
+      "loss": 2.0361,
+      "step": 17382
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.760604757242871e-06,
+      "loss": 2.0396,
+      "step": 17383
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7596559928271413e-06,
+      "loss": 2.0305,
+      "step": 17384
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7587075690681097e-06,
+      "loss": 2.0864,
+      "step": 17385
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.757759485990656e-06,
+      "loss": 2.0811,
+      "step": 17386
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7568117436196333e-06,
+      "loss": 2.0906,
+      "step": 17387
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7558643419799005e-06,
+      "loss": 2.0886,
+      "step": 17388
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7549172810962996e-06,
+      "loss": 2.0298,
+      "step": 17389
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7539705609936672e-06,
+      "loss": 2.0225,
+      "step": 17390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7530241816968317e-06,
+      "loss": 2.0398,
+      "step": 17391
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7520781432306097e-06,
+      "loss": 2.0447,
+      "step": 17392
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.751132445619805e-06,
+      "loss": 2.0764,
+      "step": 17393
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.750187088889226e-06,
+      "loss": 2.0436,
+      "step": 17394
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7492420730636566e-06,
+      "loss": 2.0884,
+      "step": 17395
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7482973981678846e-06,
+      "loss": 2.0867,
+      "step": 17396
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.747353064226674e-06,
+      "loss": 2.0352,
+      "step": 17397
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.746409071264801e-06,
+      "loss": 2.0273,
+      "step": 17398
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7454654193070116e-06,
+      "loss": 2.1255,
+      "step": 17399
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7445221083780557e-06,
+      "loss": 2.1411,
+      "step": 17400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7435791385026685e-06,
+      "loss": 2.0791,
+      "step": 17401
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742636509705577e-06,
+      "loss": 2.1387,
+      "step": 17402
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.74169422201151e-06,
+      "loss": 2.1104,
+      "step": 17403
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7407522754451652e-06,
+      "loss": 2.1001,
+      "step": 17404
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.739810670031252e-06,
+      "loss": 2.0847,
+      "step": 17405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.73886940579446e-06,
+      "loss": 2.0938,
+      "step": 17406
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7379284827594737e-06,
+      "loss": 2.0955,
+      "step": 17407
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7369879009509672e-06,
+      "loss": 2.0957,
+      "step": 17408
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.73604766039361e-06,
+      "loss": 2.0747,
+      "step": 17409
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.735107761112049e-06,
+      "loss": 2.1128,
+      "step": 17410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7341682031309424e-06,
+      "loss": 2.0737,
+      "step": 17411
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7332289864749223e-06,
+      "loss": 2.0671,
+      "step": 17412
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.732290111168624e-06,
+      "loss": 1.98,
+      "step": 17413
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.731351577236664e-06,
+      "loss": 2.0044,
+      "step": 17414
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7304133847036557e-06,
+      "loss": 2.0989,
+      "step": 17415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.729475533594203e-06,
+      "loss": 2.0793,
+      "step": 17416
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.728538023932898e-06,
+      "loss": 2.0776,
+      "step": 17417
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7276008557443253e-06,
+      "loss": 2.0928,
+      "step": 17418
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7266640290530666e-06,
+      "loss": 2.0984,
+      "step": 17419
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7257275438836816e-06,
+      "loss": 2.104,
+      "step": 17420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7247914002607327e-06,
+      "loss": 2.074,
+      "step": 17421
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.723855598208772e-06,
+      "loss": 2.0625,
+      "step": 17422
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7229201377523333e-06,
+      "loss": 2.114,
+      "step": 17423
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7219850189159535e-06,
+      "loss": 2.0964,
+      "step": 17424
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.721050241724148e-06,
+      "loss": 2.0811,
+      "step": 17425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7201158062014418e-06,
+      "loss": 2.1382,
+      "step": 17426
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7191817123723247e-06,
+      "loss": 2.0688,
+      "step": 17427
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.718247960261306e-06,
+      "loss": 2.0496,
+      "step": 17428
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.717314549892863e-06,
+      "loss": 2.0845,
+      "step": 17429
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7163814812914784e-06,
+      "loss": 2.1084,
+      "step": 17430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.715448754481617e-06,
+      "loss": 2.0942,
+      "step": 17431
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.714516369487741e-06,
+      "loss": 2.0608,
+      "step": 17432
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7135843263343013e-06,
+      "loss": 2.0632,
+      "step": 17433
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7126526250457396e-06,
+      "loss": 2.0659,
+      "step": 17434
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7117212656464862e-06,
+      "loss": 2.0991,
+      "step": 17435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.710790248160967e-06,
+      "loss": 2.1138,
+      "step": 17436
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7098595726135957e-06,
+      "loss": 2.0935,
+      "step": 17437
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7089292390287825e-06,
+      "loss": 2.062,
+      "step": 17438
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7079992474309196e-06,
+      "loss": 2.0244,
+      "step": 17439
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7070695978443945e-06,
+      "loss": 2.0654,
+      "step": 17440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7061402902935906e-06,
+      "loss": 2.0017,
+      "step": 17441
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7052113248028703e-06,
+      "loss": 2.0098,
+      "step": 17442
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7042827013966057e-06,
+      "loss": 2.0154,
+      "step": 17443
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7033544200991377e-06,
+      "loss": 2.0322,
+      "step": 17444
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7024264809348183e-06,
+      "loss": 2.0764,
+      "step": 17445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.701498883927976e-06,
+      "loss": 2.0686,
+      "step": 17446
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7005716291029364e-06,
+      "loss": 2.0784,
+      "step": 17447
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6996447164840187e-06,
+      "loss": 2.0989,
+      "step": 17448
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6987181460955284e-06,
+      "loss": 2.094,
+      "step": 17449
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6977919179617644e-06,
+      "loss": 2.095,
+      "step": 17450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.696866032107012e-06,
+      "loss": 2.0715,
+      "step": 17451
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6959404885555573e-06,
+      "loss": 2.0276,
+      "step": 17452
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6950152873316686e-06,
+      "loss": 2.0386,
+      "step": 17453
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6940904284596117e-06,
+      "loss": 2.0759,
+      "step": 17454
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6931659119636325e-06,
+      "loss": 2.0925,
+      "step": 17455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.692241737867983e-06,
+      "loss": 2.094,
+      "step": 17456
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6913179061968906e-06,
+      "loss": 2.0486,
+      "step": 17457
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.690394416974594e-06,
+      "loss": 2.0198,
+      "step": 17458
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6894712702252993e-06,
+      "loss": 2.0862,
+      "step": 17459
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.688548465973217e-06,
+      "loss": 2.0747,
+      "step": 17460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.687626004242552e-06,
+      "loss": 2.082,
+      "step": 17461
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6867038850574878e-06,
+      "loss": 2.0916,
+      "step": 17462
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6857821084422106e-06,
+      "loss": 2.0701,
+      "step": 17463
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6848606744208955e-06,
+      "loss": 2.0371,
+      "step": 17464
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6839395830176953e-06,
+      "loss": 2.0259,
+      "step": 17465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6830188342567767e-06,
+      "loss": 2.0618,
+      "step": 17466
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6820984281622766e-06,
+      "loss": 2.0337,
+      "step": 17467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6811783647583377e-06,
+      "loss": 2.0291,
+      "step": 17468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6802586440690835e-06,
+      "loss": 2.1108,
+      "step": 17469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.679339266118633e-06,
+      "loss": 2.1289,
+      "step": 17470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.678420230931098e-06,
+      "loss": 2.0212,
+      "step": 17471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6775015385305747e-06,
+      "loss": 2.0156,
+      "step": 17472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.676583188941164e-06,
+      "loss": 2.04,
+      "step": 17473
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6756651821869387e-06,
+      "loss": 2.0474,
+      "step": 17474
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6747475182919763e-06,
+      "loss": 2.0305,
+      "step": 17475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.673830197280342e-06,
+      "loss": 2.05,
+      "step": 17476
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.67291321917609e-06,
+      "loss": 2.0793,
+      "step": 17477
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6719965840032703e-06,
+      "loss": 2.0974,
+      "step": 17478
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.671080291785919e-06,
+      "loss": 2.0579,
+      "step": 17479
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6701643425480594e-06,
+      "loss": 2.074,
+      "step": 17480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6692487363137202e-06,
+      "loss": 2.0415,
+      "step": 17481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6683334731069025e-06,
+      "loss": 2.0618,
+      "step": 17482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6674185529516204e-06,
+      "loss": 2.0671,
+      "step": 17483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6665039758718547e-06,
+      "loss": 2.0818,
+      "step": 17484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.665589741891596e-06,
+      "loss": 2.094,
+      "step": 17485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.664675851034815e-06,
+      "loss": 2.0801,
+      "step": 17486
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.66376230332548e-06,
+      "loss": 2.0581,
+      "step": 17487
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6628490987875477e-06,
+      "loss": 2.0264,
+      "step": 17488
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.661936237444964e-06,
+      "loss": 2.0095,
+      "step": 17489
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6610237193216703e-06,
+      "loss": 2.0259,
+      "step": 17490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.660111544441592e-06,
+      "loss": 2.0696,
+      "step": 17491
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6591997128286567e-06,
+      "loss": 2.0745,
+      "step": 17492
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6582882245067685e-06,
+      "loss": 2.0652,
+      "step": 17493
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6573770794998385e-06,
+      "loss": 2.0374,
+      "step": 17494
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6564662778317494e-06,
+      "loss": 2.0234,
+      "step": 17495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6555558195263966e-06,
+      "loss": 2.0415,
+      "step": 17496
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6546457046076485e-06,
+      "loss": 2.043,
+      "step": 17497
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.6537359330993797e-06,
+      "loss": 2.0894,
+      "step": 17498
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.652826505025439e-06,
+      "loss": 2.1067,
+      "step": 17499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6519174204096786e-06,
+      "loss": 2.0769,
+      "step": 17500
+    },
+    {
+      "Bleu_1": 0.7725231161731716,
+      "Bleu_2": 0.6114854053735045,
+      "Bleu_3": 0.4692275465286584,
+      "Bleu_4": 0.355617628280263,
+      "CIDEr": 1.2011755971178977,
+      "METEOR": 0.2775578506824468,
+      "ROUGE_L": 0.5590541802405508,
+      "epoch": 0.88,
+      "step": 17500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.651008679275939e-06,
+      "loss": 2.0513,
+      "step": 17501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6501002816480517e-06,
+      "loss": 2.0728,
+      "step": 17502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6491922275498343e-06,
+      "loss": 2.0066,
+      "step": 17503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6482845170051057e-06,
+      "loss": 2.0061,
+      "step": 17504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6473771500376635e-06,
+      "loss": 2.0325,
+      "step": 17505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6464701266713077e-06,
+      "loss": 2.0527,
+      "step": 17506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6455634469298184e-06,
+      "loss": 2.0684,
+      "step": 17507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6446571108369773e-06,
+      "loss": 2.1089,
+      "step": 17508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6437511184165512e-06,
+      "loss": 2.113,
+      "step": 17509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6428454696922923e-06,
+      "loss": 2.1069,
+      "step": 17510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6419401646879602e-06,
+      "loss": 2.0867,
+      "step": 17511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.641035203427287e-06,
+      "loss": 2.0857,
+      "step": 17512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6401305859340103e-06,
+      "loss": 2.0879,
+      "step": 17513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.639226312231849e-06,
+      "loss": 2.0625,
+      "step": 17514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6383223823445174e-06,
+      "loss": 2.053,
+      "step": 17515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6374187962957203e-06,
+      "loss": 2.0557,
+      "step": 17516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6365155541091516e-06,
+      "loss": 2.0659,
+      "step": 17517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6356126558085015e-06,
+      "loss": 2.0601,
+      "step": 17518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6347101014174433e-06,
+      "loss": 2.0911,
+      "step": 17519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.633807890959649e-06,
+      "loss": 2.0974,
+      "step": 17520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.632906024458776e-06,
+      "loss": 2.0342,
+      "step": 17521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6320045019384753e-06,
+      "loss": 2.0142,
+      "step": 17522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.631103323422388e-06,
+      "loss": 2.0627,
+      "step": 17523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.630202488934147e-06,
+      "loss": 2.0564,
+      "step": 17524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.629301998497375e-06,
+      "loss": 2.0879,
+      "step": 17525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6284018521356913e-06,
+      "loss": 2.0393,
+      "step": 17526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6275020498726887e-06,
+      "loss": 1.9905,
+      "step": 17527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.626602591731975e-06,
+      "loss": 2.0127,
+      "step": 17528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.625703477737132e-06,
+      "loss": 2.1211,
+      "step": 17529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6248047079117417e-06,
+      "loss": 2.1233,
+      "step": 17530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.623906282279372e-06,
+      "loss": 2.0459,
+      "step": 17531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.623008200863581e-06,
+      "loss": 2.0764,
+      "step": 17532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6221104636879197e-06,
+      "loss": 2.0227,
+      "step": 17533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6212130707759323e-06,
+      "loss": 1.991,
+      "step": 17534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.620316022151152e-06,
+      "loss": 2.0657,
+      "step": 17535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.619419317837104e-06,
+      "loss": 2.0442,
+      "step": 17536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.618522957857295e-06,
+      "loss": 2.061,
+      "step": 17537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.617626942235245e-06,
+      "loss": 2.0522,
+      "step": 17538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.616731270994437e-06,
+      "loss": 2.125,
+      "step": 17539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.615835944158372e-06,
+      "loss": 2.1523,
+      "step": 17540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6149409617505178e-06,
+      "loss": 2.1101,
+      "step": 17541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6140463237943474e-06,
+      "loss": 2.0679,
+      "step": 17542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6131520303133273e-06,
+      "loss": 2.022,
+      "step": 17543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6122580813309e-06,
+      "loss": 2.0107,
+      "step": 17544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.611364476870517e-06,
+      "loss": 2.0588,
+      "step": 17545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6104712169556078e-06,
+      "loss": 2.0371,
+      "step": 17546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.609578301609595e-06,
+      "loss": 2.0674,
+      "step": 17547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6086857308559008e-06,
+      "loss": 2.0801,
+      "step": 17548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.607793504717924e-06,
+      "loss": 2.0232,
+      "step": 17549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6069016232190694e-06,
+      "loss": 1.9901,
+      "step": 17550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6060100863827234e-06,
+      "loss": 2.0625,
+      "step": 17551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.605118894232259e-06,
+      "loss": 2.0642,
+      "step": 17552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.604228046791056e-06,
+      "loss": 2.0525,
+      "step": 17553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.603337544082468e-06,
+      "loss": 2.0598,
+      "step": 17554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6024473861298574e-06,
+      "loss": 2.0278,
+      "step": 17555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.601557572956559e-06,
+      "loss": 2.0608,
+      "step": 17556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.6006681045859065e-06,
+      "loss": 2.103,
+      "step": 17557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5997789810412306e-06,
+      "loss": 2.1013,
+      "step": 17558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.598890202345845e-06,
+      "loss": 2.0662,
+      "step": 17559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.598001768523057e-06,
+      "loss": 2.0344,
+      "step": 17560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5971136795961653e-06,
+      "loss": 2.0964,
+      "step": 17561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.596225935588457e-06,
+      "loss": 2.0935,
+      "step": 17562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.595338536523215e-06,
+      "loss": 2.1047,
+      "step": 17563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5944514824237073e-06,
+      "loss": 2.0908,
+      "step": 17564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.593564773313199e-06,
+      "loss": 2.0596,
+      "step": 17565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.592678409214944e-06,
+      "loss": 2.0732,
+      "step": 17566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5917923901521797e-06,
+      "loss": 2.0613,
+      "step": 17567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.590906716148148e-06,
+      "loss": 2.0684,
+      "step": 17568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5900213872260663e-06,
+      "loss": 2.0779,
+      "step": 17569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5891364034091643e-06,
+      "loss": 2.072,
+      "step": 17570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5882517647206376e-06,
+      "loss": 2.0632,
+      "step": 17571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5873674711836887e-06,
+      "loss": 2.0898,
+      "step": 17572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5864835228215083e-06,
+      "loss": 2.0852,
+      "step": 17573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.585599919657278e-06,
+      "loss": 2.1033,
+      "step": 17574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.584716661714165e-06,
+      "loss": 2.0422,
+      "step": 17575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5838337490153364e-06,
+      "loss": 2.0225,
+      "step": 17576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5829511815839423e-06,
+      "loss": 1.9902,
+      "step": 17577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5820689594431295e-06,
+      "loss": 2.0354,
+      "step": 17578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5811870826160296e-06,
+      "loss": 2.0835,
+      "step": 17579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5803055511257734e-06,
+      "loss": 2.0654,
+      "step": 17580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5794243649954764e-06,
+      "loss": 2.0789,
+      "step": 17581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.578543524248243e-06,
+      "loss": 2.0752,
+      "step": 17582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5776630289071783e-06,
+      "loss": 2.0906,
+      "step": 17583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5767828789953658e-06,
+      "loss": 2.0869,
+      "step": 17584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.575903074535893e-06,
+      "loss": 2.094,
+      "step": 17585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5750236155518285e-06,
+      "loss": 2.0959,
+      "step": 17586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.574144502066234e-06,
+      "loss": 2.0979,
+      "step": 17587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5732657341021645e-06,
+      "loss": 2.0771,
+      "step": 17588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5723873116826657e-06,
+      "loss": 2.1016,
+      "step": 17589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.57150923483077e-06,
+      "loss": 2.0508,
+      "step": 17590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.570631503569509e-06,
+      "loss": 2.0515,
+      "step": 17591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5697541179218936e-06,
+      "loss": 2.0691,
+      "step": 17592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5688770779109376e-06,
+      "loss": 2.1013,
+      "step": 17593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5680003835596394e-06,
+      "loss": 2.1003,
+      "step": 17594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.567124034890987e-06,
+      "loss": 2.1008,
+      "step": 17595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.566248031927967e-06,
+      "loss": 2.0818,
+      "step": 17596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5653723746935425e-06,
+      "loss": 2.0203,
+      "step": 17597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5644970632106874e-06,
+      "loss": 2.0486,
+      "step": 17598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.563622097502343e-06,
+      "loss": 2.0757,
+      "step": 17599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5627474775914672e-06,
+      "loss": 2.0815,
+      "step": 17600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.561873203500989e-06,
+      "loss": 2.1499,
+      "step": 17601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5609992752538347e-06,
+      "loss": 2.1196,
+      "step": 17602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5601256928729245e-06,
+      "loss": 2.0457,
+      "step": 17603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5592524563811666e-06,
+      "loss": 2.0386,
+      "step": 17604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.558379565801459e-06,
+      "loss": 2.0503,
+      "step": 17605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.557507021156696e-06,
+      "loss": 2.0901,
+      "step": 17606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.556634822469753e-06,
+      "loss": 2.0342,
+      "step": 17607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5557629697635117e-06,
+      "loss": 2.0674,
+      "step": 17608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5548914630608224e-06,
+      "loss": 2.0122,
+      "step": 17609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5540203023845506e-06,
+      "loss": 2.0193,
+      "step": 17610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5531494877575378e-06,
+      "loss": 2.0564,
+      "step": 17611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.552279019202618e-06,
+      "loss": 2.0645,
+      "step": 17612
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5514088967426225e-06,
+      "loss": 2.0381,
+      "step": 17613
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5505391204003633e-06,
+      "loss": 2.0466,
+      "step": 17614
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5496696901986566e-06,
+      "loss": 2.0437,
+      "step": 17615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5488006061602963e-06,
+      "loss": 2.0479,
+      "step": 17616
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.547931868308076e-06,
+      "loss": 2.072,
+      "step": 17617
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.547063476664777e-06,
+      "loss": 2.0903,
+      "step": 17618
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.54619543125317e-06,
+      "loss": 2.0532,
+      "step": 17619
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.545327732096021e-06,
+      "loss": 2.0178,
+      "step": 17620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.544460379216086e-06,
+      "loss": 2.0525,
+      "step": 17621
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.543593372636101e-06,
+      "loss": 2.0715,
+      "step": 17622
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.542726712378815e-06,
+      "loss": 2.1089,
+      "step": 17623
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.541860398466943e-06,
+      "loss": 2.071,
+      "step": 17624
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5409944309232136e-06,
+      "loss": 2.0813,
+      "step": 17625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.540128809770332e-06,
+      "loss": 2.0986,
+      "step": 17626
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5392635350309946e-06,
+      "loss": 2.075,
+      "step": 17627
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.538398606727895e-06,
+      "loss": 2.0637,
+      "step": 17628
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5375340248837175e-06,
+      "loss": 2.0791,
+      "step": 17629
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.53666978952113e-06,
+      "loss": 2.0879,
+      "step": 17630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5358059006627975e-06,
+      "loss": 2.0132,
+      "step": 17631
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5349423583313773e-06,
+      "loss": 2.0039,
+      "step": 17632
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.534079162549513e-06,
+      "loss": 2.0745,
+      "step": 17633
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5332163133398393e-06,
+      "loss": 2.0481,
+      "step": 17634
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5323538107249857e-06,
+      "loss": 2.0735,
+      "step": 17635
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.531491654727568e-06,
+      "loss": 2.0718,
+      "step": 17636
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.530629845370197e-06,
+      "loss": 2.1162,
+      "step": 17637
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5297683826754764e-06,
+      "loss": 2.135,
+      "step": 17638
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.528907266665988e-06,
+      "loss": 2.0476,
+      "step": 17639
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5280464973643218e-06,
+      "loss": 2.0513,
+      "step": 17640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5271860747930437e-06,
+      "loss": 2.0435,
+      "step": 17641
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.526325998974725e-06,
+      "loss": 2.0312,
+      "step": 17642
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5254662699319147e-06,
+      "loss": 2.0737,
+      "step": 17643
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5246068876871564e-06,
+      "loss": 2.0793,
+      "step": 17644
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5237478522629936e-06,
+      "loss": 2.0891,
+      "step": 17645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.522889163681947e-06,
+      "loss": 2.1025,
+      "step": 17646
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.522030821966538e-06,
+      "loss": 2.1006,
+      "step": 17647
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5211728271392737e-06,
+      "loss": 2.0667,
+      "step": 17648
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5203151792226565e-06,
+      "loss": 2.0811,
+      "step": 17649
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5194578782391746e-06,
+      "loss": 2.0793,
+      "step": 17650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5186009242113105e-06,
+      "loss": 2.0295,
+      "step": 17651
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5177443171615395e-06,
+      "loss": 2.0308,
+      "step": 17652
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.516888057112323e-06,
+      "loss": 2.0662,
+      "step": 17653
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.516032144086112e-06,
+      "loss": 2.0608,
+      "step": 17654
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5151765781053586e-06,
+      "loss": 2.0337,
+      "step": 17655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.514321359192491e-06,
+      "loss": 2.033,
+      "step": 17656
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5134664873699453e-06,
+      "loss": 2.0435,
+      "step": 17657
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5126119626601346e-06,
+      "loss": 2.0637,
+      "step": 17658
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5117577850854676e-06,
+      "loss": 1.9937,
+      "step": 17659
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.510903954668344e-06,
+      "loss": 2.0178,
+      "step": 17660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.510050471431156e-06,
+      "loss": 2.0923,
+      "step": 17661
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.509197335396282e-06,
+      "loss": 2.0439,
+      "step": 17662
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5083445465861004e-06,
+      "loss": 2.1111,
+      "step": 17663
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.50749210502297e-06,
+      "loss": 2.1042,
+      "step": 17664
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5066400107292458e-06,
+      "loss": 2.1079,
+      "step": 17665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.505788263727274e-06,
+      "loss": 2.0981,
+      "step": 17666
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.504936864039391e-06,
+      "loss": 2.0737,
+      "step": 17667
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5040858116879243e-06,
+      "loss": 2.0623,
+      "step": 17668
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5032351066951863e-06,
+      "loss": 2.0764,
+      "step": 17669
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5023847490834937e-06,
+      "loss": 2.0698,
+      "step": 17670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5015347388751393e-06,
+      "loss": 2.1057,
+      "step": 17671
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.5006850760924195e-06,
+      "loss": 2.0757,
+      "step": 17672
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4998357607576116e-06,
+      "loss": 2.0889,
+      "step": 17673
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4989867928929884e-06,
+      "loss": 2.0759,
+      "step": 17674
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4981381725208146e-06,
+      "loss": 2.0972,
+      "step": 17675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.497289899663343e-06,
+      "loss": 2.0955,
+      "step": 17676
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.496441974342821e-06,
+      "loss": 2.0261,
+      "step": 17677
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4955943965814816e-06,
+      "loss": 2.0476,
+      "step": 17678
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4947471664015493e-06,
+      "loss": 2.0137,
+      "step": 17679
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.493900283825248e-06,
+      "loss": 2.0559,
+      "step": 17680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.493053748874783e-06,
+      "loss": 2.0181,
+      "step": 17681
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.492207561572353e-06,
+      "loss": 2.0322,
+      "step": 17682
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.491361721940153e-06,
+      "loss": 2.0425,
+      "step": 17683
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.490516230000356e-06,
+      "loss": 2.0671,
+      "step": 17684
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4896710857751394e-06,
+      "loss": 2.0674,
+      "step": 17685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4888262892866637e-06,
+      "loss": 2.1079,
+      "step": 17686
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.487981840557088e-06,
+      "loss": 2.0457,
+      "step": 17687
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4871377396085513e-06,
+      "loss": 2.0835,
+      "step": 17688
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4862939864631896e-06,
+      "loss": 2.083,
+      "step": 17689
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.485450581143131e-06,
+      "loss": 2.0806,
+      "step": 17690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4846075236704937e-06,
+      "loss": 2.0791,
+      "step": 17691
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4837648140673826e-06,
+      "loss": 2.0491,
+      "step": 17692
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4829224523559034e-06,
+      "loss": 2.0647,
+      "step": 17693
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.482080438558134e-06,
+      "loss": 2.0562,
+      "step": 17694
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4812387726961714e-06,
+      "loss": 2.0552,
+      "step": 17695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4803974547920714e-06,
+      "loss": 2.0393,
+      "step": 17696
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.47955648486791e-06,
+      "loss": 2.0916,
+      "step": 17697
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4787158629457303e-06,
+      "loss": 2.1003,
+      "step": 17698
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.4778755890475837e-06,
+      "loss": 2.0505,
+      "step": 17699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4770356631954986e-06,
+      "loss": 2.0591,
+      "step": 17700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4761960854115083e-06,
+      "loss": 2.0579,
+      "step": 17701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4753568557176266e-06,
+      "loss": 2.0278,
+      "step": 17702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4745179741358585e-06,
+      "loss": 2.0657,
+      "step": 17703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.473679440688209e-06,
+      "loss": 2.0684,
+      "step": 17704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.472841255396662e-06,
+      "loss": 2.1125,
+      "step": 17705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.472003418283201e-06,
+      "loss": 2.0901,
+      "step": 17706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4711659293697952e-06,
+      "loss": 2.0889,
+      "step": 17707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.470328788678411e-06,
+      "loss": 2.0935,
+      "step": 17708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4694919962309935e-06,
+      "loss": 2.0317,
+      "step": 17709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4686555520494975e-06,
+      "loss": 2.0393,
+      "step": 17710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.467819456155845e-06,
+      "loss": 2.0315,
+      "step": 17711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4669837085719747e-06,
+      "loss": 2.0505,
+      "step": 17712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4661483093197933e-06,
+      "loss": 2.0176,
+      "step": 17713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4653132584212112e-06,
+      "loss": 2.0342,
+      "step": 17714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.464478555898129e-06,
+      "loss": 2.0864,
+      "step": 17715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4636442017724332e-06,
+      "loss": 2.1096,
+      "step": 17716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.462810196066004e-06,
+      "loss": 2.0664,
+      "step": 17717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.461976538800714e-06,
+      "loss": 2.0564,
+      "step": 17718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.461143229998422e-06,
+      "loss": 2.0564,
+      "step": 17719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4603102696809847e-06,
+      "loss": 2.0459,
+      "step": 17720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.459477657870241e-06,
+      "loss": 2.0425,
+      "step": 17721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4586453945880266e-06,
+      "loss": 2.063,
+      "step": 17722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4578134798561706e-06,
+      "loss": 2.0117,
+      "step": 17723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.456981913696481e-06,
+      "loss": 2.0337,
+      "step": 17724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.456150696130773e-06,
+      "loss": 2.1067,
+      "step": 17725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4553198271808386e-06,
+      "loss": 2.0735,
+      "step": 17726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.454489306868471e-06,
+      "loss": 2.1033,
+      "step": 17727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4536591352154445e-06,
+      "loss": 2.0784,
+      "step": 17728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4528293122435326e-06,
+      "loss": 2.0317,
+      "step": 17729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4519998379744967e-06,
+      "loss": 2.0576,
+      "step": 17730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.451170712430085e-06,
+      "loss": 2.0798,
+      "step": 17731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4503419356320457e-06,
+      "loss": 2.0869,
+      "step": 17732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4495135076021087e-06,
+      "loss": 2.0945,
+      "step": 17733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4486854283620003e-06,
+      "loss": 2.116,
+      "step": 17734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.447857697933435e-06,
+      "loss": 2.0486,
+      "step": 17735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.447030316338122e-06,
+      "loss": 2.0881,
+      "step": 17736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.446203283597754e-06,
+      "loss": 2.074,
+      "step": 17737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4453765997340233e-06,
+      "loss": 2.0972,
+      "step": 17738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4445502647686014e-06,
+      "loss": 2.0125,
+      "step": 17739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.443724278723169e-06,
+      "loss": 2.0566,
+      "step": 17740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4428986416193738e-06,
+      "loss": 2.0894,
+      "step": 17741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4420733534788803e-06,
+      "loss": 2.0842,
+      "step": 17742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4412484143233227e-06,
+      "loss": 2.1013,
+      "step": 17743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4404238241743338e-06,
+      "loss": 2.1113,
+      "step": 17744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.439599583053541e-06,
+      "loss": 2.071,
+      "step": 17745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.438775690982556e-06,
+      "loss": 2.0996,
+      "step": 17746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4379521479829876e-06,
+      "loss": 2.0269,
+      "step": 17747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4371289540764317e-06,
+      "loss": 2.0601,
+      "step": 17748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.43630610928447e-06,
+      "loss": 2.0947,
+      "step": 17749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4354836136286886e-06,
+      "loss": 2.0652,
+      "step": 17750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4346614671306525e-06,
+      "loss": 2.053,
+      "step": 17751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4338396698119227e-06,
+      "loss": 2.0637,
+      "step": 17752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4330182216940505e-06,
+      "loss": 2.0747,
+      "step": 17753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4321971227985714e-06,
+      "loss": 2.1003,
+      "step": 17754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.431376373147029e-06,
+      "loss": 2.0601,
+      "step": 17755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.430555972760935e-06,
+      "loss": 2.0903,
+      "step": 17756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4297359216618105e-06,
+      "loss": 2.0815,
+      "step": 17757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4289162198711563e-06,
+      "loss": 2.0542,
+      "step": 17758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4280968674104737e-06,
+      "loss": 2.0356,
+      "step": 17759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4272778643012454e-06,
+      "loss": 2.0469,
+      "step": 17760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.426459210564948e-06,
+      "loss": 2.0515,
+      "step": 17761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4256409062230505e-06,
+      "loss": 2.04,
+      "step": 17762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.424822951297014e-06,
+      "loss": 2.0737,
+      "step": 17763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4240053458082857e-06,
+      "loss": 2.0991,
+      "step": 17764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4231880897783088e-06,
+      "loss": 2.1211,
+      "step": 17765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4223711832285118e-06,
+      "loss": 2.1221,
+      "step": 17766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4215546261803225e-06,
+      "loss": 2.0183,
+      "step": 17767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4207384186551433e-06,
+      "loss": 2.0234,
+      "step": 17768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.419922560674393e-06,
+      "loss": 2.0386,
+      "step": 17769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4191070522594572e-06,
+      "loss": 2.0146,
+      "step": 17770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.418291893431722e-06,
+      "loss": 2.0317,
+      "step": 17771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.417477084212566e-06,
+      "loss": 2.0518,
+      "step": 17772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4166626246233583e-06,
+      "loss": 2.0693,
+      "step": 17773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4158485146854515e-06,
+      "loss": 2.0969,
+      "step": 17774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4150347544202006e-06,
+      "loss": 2.0417,
+      "step": 17775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.414221343848942e-06,
+      "loss": 2.0667,
+      "step": 17776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4134082829930083e-06,
+      "loss": 2.0764,
+      "step": 17777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.412595571873721e-06,
+      "loss": 2.0525,
+      "step": 17778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.411783210512391e-06,
+      "loss": 2.0613,
+      "step": 17779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4109711989303236e-06,
+      "loss": 2.0801,
+      "step": 17780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4101595371488096e-06,
+      "loss": 2.054,
+      "step": 17781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4093482251891385e-06,
+      "loss": 2.0564,
+      "step": 17782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4085372630725806e-06,
+      "loss": 2.0229,
+      "step": 17783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.40772665082041e-06,
+      "loss": 2.0503,
+      "step": 17784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4069163884538766e-06,
+      "loss": 2.0513,
+      "step": 17785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.406106475994231e-06,
+      "loss": 2.0608,
+      "step": 17786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4052969134627127e-06,
+      "loss": 2.0483,
+      "step": 17787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4044877008805503e-06,
+      "loss": 2.0833,
+      "step": 17788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.403678838268969e-06,
+      "loss": 2.0569,
+      "step": 17789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.402870325649173e-06,
+      "loss": 2.063,
+      "step": 17790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4020621630423723e-06,
+      "loss": 2.0857,
+      "step": 17791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.401254350469753e-06,
+      "loss": 2.0938,
+      "step": 17792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.4004468879525037e-06,
+      "loss": 2.0254,
+      "step": 17793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3996397755117987e-06,
+      "loss": 2.0276,
+      "step": 17794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.398833013168803e-06,
+      "loss": 2.052,
+      "step": 17795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3980266009446693e-06,
+      "loss": 2.0959,
+      "step": 17796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3972205388605527e-06,
+      "loss": 2.041,
+      "step": 17797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3964148269375814e-06,
+      "loss": 2.0303,
+      "step": 17798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.395609465196896e-06,
+      "loss": 2.0249,
+      "step": 17799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.394804453659606e-06,
+      "loss": 2.0461,
+      "step": 17800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3939997923468265e-06,
+      "loss": 2.0393,
+      "step": 17801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3931954812796562e-06,
+      "loss": 2.0308,
+      "step": 17802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.39239152047919e-06,
+      "loss": 2.0696,
+      "step": 17803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3915879099665092e-06,
+      "loss": 2.0547,
+      "step": 17804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.390784649762689e-06,
+      "loss": 2.0715,
+      "step": 17805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3899817398887937e-06,
+      "loss": 2.0627,
+      "step": 17806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.389179180365878e-06,
+      "loss": 2.0674,
+      "step": 17807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3883769712149874e-06,
+      "loss": 2.0737,
+      "step": 17808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3875751124571608e-06,
+      "loss": 2.1023,
+      "step": 17809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.386773604113426e-06,
+      "loss": 2.0852,
+      "step": 17810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.385972446204797e-06,
+      "loss": 2.092,
+      "step": 17811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3851716387522896e-06,
+      "loss": 2.1121,
+      "step": 17812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.384371181776897e-06,
+      "loss": 2.0349,
+      "step": 17813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.38357107529962e-06,
+      "loss": 2.0583,
+      "step": 17814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.382771319341433e-06,
+      "loss": 2.1023,
+      "step": 17815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.38197191392331e-06,
+      "loss": 2.0984,
+      "step": 17816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.381172859066215e-06,
+      "loss": 2.071,
+      "step": 17817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.380374154791104e-06,
+      "loss": 2.0627,
+      "step": 17818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3795758011189194e-06,
+      "loss": 2.0703,
+      "step": 17819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3787777980705982e-06,
+      "loss": 2.0894,
+      "step": 17820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.377980145667067e-06,
+      "loss": 2.0747,
+      "step": 17821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.377182843929244e-06,
+      "loss": 2.0676,
+      "step": 17822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.376385892878035e-06,
+      "loss": 2.063,
+      "step": 17823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.375589292534343e-06,
+      "loss": 2.0579,
+      "step": 17824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3747930429190575e-06,
+      "loss": 2.0691,
+      "step": 17825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.373997144053053e-06,
+      "loss": 2.084,
+      "step": 17826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.373201595957211e-06,
+      "loss": 2.0457,
+      "step": 17827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.372406398652384e-06,
+      "loss": 2.0427,
+      "step": 17828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3716115521594335e-06,
+      "loss": 2.0884,
+      "step": 17829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.370817056499197e-06,
+      "loss": 2.0854,
+      "step": 17830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3700229116925138e-06,
+      "loss": 1.9688,
+      "step": 17831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.369229117760205e-06,
+      "loss": 1.9714,
+      "step": 17832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3684356747230905e-06,
+      "loss": 2.0305,
+      "step": 17833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.367642582601976e-06,
+      "loss": 2.0908,
+      "step": 17834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3668498414176607e-06,
+      "loss": 2.1147,
+      "step": 17835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3660574511909306e-06,
+      "loss": 2.1204,
+      "step": 17836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.36526541194257e-06,
+      "loss": 2.0046,
+      "step": 17837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3644737236933423e-06,
+      "loss": 2.0129,
+      "step": 17838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3636823864640146e-06,
+      "loss": 2.0786,
+      "step": 17839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3628914002753365e-06,
+      "loss": 2.0608,
+      "step": 17840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.362100765148052e-06,
+      "loss": 2.0798,
+      "step": 17841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3613104811028907e-06,
+      "loss": 2.0708,
+      "step": 17842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3605205481605815e-06,
+      "loss": 2.0613,
+      "step": 17843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.359730966341837e-06,
+      "loss": 2.0466,
+      "step": 17844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.358941735667365e-06,
+      "loss": 2.0413,
+      "step": 17845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.358152856157858e-06,
+      "loss": 2.0581,
+      "step": 17846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3573643278340085e-06,
+      "loss": 2.0879,
+      "step": 17847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3565761507164913e-06,
+      "loss": 2.0896,
+      "step": 17848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.355788324825978e-06,
+      "loss": 2.074,
+      "step": 17849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.355000850183129e-06,
+      "loss": 2.0688,
+      "step": 17850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3542137268085885e-06,
+      "loss": 2.0591,
+      "step": 17851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3534269547230052e-06,
+      "loss": 2.0618,
+      "step": 17852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3526405339470054e-06,
+      "loss": 2.0857,
+      "step": 17853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3518544645012185e-06,
+      "loss": 2.0908,
+      "step": 17854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3510687464062537e-06,
+      "loss": 2.0889,
+      "step": 17855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3502833796827155e-06,
+      "loss": 2.114,
+      "step": 17856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.349498364351203e-06,
+      "loss": 2.1064,
+      "step": 17857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3487137004322996e-06,
+      "loss": 2.0911,
+      "step": 17858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.34792938794658e-06,
+      "loss": 2.1035,
+      "step": 17859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3471454269146154e-06,
+      "loss": 2.1084,
+      "step": 17860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3463618173569634e-06,
+      "loss": 2.0654,
+      "step": 17861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.345578559294173e-06,
+      "loss": 2.0945,
+      "step": 17862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3447956527467844e-06,
+      "loss": 2.0972,
+      "step": 17863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.344013097735328e-06,
+      "loss": 2.0771,
+      "step": 17864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.343230894280327e-06,
+      "loss": 2.0535,
+      "step": 17865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3424490424022886e-06,
+      "loss": 2.0808,
+      "step": 17866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3416675421217247e-06,
+      "loss": 2.0237,
+      "step": 17867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3408863934591173e-06,
+      "loss": 2.0127,
+      "step": 17868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3401055964349648e-06,
+      "loss": 2.0776,
+      "step": 17869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3393251510697346e-06,
+      "loss": 2.0691,
+      "step": 17870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3385450573838927e-06,
+      "loss": 2.0696,
+      "step": 17871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3377653153979005e-06,
+      "loss": 2.094,
+      "step": 17872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3369859251321997e-06,
+      "loss": 2.0627,
+      "step": 17873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.336206886607235e-06,
+      "loss": 2.0684,
+      "step": 17874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3354281998434322e-06,
+      "loss": 2.0422,
+      "step": 17875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.334649864861212e-06,
+      "loss": 2.0403,
+      "step": 17876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3338718816809854e-06,
+      "loss": 2.1287,
+      "step": 17877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3330942503231556e-06,
+      "loss": 2.0811,
+      "step": 17878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.332316970808114e-06,
+      "loss": 2.0823,
+      "step": 17879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.331540043156244e-06,
+      "loss": 2.0769,
+      "step": 17880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.330763467387918e-06,
+      "loss": 2.0994,
+      "step": 17881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3299872435235063e-06,
+      "loss": 2.0498,
+      "step": 17882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3292113715833545e-06,
+      "loss": 2.0862,
+      "step": 17883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3284358515878194e-06,
+      "loss": 2.0894,
+      "step": 17884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.327660683557228e-06,
+      "loss": 2.01,
+      "step": 17885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3268858675119215e-06,
+      "loss": 2.0291,
+      "step": 17886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3261114034722056e-06,
+      "loss": 2.0581,
+      "step": 17887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3253372914583954e-06,
+      "loss": 2.1128,
+      "step": 17888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.324563531490792e-06,
+      "loss": 2.0261,
+      "step": 17889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.323790123589685e-06,
+      "loss": 2.0176,
+      "step": 17890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.323017067775354e-06,
+      "loss": 1.998,
+      "step": 17891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3222443640680747e-06,
+      "loss": 2.0132,
+      "step": 17892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3214720124881093e-06,
+      "loss": 2.0837,
+      "step": 17893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.320700013055712e-06,
+      "loss": 2.0901,
+      "step": 17894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.319928365791128e-06,
+      "loss": 2.064,
+      "step": 17895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.319157070714591e-06,
+      "loss": 2.0779,
+      "step": 17896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3183861278463298e-06,
+      "loss": 2.1394,
+      "step": 17897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.3176155372065583e-06,
+      "loss": 2.1284,
+      "step": 17898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.316845298815489e-06,
+      "loss": 2.0754,
+      "step": 17899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3160754126933133e-06,
+      "loss": 2.0618,
+      "step": 17900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3153058788602306e-06,
+      "loss": 2.0476,
+      "step": 17901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3145366973364115e-06,
+      "loss": 2.0276,
+      "step": 17902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.313767868142031e-06,
+      "loss": 2.0857,
+      "step": 17903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.312999391297252e-06,
+      "loss": 2.0872,
+      "step": 17904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3122312668222244e-06,
+      "loss": 2.0688,
+      "step": 17905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.311463494737093e-06,
+      "loss": 2.1033,
+      "step": 17906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.310696075061994e-06,
+      "loss": 2.043,
+      "step": 17907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3099290078170433e-06,
+      "loss": 2.0376,
+      "step": 17908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.309162293022368e-06,
+      "loss": 2.104,
+      "step": 17909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3083959306980628e-06,
+      "loss": 2.1101,
+      "step": 17910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3076299208642377e-06,
+      "loss": 2.0759,
+      "step": 17911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.306864263540968e-06,
+      "loss": 2.0989,
+      "step": 17912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.30609895874834e-06,
+      "loss": 2.0249,
+      "step": 17913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3053340065064186e-06,
+      "loss": 2.0068,
+      "step": 17914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3045694068352652e-06,
+      "loss": 2.1157,
+      "step": 17915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3038051597549347e-06,
+      "loss": 2.105,
+      "step": 17916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3030412652854597e-06,
+      "loss": 2.0776,
+      "step": 17917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.302277723446881e-06,
+      "loss": 2.05,
+      "step": 17918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3015145342592157e-06,
+      "loss": 2.0679,
+      "step": 17919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.300751697742483e-06,
+      "loss": 2.0642,
+      "step": 17920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2999892139166817e-06,
+      "loss": 2.095,
+      "step": 17921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2992270828018125e-06,
+      "loss": 2.1128,
+      "step": 17922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2984653044178545e-06,
+      "loss": 2.0669,
+      "step": 17923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2977038787847933e-06,
+      "loss": 2.0049,
+      "step": 17924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.296942805922586e-06,
+      "loss": 2.0059,
+      "step": 17925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2961820858512023e-06,
+      "loss": 2.0278,
+      "step": 17926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.295421718590582e-06,
+      "loss": 2.1165,
+      "step": 17927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.294661704160669e-06,
+      "loss": 2.1064,
+      "step": 17928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.293902042581394e-06,
+      "loss": 2.0725,
+      "step": 17929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.293142733872677e-06,
+      "loss": 2.0593,
+      "step": 17930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2923837780544283e-06,
+      "loss": 2.0503,
+      "step": 17931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.291625175146554e-06,
+      "loss": 2.0408,
+      "step": 17932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2908669251689464e-06,
+      "loss": 2.1091,
+      "step": 17933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2901090281414893e-06,
+      "loss": 2.0645,
+      "step": 17934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.289351484084059e-06,
+      "loss": 2.0876,
+      "step": 17935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.288594293016519e-06,
+      "loss": 2.0898,
+      "step": 17936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2878374549587295e-06,
+      "loss": 1.9805,
+      "step": 17937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.28708096993053e-06,
+      "loss": 1.9807,
+      "step": 17938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.286324837951769e-06,
+      "loss": 2.0723,
+      "step": 17939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2855690590422647e-06,
+      "loss": 2.0498,
+      "step": 17940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2848136332218467e-06,
+      "loss": 2.0925,
+      "step": 17941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2840585605103178e-06,
+      "loss": 2.1082,
+      "step": 17942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2833038409274797e-06,
+      "loss": 2.0596,
+      "step": 17943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.282549474493126e-06,
+      "loss": 2.0776,
+      "step": 17944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.28179546122704e-06,
+      "loss": 2.1047,
+      "step": 17945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2810418011489907e-06,
+      "loss": 2.1321,
+      "step": 17946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2802884942787467e-06,
+      "loss": 2.0435,
+      "step": 17947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2795355406360596e-06,
+      "loss": 2.0522,
+      "step": 17948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2787829402406756e-06,
+      "loss": 2.0552,
+      "step": 17949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2780306931123326e-06,
+      "loss": 2.0796,
+      "step": 17950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2772787992707554e-06,
+      "loss": 2.0576,
+      "step": 17951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.276527258735661e-06,
+      "loss": 2.0679,
+      "step": 17952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2757760715267567e-06,
+      "loss": 2.0977,
+      "step": 17953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2750252376637483e-06,
+      "loss": 2.0693,
+      "step": 17954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2742747571663147e-06,
+      "loss": 2.0823,
+      "step": 17955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.273524630054146e-06,
+      "loss": 2.0981,
+      "step": 17956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2727748563469086e-06,
+      "loss": 2.1082,
+      "step": 17957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2720254360642643e-06,
+      "loss": 2.105,
+      "step": 17958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2712763692258693e-06,
+      "loss": 2.0566,
+      "step": 17959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2705276558513623e-06,
+      "loss": 2.0674,
+      "step": 17960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.269779295960381e-06,
+      "loss": 2.0498,
+      "step": 17961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.269031289572548e-06,
+      "loss": 2.0579,
+      "step": 17962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.268283636707481e-06,
+      "loss": 1.9783,
+      "step": 17963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2675363373847845e-06,
+      "loss": 1.9875,
+      "step": 17964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2667893916240577e-06,
+      "loss": 2.0828,
+      "step": 17965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2660427994448853e-06,
+      "loss": 2.0864,
+      "step": 17966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.265296560866851e-06,
+      "loss": 2.0715,
+      "step": 17967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.264550675909516e-06,
+      "loss": 2.0637,
+      "step": 17968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2638051445924494e-06,
+      "loss": 2.0957,
+      "step": 17969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2630599669351932e-06,
+      "loss": 2.0894,
+      "step": 17970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.262315142957294e-06,
+      "loss": 2.0923,
+      "step": 17971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2615706726782853e-06,
+      "loss": 2.0801,
+      "step": 17972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.260826556117686e-06,
+      "loss": 2.0439,
+      "step": 17973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.260082793295011e-06,
+      "loss": 2.0762,
+      "step": 17974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.259339384229764e-06,
+      "loss": 2.0481,
+      "step": 17975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.258596328941443e-06,
+      "loss": 2.0701,
+      "step": 17976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2578536274495327e-06,
+      "loss": 1.9729,
+      "step": 17977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.257111279773504e-06,
+      "loss": 1.9871,
+      "step": 17978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2563692859328344e-06,
+      "loss": 2.1052,
+      "step": 17979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2556276459469704e-06,
+      "loss": 2.0845,
+      "step": 17980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2548863598353714e-06,
+      "loss": 2.0293,
+      "step": 17981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2541454276174683e-06,
+      "loss": 2.0271,
+      "step": 17982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2534048493126975e-06,
+      "loss": 2.1208,
+      "step": 17983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2526646249404747e-06,
+      "loss": 2.0815,
+      "step": 17984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2519247545202114e-06,
+      "loss": 2.0618,
+      "step": 17985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.251185238071317e-06,
+      "loss": 2.0483,
+      "step": 17986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2504460756131778e-06,
+      "loss": 2.0576,
+      "step": 17987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2497072671651782e-06,
+      "loss": 2.0801,
+      "step": 17988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2489688127466938e-06,
+      "loss": 2.1052,
+      "step": 17989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.248230712377092e-06,
+      "loss": 2.0725,
+      "step": 17990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.247492966075723e-06,
+      "loss": 2.1174,
+      "step": 17991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2467555738619394e-06,
+      "loss": 2.0911,
+      "step": 17992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2460185357550716e-06,
+      "loss": 2.0229,
+      "step": 17993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2452818517744556e-06,
+      "loss": 2.0376,
+      "step": 17994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2445455219394024e-06,
+      "loss": 2.0874,
+      "step": 17995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.24380954626923e-06,
+      "loss": 2.1052,
+      "step": 17996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2430739247832294e-06,
+      "loss": 2.0327,
+      "step": 17997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2423386575006956e-06,
+      "loss": 2.0344,
+      "step": 17998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2416037444409127e-06,
+      "loss": 2.0444,
+      "step": 17999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2408691856231472e-06,
+      "loss": 2.0605,
+      "step": 18000
+    },
+    {
+      "Bleu_1": 0.7735965191084472,
+      "Bleu_2": 0.6131911367904052,
+      "Bleu_3": 0.471303573772225,
+      "Bleu_4": 0.3579713267122788,
+      "CIDEr": 1.202034902350407,
+      "METEOR": 0.2780439641612939,
+      "ROUGE_L": 0.5589098841338522,
+      "epoch": 0.9,
+      "step": 18000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.240134981066667e-06,
+      "loss": 2.0388,
+      "step": 18001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2394011307907237e-06,
+      "loss": 2.0347,
+      "step": 18002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2386676348145615e-06,
+      "loss": 2.1011,
+      "step": 18003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2379344931574163e-06,
+      "loss": 2.0725,
+      "step": 18004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2372017058385153e-06,
+      "loss": 2.0542,
+      "step": 18005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.236469272877069e-06,
+      "loss": 2.0488,
+      "step": 18006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2357371942922912e-06,
+      "loss": 2.053,
+      "step": 18007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2350054701033773e-06,
+      "loss": 2.0635,
+      "step": 18008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2342741003295174e-06,
+      "loss": 2.0518,
+      "step": 18009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.233543084989886e-06,
+      "loss": 2.0784,
+      "step": 18010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.23281242410366e-06,
+      "loss": 2.0557,
+      "step": 18011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.232082117689995e-06,
+      "loss": 2.0491,
+      "step": 18012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2313521657680462e-06,
+      "loss": 2.1062,
+      "step": 18013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2306225683569515e-06,
+      "loss": 2.0935,
+      "step": 18014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.229893325475847e-06,
+      "loss": 2.082,
+      "step": 18015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.229164437143854e-06,
+      "loss": 2.1016,
+      "step": 18016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2284359033800895e-06,
+      "loss": 2.1443,
+      "step": 18017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.227707724203654e-06,
+      "loss": 2.1326,
+      "step": 18018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.226979899633648e-06,
+      "loss": 2.053,
+      "step": 18019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.226252429689156e-06,
+      "loss": 2.0632,
+      "step": 18020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2255253143892564e-06,
+      "loss": 2.0217,
+      "step": 18021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2247985537530115e-06,
+      "loss": 2.0044,
+      "step": 18022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2240721477994863e-06,
+      "loss": 2.0437,
+      "step": 18023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.223346096547729e-06,
+      "loss": 2.0681,
+      "step": 18024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2226204000167734e-06,
+      "loss": 2.0703,
+      "step": 18025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2218950582256597e-06,
+      "loss": 2.0574,
+      "step": 18026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2211700711934e-06,
+      "loss": 2.0811,
+      "step": 18027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.220445438939013e-06,
+      "loss": 2.071,
+      "step": 18028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2197211614814973e-06,
+      "loss": 2.0442,
+      "step": 18029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2189972388398465e-06,
+      "loss": 2.0454,
+      "step": 18030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.218273671033046e-06,
+      "loss": 2.1084,
+      "step": 18031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.21755045808007e-06,
+      "loss": 2.0498,
+      "step": 18032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.216827599999885e-06,
+      "loss": 2.0862,
+      "step": 18033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.216105096811446e-06,
+      "loss": 2.073,
+      "step": 18034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2153829485336998e-06,
+      "loss": 2.0649,
+      "step": 18035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.214661155185584e-06,
+      "loss": 2.0679,
+      "step": 18036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.213939716786025e-06,
+      "loss": 2.0979,
+      "step": 18037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.213218633353947e-06,
+      "loss": 2.0854,
+      "step": 18038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2124979049082578e-06,
+      "loss": 2.1021,
+      "step": 18039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2117775314678502e-06,
+      "loss": 2.0972,
+      "step": 18040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2110575130516244e-06,
+      "loss": 2.0676,
+      "step": 18041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.210337849678455e-06,
+      "loss": 2.073,
+      "step": 18042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2096185413672237e-06,
+      "loss": 2.0815,
+      "step": 18043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.208899588136784e-06,
+      "loss": 2.0754,
+      "step": 18044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.208180990005994e-06,
+      "loss": 2.05,
+      "step": 18045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2074627469936986e-06,
+      "loss": 2.054,
+      "step": 18046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2067448591187296e-06,
+      "loss": 2.0298,
+      "step": 18047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2060273263999164e-06,
+      "loss": 2.0632,
+      "step": 18048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.205310148856076e-06,
+      "loss": 2.0808,
+      "step": 18049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2045933265060088e-06,
+      "loss": 2.0645,
+      "step": 18050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2038768593685222e-06,
+      "loss": 2.0449,
+      "step": 18051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.203160747462395e-06,
+      "loss": 2.054,
+      "step": 18052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2024449908064154e-06,
+      "loss": 1.9954,
+      "step": 18053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2017295894193483e-06,
+      "loss": 2.0198,
+      "step": 18054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2010145433199524e-06,
+      "loss": 2.0442,
+      "step": 18055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.2002998525269856e-06,
+      "loss": 2.0149,
+      "step": 18056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1995855170591834e-06,
+      "loss": 2.0591,
+      "step": 18057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1988715369352835e-06,
+      "loss": 2.0598,
+      "step": 18058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.198157912174005e-06,
+      "loss": 2.0593,
+      "step": 18059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1974446427940636e-06,
+      "loss": 2.083,
+      "step": 18060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.196731728814164e-06,
+      "loss": 2.0679,
+      "step": 18061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1960191702530012e-06,
+      "loss": 2.0747,
+      "step": 18062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.19530696712926e-06,
+      "loss": 2.0591,
+      "step": 18063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1945951194616235e-06,
+      "loss": 2.0381,
+      "step": 18064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1938836272687503e-06,
+      "loss": 2.1025,
+      "step": 18065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1931724905693044e-06,
+      "loss": 2.0886,
+      "step": 18066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1924617093819306e-06,
+      "loss": 2.0374,
+      "step": 18067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.191751283725271e-06,
+      "loss": 2.0237,
+      "step": 18068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1910412136179535e-06,
+      "loss": 2.0493,
+      "step": 18069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1903314990786026e-06,
+      "loss": 2.0762,
+      "step": 18070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1896221401258255e-06,
+      "loss": 2.0806,
+      "step": 18071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.188913136778228e-06,
+      "loss": 2.1138,
+      "step": 18072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1882044890543983e-06,
+      "loss": 2.0166,
+      "step": 18073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.187496196972924e-06,
+      "loss": 2.0105,
+      "step": 18074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1867882605523787e-06,
+      "loss": 2.0872,
+      "step": 18075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.186080679811326e-06,
+      "loss": 2.0542,
+      "step": 18076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1853734547683223e-06,
+      "loss": 2.0754,
+      "step": 18077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1846665854419117e-06,
+      "loss": 2.0906,
+      "step": 18078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1839600718506357e-06,
+      "loss": 2.0557,
+      "step": 18079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.183253914013013e-06,
+      "loss": 2.0947,
+      "step": 18080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1825481119475736e-06,
+      "loss": 2.0637,
+      "step": 18081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1818426656728145e-06,
+      "loss": 2.0784,
+      "step": 18082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.181137575207246e-06,
+      "loss": 2.095,
+      "step": 18083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1804328405693514e-06,
+      "loss": 2.1289,
+      "step": 18084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1797284617776125e-06,
+      "loss": 2.0491,
+      "step": 18085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1790244388505017e-06,
+      "loss": 2.0549,
+      "step": 18086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1783207718064817e-06,
+      "loss": 2.1211,
+      "step": 18087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.177617460664004e-06,
+      "loss": 2.1125,
+      "step": 18088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.176914505441514e-06,
+      "loss": 2.0576,
+      "step": 18089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1762119061574433e-06,
+      "loss": 2.0825,
+      "step": 18090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1755096628302196e-06,
+      "loss": 2.0291,
+      "step": 18091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.174807775478257e-06,
+      "loss": 2.0205,
+      "step": 18092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1741062441199605e-06,
+      "loss": 2.0952,
+      "step": 18093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1734050687737307e-06,
+      "loss": 2.0742,
+      "step": 18094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1727042494579492e-06,
+      "loss": 2.0039,
+      "step": 18095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1720037861910014e-06,
+      "loss": 2.0273,
+      "step": 18096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.171303678991248e-06,
+      "loss": 2.0471,
+      "step": 18097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.170603927877055e-06,
+      "loss": 2.0593,
+      "step": 18098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.1699045328667706e-06,
+      "loss": 2.0718,
+      "step": 18099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.169205493978734e-06,
+      "loss": 2.075,
+      "step": 18100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1685068112312784e-06,
+      "loss": 2.0513,
+      "step": 18101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1678084846427266e-06,
+      "loss": 2.0298,
+      "step": 18102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1671105142313905e-06,
+      "loss": 2.0718,
+      "step": 18103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.166412900015573e-06,
+      "loss": 2.0972,
+      "step": 18104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1657156420135687e-06,
+      "loss": 2.0889,
+      "step": 18105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1650187402436615e-06,
+      "loss": 2.0991,
+      "step": 18106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.164322194724129e-06,
+      "loss": 2.0667,
+      "step": 18107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1636260054732353e-06,
+      "loss": 2.0825,
+      "step": 18108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1629301725092386e-06,
+      "loss": 2.0669,
+      "step": 18109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1622346958503825e-06,
+      "loss": 2.0667,
+      "step": 18110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1615395755149135e-06,
+      "loss": 2.0723,
+      "step": 18111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1608448115210495e-06,
+      "loss": 2.095,
+      "step": 18112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.16015040388702e-06,
+      "loss": 2.0916,
+      "step": 18113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1594563526310297e-06,
+      "loss": 2.0791,
+      "step": 18114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.158762657771278e-06,
+      "loss": 2.075,
+      "step": 18115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.158069319325958e-06,
+      "loss": 2.1108,
+      "step": 18116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1573763373132534e-06,
+      "loss": 2.0356,
+      "step": 18117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.156683711751334e-06,
+      "loss": 2.0076,
+      "step": 18118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1559914426583672e-06,
+      "loss": 2.085,
+      "step": 18119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1552995300525e-06,
+      "loss": 2.0728,
+      "step": 18120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1546079739518856e-06,
+      "loss": 2.075,
+      "step": 18121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1539167743746515e-06,
+      "loss": 2.0518,
+      "step": 18122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.153225931338927e-06,
+      "loss": 2.0793,
+      "step": 18123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.152535444862831e-06,
+      "loss": 2.0566,
+      "step": 18124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1518453149644657e-06,
+      "loss": 2.0188,
+      "step": 18125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.151155541661935e-06,
+      "loss": 2.0303,
+      "step": 18126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.150466124973321e-06,
+      "loss": 2.0991,
+      "step": 18127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1497770649167077e-06,
+      "loss": 2.0769,
+      "step": 18128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1490883615101602e-06,
+      "loss": 2.1057,
+      "step": 18129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1484000147717464e-06,
+      "loss": 2.0906,
+      "step": 18130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.14771202471951e-06,
+      "loss": 2.1204,
+      "step": 18131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.147024391371496e-06,
+      "loss": 2.1033,
+      "step": 18132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1463371147457374e-06,
+      "loss": 2.0276,
+      "step": 18133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1456501948602543e-06,
+      "loss": 2.0369,
+      "step": 18134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1449636317330622e-06,
+      "loss": 2.0669,
+      "step": 18135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.144277425382169e-06,
+      "loss": 2.0886,
+      "step": 18136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1435915758255612e-06,
+      "loss": 2.0808,
+      "step": 18137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1429060830812317e-06,
+      "loss": 2.0562,
+      "step": 18138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.142220947167153e-06,
+      "loss": 2.1099,
+      "step": 18139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1415361681012963e-06,
+      "loss": 2.0952,
+      "step": 18140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1408517459016148e-06,
+      "loss": 2.1177,
+      "step": 18141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1401676805860586e-06,
+      "loss": 2.1038,
+      "step": 18142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.139483972172564e-06,
+      "loss": 2.0757,
+      "step": 18143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.138800620679063e-06,
+      "loss": 2.0955,
+      "step": 18144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1381176261234776e-06,
+      "loss": 2.1409,
+      "step": 18145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1374349885237136e-06,
+      "loss": 2.1221,
+      "step": 18146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1367527078976755e-06,
+      "loss": 2.0442,
+      "step": 18147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1360707842632554e-06,
+      "loss": 2.0334,
+      "step": 18148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.135389217638334e-06,
+      "loss": 2.0642,
+      "step": 18149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.134708008040785e-06,
+      "loss": 2.0413,
+      "step": 18150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1340271554884777e-06,
+      "loss": 2.105,
+      "step": 18151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.133346659999258e-06,
+      "loss": 2.0579,
+      "step": 18152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1326665215909778e-06,
+      "loss": 2.0911,
+      "step": 18153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1319867402814668e-06,
+      "loss": 2.0615,
+      "step": 18154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1313073160885604e-06,
+      "loss": 2.0547,
+      "step": 18155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.130628249030066e-06,
+      "loss": 2.0752,
+      "step": 18156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1299495391237976e-06,
+      "loss": 2.0574,
+      "step": 18157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1292711863875514e-06,
+      "loss": 2.0786,
+      "step": 18158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1285931908391178e-06,
+      "loss": 2.0891,
+      "step": 18159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.127915552496274e-06,
+      "loss": 2.1074,
+      "step": 18160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1272382713767896e-06,
+      "loss": 2.0964,
+      "step": 18161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1265613474984304e-06,
+      "loss": 2.0703,
+      "step": 18162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.125884780878945e-06,
+      "loss": 2.0901,
+      "step": 18163
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1252085715360764e-06,
+      "loss": 2.0845,
+      "step": 18164
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.124532719487555e-06,
+      "loss": 2.1045,
+      "step": 18165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.123857224751109e-06,
+      "loss": 2.0903,
+      "step": 18166
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1231820873444436e-06,
+      "loss": 2.0667,
+      "step": 18167
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1225073072852747e-06,
+      "loss": 2.0737,
+      "step": 18168
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1218328845912883e-06,
+      "loss": 2.092,
+      "step": 18169
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1211588192801785e-06,
+      "loss": 2.0859,
+      "step": 18170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1204851113696144e-06,
+      "loss": 2.0698,
+      "step": 18171
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1198117608772673e-06,
+      "loss": 2.0657,
+      "step": 18172
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1191387678207943e-06,
+      "loss": 2.0291,
+      "step": 18173
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1184661322178443e-06,
+      "loss": 2.0205,
+      "step": 18174
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.117793854086056e-06,
+      "loss": 2.0566,
+      "step": 18175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1171219334430574e-06,
+      "loss": 2.0576,
+      "step": 18176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1164503703064702e-06,
+      "loss": 2.0806,
+      "step": 18177
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1157791646939084e-06,
+      "loss": 2.0603,
+      "step": 18178
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1151083166229677e-06,
+      "loss": 2.0962,
+      "step": 18179
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1144378261112445e-06,
+      "loss": 2.0867,
+      "step": 18180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1137676931763226e-06,
+      "loss": 2.0476,
+      "step": 18181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.11309791783577e-06,
+      "loss": 2.0598,
+      "step": 18182
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1124285001071576e-06,
+      "loss": 2.0259,
+      "step": 18183
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1117594400080325e-06,
+      "loss": 2.0215,
+      "step": 18184
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.111090737555948e-06,
+      "loss": 2.1042,
+      "step": 18185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.110422392768433e-06,
+      "loss": 2.0952,
+      "step": 18186
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1097544056630203e-06,
+      "loss": 2.0569,
+      "step": 18187
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1090867762572225e-06,
+      "loss": 2.0874,
+      "step": 18188
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.108419504568549e-06,
+      "loss": 2.054,
+      "step": 18189
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.107752590614499e-06,
+      "loss": 2.0229,
+      "step": 18190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.10708603441256e-06,
+      "loss": 2.0696,
+      "step": 18191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1064198359802106e-06,
+      "loss": 2.0522,
+      "step": 18192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.105753995334926e-06,
+      "loss": 2.0825,
+      "step": 18193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1050885124941622e-06,
+      "loss": 2.0676,
+      "step": 18194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.104423387475374e-06,
+      "loss": 2.092,
+      "step": 18195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.103758620296002e-06,
+      "loss": 2.0752,
+      "step": 18196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1030942109734757e-06,
+      "loss": 2.0667,
+      "step": 18197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1024301595252267e-06,
+      "loss": 2.0728,
+      "step": 18198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1017664659686586e-06,
+      "loss": 2.0283,
+      "step": 18199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1011031303211862e-06,
+      "loss": 2.0447,
+      "step": 18200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1004401526001976e-06,
+      "loss": 2.095,
+      "step": 18201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.099777532823082e-06,
+      "loss": 2.0808,
+      "step": 18202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0991152710072133e-06,
+      "loss": 2.0776,
+      "step": 18203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0984533671699612e-06,
+      "loss": 2.0884,
+      "step": 18204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.097791821328682e-06,
+      "loss": 2.0867,
+      "step": 18205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0971306335007255e-06,
+      "loss": 2.114,
+      "step": 18206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0964698037034257e-06,
+      "loss": 2.0239,
+      "step": 18207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0958093319541186e-06,
+      "loss": 2.0088,
+      "step": 18208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0951492182701193e-06,
+      "loss": 2.0786,
+      "step": 18209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0944894626687432e-06,
+      "loss": 2.1025,
+      "step": 18210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.093830065167289e-06,
+      "loss": 2.0769,
+      "step": 18211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0931710257830468e-06,
+      "loss": 2.0608,
+      "step": 18212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0925123445333e-06,
+      "loss": 2.1003,
+      "step": 18213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.091854021435323e-06,
+      "loss": 2.0991,
+      "step": 18214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.091196056506381e-06,
+      "loss": 2.0879,
+      "step": 18215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0905384497637257e-06,
+      "loss": 2.0869,
+      "step": 18216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0898812012246043e-06,
+      "loss": 2.0371,
+      "step": 18217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.089224310906249e-06,
+      "loss": 2.0544,
+      "step": 18218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.08856777882589e-06,
+      "loss": 2.0608,
+      "step": 18219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.087911605000741e-06,
+      "loss": 2.0425,
+      "step": 18220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0872557894480126e-06,
+      "loss": 2.0652,
+      "step": 18221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.086600332184899e-06,
+      "loss": 2.0952,
+      "step": 18222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0859452332285903e-06,
+      "loss": 2.0232,
+      "step": 18223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0852904925962646e-06,
+      "loss": 2.0374,
+      "step": 18224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.084636110305097e-06,
+      "loss": 2.0269,
+      "step": 18225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0839820863722425e-06,
+      "loss": 2.0269,
+      "step": 18226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0833284208148523e-06,
+      "loss": 2.1094,
+      "step": 18227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.082675113650071e-06,
+      "loss": 2.0771,
+      "step": 18228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0820221648950284e-06,
+      "loss": 2.0442,
+      "step": 18229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.081369574566849e-06,
+      "loss": 2.0771,
+      "step": 18230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.080717342682642e-06,
+      "loss": 2.0251,
+      "step": 18231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0800654692595197e-06,
+      "loss": 2.0459,
+      "step": 18232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0794139543145672e-06,
+      "loss": 2.0684,
+      "step": 18233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.078762797864877e-06,
+      "loss": 2.0818,
+      "step": 18234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0781119999275215e-06,
+      "loss": 1.9993,
+      "step": 18235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.077461560519569e-06,
+      "loss": 2.0159,
+      "step": 18236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.076811479658072e-06,
+      "loss": 2.0461,
+      "step": 18237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.076161757360086e-06,
+      "loss": 2.0427,
+      "step": 18238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0755123936426416e-06,
+      "loss": 2.0269,
+      "step": 18239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0748633885227743e-06,
+      "loss": 2.073,
+      "step": 18240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.074214742017496e-06,
+      "loss": 2.0637,
+      "step": 18241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.073566454143821e-06,
+      "loss": 2.0771,
+      "step": 18242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0729185249187516e-06,
+      "loss": 2.0928,
+      "step": 18243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0722709543592745e-06,
+      "loss": 2.0969,
+      "step": 18244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0716237424823757e-06,
+      "loss": 2.0498,
+      "step": 18245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0709768893050255e-06,
+      "loss": 2.0386,
+      "step": 18246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.070330394844187e-06,
+      "loss": 2.0642,
+      "step": 18247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.069684259116813e-06,
+      "loss": 2.0779,
+      "step": 18248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.06903848213985e-06,
+      "loss": 2.0444,
+      "step": 18249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0683930639302297e-06,
+      "loss": 2.0549,
+      "step": 18250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0677480045048797e-06,
+      "loss": 2.0925,
+      "step": 18251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.067103303880717e-06,
+      "loss": 2.0659,
+      "step": 18252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.066458962074646e-06,
+      "loss": 2.0552,
+      "step": 18253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.065814979103562e-06,
+      "loss": 2.0442,
+      "step": 18254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0651713549843592e-06,
+      "loss": 2.0413,
+      "step": 18255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0645280897339062e-06,
+      "loss": 2.0242,
+      "step": 18256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0638851833690815e-06,
+      "loss": 2.0894,
+      "step": 18257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0632426359067375e-06,
+      "loss": 2.0632,
+      "step": 18258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0626004473637295e-06,
+      "loss": 2.0923,
+      "step": 18259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0619586177568955e-06,
+      "loss": 2.0742,
+      "step": 18260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.061317147103066e-06,
+      "loss": 2.0806,
+      "step": 18261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.060676035419063e-06,
+      "loss": 2.0769,
+      "step": 18262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0600352827217005e-06,
+      "loss": 2.0149,
+      "step": 18263
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.059394889027782e-06,
+      "loss": 2.0339,
+      "step": 18264
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0587548543541e-06,
+      "loss": 2.0979,
+      "step": 18265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0581151787174386e-06,
+      "loss": 2.0781,
+      "step": 18266
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0574758621345703e-06,
+      "loss": 2.0986,
+      "step": 18267
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.056836904622265e-06,
+      "loss": 2.1675,
+      "step": 18268
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0561983061972734e-06,
+      "loss": 2.0989,
+      "step": 18269
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0555600668763497e-06,
+      "loss": 2.0559,
+      "step": 18270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0549221866762226e-06,
+      "loss": 2.0229,
+      "step": 18271
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.054284665613625e-06,
+      "loss": 2.0103,
+      "step": 18272
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.053647503705272e-06,
+      "loss": 2.0303,
+      "step": 18273
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.053010700967874e-06,
+      "loss": 2.0056,
+      "step": 18274
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0523742574181303e-06,
+      "loss": 2.0066,
+      "step": 18275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0517381730727297e-06,
+      "loss": 1.9922,
+      "step": 18276
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0511024479483553e-06,
+      "loss": 2.0942,
+      "step": 18277
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0504670820616787e-06,
+      "loss": 2.1064,
+      "step": 18278
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.049832075429357e-06,
+      "loss": 2.0623,
+      "step": 18279
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0491974280680483e-06,
+      "loss": 2.0815,
+      "step": 18280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0485631399943887e-06,
+      "loss": 2.1104,
+      "step": 18281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.047929211225019e-06,
+      "loss": 2.1289,
+      "step": 18282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0472956417765594e-06,
+      "loss": 2.0676,
+      "step": 18283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.046662431665624e-06,
+      "loss": 2.0886,
+      "step": 18284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.046029580908817e-06,
+      "loss": 2.1282,
+      "step": 18285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.045397089522738e-06,
+      "loss": 2.0911,
+      "step": 18286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0447649575239726e-06,
+      "loss": 2.0288,
+      "step": 18287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0441331849290932e-06,
+      "loss": 2.0205,
+      "step": 18288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0435017717546727e-06,
+      "loss": 2.0215,
+      "step": 18289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.042870718017269e-06,
+      "loss": 2.0454,
+      "step": 18290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.042240023733427e-06,
+      "loss": 2.0967,
+      "step": 18291
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0416096889196873e-06,
+      "loss": 2.0718,
+      "step": 18292
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0409797135925813e-06,
+      "loss": 2.0767,
+      "step": 18293
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0403500977686258e-06,
+      "loss": 2.0376,
+      "step": 18294
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0397208414643378e-06,
+      "loss": 2.083,
+      "step": 18295
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.039091944696212e-06,
+      "loss": 2.0754,
+      "step": 18296
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0384634074807475e-06,
+      "loss": 2.0657,
+      "step": 18297
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0378352298344206e-06,
+      "loss": 2.0737,
+      "step": 18298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0372074117737083e-06,
+      "loss": 2.0793,
+      "step": 18299
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.036579953315071e-06,
+      "loss": 2.0747,
+      "step": 18300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.035952854474967e-06,
+      "loss": 2.0415,
+      "step": 18301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0353261152698395e-06,
+      "loss": 2.0427,
+      "step": 18302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.034699735716124e-06,
+      "loss": 2.0706,
+      "step": 18303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.034073715830246e-06,
+      "loss": 2.0847,
+      "step": 18304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.033448055628623e-06,
+      "loss": 2.0193,
+      "step": 18305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0328227551276617e-06,
+      "loss": 2.05,
+      "step": 18306
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0321978143437612e-06,
+      "loss": 2.0703,
+      "step": 18307
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.031573233293309e-06,
+      "loss": 2.063,
+      "step": 18308
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0309490119926815e-06,
+      "loss": 2.0588,
+      "step": 18309
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0303251504582544e-06,
+      "loss": 2.0527,
+      "step": 18310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.029701648706379e-06,
+      "loss": 2.1367,
+      "step": 18311
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0290785067534163e-06,
+      "loss": 2.1162,
+      "step": 18312
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0284557246156977e-06,
+      "loss": 2.0603,
+      "step": 18313
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0278333023095602e-06,
+      "loss": 2.071,
+      "step": 18314
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0272112398513258e-06,
+      "loss": 2.0925,
+      "step": 18315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.026589537257305e-06,
+      "loss": 2.0464,
+      "step": 18316
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0259681945438047e-06,
+      "loss": 2.1255,
+      "step": 18317
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0253472117271165e-06,
+      "loss": 2.0801,
+      "step": 18318
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0247265888235246e-06,
+      "loss": 2.0972,
+      "step": 18319
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0241063258493053e-06,
+      "loss": 2.0881,
+      "step": 18320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0234864228207244e-06,
+      "loss": 2.0625,
+      "step": 18321
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.022866879754037e-06,
+      "loss": 2.0718,
+      "step": 18322
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.022247696665494e-06,
+      "loss": 2.0947,
+      "step": 18323
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0216288735713253e-06,
+      "loss": 2.1045,
+      "step": 18324
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0210104104877657e-06,
+      "loss": 2.0962,
+      "step": 18325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0203923074310283e-06,
+      "loss": 2.0811,
+      "step": 18326
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.019774564417328e-06,
+      "loss": 2.0549,
+      "step": 18327
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.019157181462858e-06,
+      "loss": 2.0732,
+      "step": 18328
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0185401585838124e-06,
+      "loss": 2.1304,
+      "step": 18329
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0179234957963723e-06,
+      "loss": 2.1094,
+      "step": 18330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0173071931167053e-06,
+      "loss": 1.9983,
+      "step": 18331
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0166912505609792e-06,
+      "loss": 2.0327,
+      "step": 18332
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0160756681453406e-06,
+      "loss": 2.1138,
+      "step": 18333
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0154604458859338e-06,
+      "loss": 2.1099,
+      "step": 18334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0148455837988957e-06,
+      "loss": 2.0825,
+      "step": 18335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.014231081900346e-06,
+      "loss": 2.1118,
+      "step": 18336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0136169402064013e-06,
+      "loss": 2.083,
+      "step": 18337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0130031587331683e-06,
+      "loss": 2.1021,
+      "step": 18338
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0123897374967386e-06,
+      "loss": 2.0164,
+      "step": 18339
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0117766765132033e-06,
+      "loss": 1.9993,
+      "step": 18340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0111639757986345e-06,
+      "loss": 2.0999,
+      "step": 18341
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.010551635369102e-06,
+      "loss": 2.1011,
+      "step": 18342
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0099396552406658e-06,
+      "loss": 2.04,
+      "step": 18343
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0093280354293704e-06,
+      "loss": 2.0579,
+      "step": 18344
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0087167759512564e-06,
+      "loss": 2.1045,
+      "step": 18345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0081058768223513e-06,
+      "loss": 2.103,
+      "step": 18346
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.007495338058681e-06,
+      "loss": 2.0366,
+      "step": 18347
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.006885159676253e-06,
+      "loss": 2.0469,
+      "step": 18348
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0062753416910645e-06,
+      "loss": 2.0583,
+      "step": 18349
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.005665884119115e-06,
+      "loss": 2.0752,
+      "step": 18350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0050567869763785e-06,
+      "loss": 2.0903,
+      "step": 18351
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.004448050278836e-06,
+      "loss": 2.061,
+      "step": 18352
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0038396740424443e-06,
+      "loss": 2.0115,
+      "step": 18353
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0032316582831623e-06,
+      "loss": 2.0034,
+      "step": 18354
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0026240030169293e-06,
+      "loss": 2.0554,
+      "step": 18355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0020167082596835e-06,
+      "loss": 2.0508,
+      "step": 18356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.001409774027353e-06,
+      "loss": 2.0325,
+      "step": 18357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.000803200335848e-06,
+      "loss": 2.0166,
+      "step": 18358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0001969872010807e-06,
+      "loss": 2.0112,
+      "step": 18359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9995911346389467e-06,
+      "loss": 1.9971,
+      "step": 18360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.99898564266533e-06,
+      "loss": 2.0022,
+      "step": 18361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9983805112961167e-06,
+      "loss": 2.031,
+      "step": 18362
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9977757405471687e-06,
+      "loss": 2.1094,
+      "step": 18363
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9971713304343463e-06,
+      "loss": 2.1155,
+      "step": 18364
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9965672809735043e-06,
+      "loss": 2.0623,
+      "step": 18365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.995963592180474e-06,
+      "loss": 2.0789,
+      "step": 18366
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.995360264071099e-06,
+      "loss": 2.0652,
+      "step": 18367
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.994757296661189e-06,
+      "loss": 2.0955,
+      "step": 18368
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9941546899665647e-06,
+      "loss": 2.075,
+      "step": 18369
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.993552444003024e-06,
+      "loss": 2.0684,
+      "step": 18370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9929505587863606e-06,
+      "loss": 2.0645,
+      "step": 18371
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9923490343323602e-06,
+      "loss": 2.0439,
+      "step": 18372
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9917478706567954e-06,
+      "loss": 2.0503,
+      "step": 18373
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9911470677754316e-06,
+      "loss": 2.0415,
+      "step": 18374
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.990546625704023e-06,
+      "loss": 2.0615,
+      "step": 18375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9899465444583194e-06,
+      "loss": 2.0754,
+      "step": 18376
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.989346824054052e-06,
+      "loss": 2.0554,
+      "step": 18377
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9887474645069504e-06,
+      "loss": 2.05,
+      "step": 18378
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.988148465832732e-06,
+      "loss": 2.0679,
+      "step": 18379
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9875498280471077e-06,
+      "loss": 2.0662,
+      "step": 18380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9869515511657698e-06,
+      "loss": 2.0283,
+      "step": 18381
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9863536352044126e-06,
+      "loss": 2.001,
+      "step": 18382
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9857560801787117e-06,
+      "loss": 2.0762,
+      "step": 18383
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.985158886104343e-06,
+      "loss": 2.0837,
+      "step": 18384
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9845620529969603e-06,
+      "loss": 2.0984,
+      "step": 18385
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.98396558087222e-06,
+      "loss": 2.0977,
+      "step": 18386
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.983369469745761e-06,
+      "loss": 2.074,
+      "step": 18387
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9827737196332172e-06,
+      "loss": 2.082,
+      "step": 18388
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9821783305502116e-06,
+      "loss": 2.0803,
+      "step": 18389
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9815833025123563e-06,
+      "loss": 2.0955,
+      "step": 18390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9809886355352576e-06,
+      "loss": 2.0427,
+      "step": 18391
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.980394329634507e-06,
+      "loss": 2.0208,
+      "step": 18392
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9798003848256917e-06,
+      "loss": 2.0955,
+      "step": 18393
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.979206801124385e-06,
+      "loss": 2.104,
+      "step": 18394
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9786135785461566e-06,
+      "loss": 2.1091,
+      "step": 18395
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9780207171065555e-06,
+      "loss": 2.0977,
+      "step": 18396
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9774282168211393e-06,
+      "loss": 1.9963,
+      "step": 18397
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.976836077705437e-06,
+      "loss": 2.0508,
+      "step": 18398
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9762442997749836e-06,
+      "loss": 2.106,
+      "step": 18399
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9756528830452924e-06,
+      "loss": 2.1162,
+      "step": 18400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9750618275318732e-06,
+      "loss": 2.1414,
+      "step": 18401
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.97447113325023e-06,
+      "loss": 2.0879,
+      "step": 18402
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.973880800215848e-06,
+      "loss": 2.0693,
+      "step": 18403
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.973290828444211e-06,
+      "loss": 2.1099,
+      "step": 18404
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9727012179507892e-06,
+      "loss": 2.0691,
+      "step": 18405
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9721119687510437e-06,
+      "loss": 2.0596,
+      "step": 18406
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.971523080860429e-06,
+      "loss": 2.0251,
+      "step": 18407
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.970934554294386e-06,
+      "loss": 2.0232,
+      "step": 18408
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9703463890683514e-06,
+      "loss": 2.1067,
+      "step": 18409
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9697585851977472e-06,
+      "loss": 2.0959,
+      "step": 18410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.969171142697983e-06,
+      "loss": 2.1284,
+      "step": 18411
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9685840615844716e-06,
+      "loss": 2.1221,
+      "step": 18412
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9679973418726032e-06,
+      "loss": 2.0139,
+      "step": 18413
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9674109835777664e-06,
+      "loss": 2.0149,
+      "step": 18414
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9668249867153386e-06,
+      "loss": 2.1086,
+      "step": 18415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9662393513006845e-06,
+      "loss": 2.1233,
+      "step": 18416
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9656540773491625e-06,
+      "loss": 2.0239,
+      "step": 18417
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9650691648761217e-06,
+      "loss": 2.0273,
+      "step": 18418
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9644846138968986e-06,
+      "loss": 2.0254,
+      "step": 18419
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9639004244268265e-06,
+      "loss": 2.0518,
+      "step": 18420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9633165964812184e-06,
+      "loss": 2.0566,
+      "step": 18421
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9627331300753914e-06,
+      "loss": 2.0686,
+      "step": 18422
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.962150025224642e-06,
+      "loss": 2.0718,
+      "step": 18423
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.961567281944265e-06,
+      "loss": 2.0471,
+      "step": 18424
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9609849002495374e-06,
+      "loss": 2.1133,
+      "step": 18425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.960402880155734e-06,
+      "loss": 2.1282,
+      "step": 18426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9598212216781177e-06,
+      "loss": 2.0042,
+      "step": 18427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9592399248319407e-06,
+      "loss": 2.0364,
+      "step": 18428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9586589896324488e-06,
+      "loss": 2.0591,
+      "step": 18429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9580784160948766e-06,
+      "loss": 2.0916,
+      "step": 18430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.957498204234446e-06,
+      "loss": 2.0649,
+      "step": 18431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9569183540663753e-06,
+      "loss": 2.0901,
+      "step": 18432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.95633886560587e-06,
+      "loss": 2.073,
+      "step": 18433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.955759738868123e-06,
+      "loss": 2.0427,
+      "step": 18434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.955180973868327e-06,
+      "loss": 2.1001,
+      "step": 18435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9546025706216524e-06,
+      "loss": 2.0701,
+      "step": 18436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9540245291432762e-06,
+      "loss": 2.0454,
+      "step": 18437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9534468494483483e-06,
+      "loss": 2.0696,
+      "step": 18438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.952869531552023e-06,
+      "loss": 2.0964,
+      "step": 18439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9522925754694364e-06,
+      "loss": 2.1199,
+      "step": 18440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.951715981215722e-06,
+      "loss": 2.106,
+      "step": 18441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9511397488059955e-06,
+      "loss": 2.1221,
+      "step": 18442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9505638782553737e-06,
+      "loss": 2.103,
+      "step": 18443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.949988369578954e-06,
+      "loss": 2.0793,
+      "step": 18444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9494132227918295e-06,
+      "loss": 2.0825,
+      "step": 18445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9488384379090847e-06,
+      "loss": 2.0662,
+      "step": 18446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.948264014945789e-06,
+      "loss": 2.0891,
+      "step": 18447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.947689953917008e-06,
+      "loss": 2.085,
+      "step": 18448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.947116254837795e-06,
+      "loss": 2.0242,
+      "step": 18449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9465429177231993e-06,
+      "loss": 2.0723,
+      "step": 18450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.945969942588246e-06,
+      "loss": 2.0525,
+      "step": 18451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.945397329447972e-06,
+      "loss": 2.0625,
+      "step": 18452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9448250783173855e-06,
+      "loss": 2.093,
+      "step": 18453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9442531892114974e-06,
+      "loss": 2.0679,
+      "step": 18454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9436816621453025e-06,
+      "loss": 2.0493,
+      "step": 18455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.943110497133787e-06,
+      "loss": 2.0547,
+      "step": 18456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9425396941919345e-06,
+      "loss": 2.0867,
+      "step": 18457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.941969253334709e-06,
+      "loss": 2.0945,
+      "step": 18458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9413991745770686e-06,
+      "loss": 2.0347,
+      "step": 18459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9408294579339687e-06,
+      "loss": 2.0645,
+      "step": 18460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.940260103420343e-06,
+      "loss": 2.0334,
+      "step": 18461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.939691111051127e-06,
+      "loss": 2.0566,
+      "step": 18462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9391224808412405e-06,
+      "loss": 2.0356,
+      "step": 18463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9385542128055943e-06,
+      "loss": 2.072,
+      "step": 18464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.937986306959092e-06,
+      "loss": 2.0515,
+      "step": 18465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.937418763316623e-06,
+      "loss": 2.0696,
+      "step": 18466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9368515818930773e-06,
+      "loss": 2.051,
+      "step": 18467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9362847627033177e-06,
+      "loss": 2.0637,
+      "step": 18468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9357183057622193e-06,
+      "loss": 2.1123,
+      "step": 18469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9351522110846297e-06,
+      "loss": 2.1079,
+      "step": 18470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.934586478685398e-06,
+      "loss": 2.0123,
+      "step": 18471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.934021108579358e-06,
+      "loss": 2.0442,
+      "step": 18472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.933456100781335e-06,
+      "loss": 2.1011,
+      "step": 18473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.932891455306149e-06,
+      "loss": 2.0955,
+      "step": 18474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.932327172168604e-06,
+      "loss": 2.0889,
+      "step": 18475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9317632513834993e-06,
+      "loss": 2.1118,
+      "step": 18476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.931199692965622e-06,
+      "loss": 2.0547,
+      "step": 18477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9306364969297512e-06,
+      "loss": 2.0442,
+      "step": 18478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.930073663290658e-06,
+      "loss": 2.0042,
+      "step": 18479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.929511192063102e-06,
+      "loss": 2.0393,
+      "step": 18480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.928949083261827e-06,
+      "loss": 2.0498,
+      "step": 18481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.928387336901582e-06,
+      "loss": 2.0742,
+      "step": 18482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.927825952997089e-06,
+      "loss": 2.0903,
+      "step": 18483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.92726493156308e-06,
+      "loss": 2.0454,
+      "step": 18484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9267042726142598e-06,
+      "loss": 2.0657,
+      "step": 18485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.926143976165336e-06,
+      "loss": 2.0559,
+      "step": 18486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.925584042230997e-06,
+      "loss": 2.064,
+      "step": 18487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.92502447082593e-06,
+      "loss": 2.0449,
+      "step": 18488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.924465261964806e-06,
+      "loss": 2.0811,
+      "step": 18489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.923906415662292e-06,
+      "loss": 2.0793,
+      "step": 18490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9233479319330443e-06,
+      "loss": 2.0496,
+      "step": 18491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.922789810791706e-06,
+      "loss": 2.0479,
+      "step": 18492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.922232052252911e-06,
+      "loss": 2.0801,
+      "step": 18493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9216746563312925e-06,
+      "loss": 2.0779,
+      "step": 18494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.92111762304146e-06,
+      "loss": 2.0283,
+      "step": 18495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9205609523980272e-06,
+      "loss": 2.0205,
+      "step": 18496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.920004644415589e-06,
+      "loss": 2.04,
+      "step": 18497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9194486991087336e-06,
+      "loss": 2.0708,
+      "step": 18498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9188931164920444e-06,
+      "loss": 2.0857,
+      "step": 18499
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.918337896580083e-06,
+      "loss": 2.0667,
+      "step": 18500
+    },
+    {
+      "Bleu_1": 0.7734624347171721,
+      "Bleu_2": 0.6125478831871777,
+      "Bleu_3": 0.47037559675681007,
+      "Bleu_4": 0.3570016548912108,
+      "CIDEr": 1.2000467668133745,
+      "METEOR": 0.2776440074932206,
+      "ROUGE_L": 0.5588531311320349,
+      "epoch": 0.93,
+      "step": 18500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9177830393874164e-06,
+      "loss": 2.0771,
+      "step": 18501
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9172285449285944e-06,
+      "loss": 2.1045,
+      "step": 18502
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9166744132181534e-06,
+      "loss": 2.0288,
+      "step": 18503
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9161206442706275e-06,
+      "loss": 2.042,
+      "step": 18504
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9155672381005397e-06,
+      "loss": 2.1194,
+      "step": 18505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.915014194722403e-06,
+      "loss": 2.1128,
+      "step": 18506
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.914461514150718e-06,
+      "loss": 2.0718,
+      "step": 18507
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9139091963999765e-06,
+      "loss": 2.0908,
+      "step": 18508
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9133572414846684e-06,
+      "loss": 2.0981,
+      "step": 18509
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.912805649419262e-06,
+      "loss": 2.1062,
+      "step": 18510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9122544202182276e-06,
+      "loss": 2.0454,
+      "step": 18511
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9117035538960164e-06,
+      "loss": 2.061,
+      "step": 18512
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9111530504670745e-06,
+      "loss": 2.0837,
+      "step": 18513
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.910602909945838e-06,
+      "loss": 2.0405,
+      "step": 18514
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9100531323467367e-06,
+      "loss": 2.0464,
+      "step": 18515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9095037176841866e-06,
+      "loss": 2.074,
+      "step": 18516
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.908954665972594e-06,
+      "loss": 2.074,
+      "step": 18517
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.908405977226358e-06,
+      "loss": 2.0569,
+      "step": 18518
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.907857651459868e-06,
+      "loss": 2.0571,
+      "step": 18519
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9073096886875025e-06,
+      "loss": 2.0535,
+      "step": 18520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.906762088923632e-06,
+      "loss": 2.053,
+      "step": 18521
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.906214852182615e-06,
+      "loss": 2.0701,
+      "step": 18522
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9056679784788033e-06,
+      "loss": 2.0588,
+      "step": 18523
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9051214678265388e-06,
+      "loss": 2.1042,
+      "step": 18524
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.90457532024015e-06,
+      "loss": 2.0342,
+      "step": 18525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9040295357339638e-06,
+      "loss": 2.0439,
+      "step": 18526
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9034841143222887e-06,
+      "loss": 2.0364,
+      "step": 18527
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.90293905601943e-06,
+      "loss": 2.0576,
+      "step": 18528
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9023943608396793e-06,
+      "loss": 2.0669,
+      "step": 18529
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9018500287973222e-06,
+      "loss": 2.0598,
+      "step": 18530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9013060599066306e-06,
+      "loss": 2.0918,
+      "step": 18531
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9007624541818747e-06,
+      "loss": 2.104,
+      "step": 18532
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9002192116373046e-06,
+      "loss": 2.0725,
+      "step": 18533
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8996763322871708e-06,
+      "loss": 2.0701,
+      "step": 18534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.899133816145705e-06,
+      "loss": 2.0796,
+      "step": 18535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.898591663227136e-06,
+      "loss": 2.075,
+      "step": 18536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.898049873545684e-06,
+      "loss": 2.1252,
+      "step": 18537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.897508447115553e-06,
+      "loss": 2.0798,
+      "step": 18538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.896967383950941e-06,
+      "loss": 2.0825,
+      "step": 18539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.896426684066039e-06,
+      "loss": 2.0916,
+      "step": 18540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8958863474750282e-06,
+      "loss": 2.0615,
+      "step": 18541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.895346374192073e-06,
+      "loss": 2.0625,
+      "step": 18542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8948067642313349e-06,
+      "loss": 2.0227,
+      "step": 18543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.894267517606968e-06,
+      "loss": 2.0444,
+      "step": 18544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8937286343331106e-06,
+      "loss": 2.0713,
+      "step": 18545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8931901144238933e-06,
+      "loss": 2.0486,
+      "step": 18546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8926519578934415e-06,
+      "loss": 2.0984,
+      "step": 18547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.892114164755867e-06,
+      "loss": 2.1306,
+      "step": 18548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8915767350252706e-06,
+      "loss": 2.0662,
+      "step": 18549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8910396687157472e-06,
+      "loss": 2.082,
+      "step": 18550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8905029658413792e-06,
+      "loss": 2.072,
+      "step": 18551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.889966626416245e-06,
+      "loss": 2.0698,
+      "step": 18552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.889430650454404e-06,
+      "loss": 2.0935,
+      "step": 18553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8888950379699177e-06,
+      "loss": 2.0884,
+      "step": 18554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8883597889768234e-06,
+      "loss": 2.0471,
+      "step": 18555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.887824903489168e-06,
+      "loss": 2.0554,
+      "step": 18556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8872903815209709e-06,
+      "loss": 2.0684,
+      "step": 18557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.88675622308625e-06,
+      "loss": 2.114,
+      "step": 18558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8862224281990132e-06,
+      "loss": 2.0476,
+      "step": 18559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.885688996873262e-06,
+      "loss": 2.0574,
+      "step": 18560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8851559291229825e-06,
+      "loss": 2.116,
+      "step": 18561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8846232249621548e-06,
+      "loss": 2.1067,
+      "step": 18562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.884090884404746e-06,
+      "loss": 2.0508,
+      "step": 18563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8835589074647187e-06,
+      "loss": 2.0544,
+      "step": 18564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8830272941560196e-06,
+      "loss": 2.0496,
+      "step": 18565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8824960444925958e-06,
+      "loss": 2.0544,
+      "step": 18566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8819651584883735e-06,
+      "loss": 2.0972,
+      "step": 18567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8814346361572777e-06,
+      "loss": 2.0852,
+      "step": 18568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8809044775132205e-06,
+      "loss": 2.0488,
+      "step": 18569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8803746825700996e-06,
+      "loss": 2.0767,
+      "step": 18570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8798452513418152e-06,
+      "loss": 2.0444,
+      "step": 18571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.879316183842248e-06,
+      "loss": 2.0405,
+      "step": 18572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8787874800852717e-06,
+      "loss": 2.054,
+      "step": 18573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.878259140084752e-06,
+      "loss": 2.0542,
+      "step": 18574
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8777311638545448e-06,
+      "loss": 2.0747,
+      "step": 18575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8772035514084902e-06,
+      "loss": 2.0923,
+      "step": 18576
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8766763027604323e-06,
+      "loss": 2.0688,
+      "step": 18577
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8761494179241916e-06,
+      "loss": 2.0481,
+      "step": 18578
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8756228969135908e-06,
+      "loss": 2.0925,
+      "step": 18579
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.87509673974243e-06,
+      "loss": 2.085,
+      "step": 18580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.874570946424512e-06,
+      "loss": 2.0908,
+      "step": 18581
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8740455169736236e-06,
+      "loss": 2.0466,
+      "step": 18582
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8735204514035442e-06,
+      "loss": 2.0681,
+      "step": 18583
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8729957497280426e-06,
+      "loss": 2.0793,
+      "step": 18584
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8724714119608776e-06,
+      "loss": 2.0522,
+      "step": 18585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8719474381158004e-06,
+      "loss": 2.0457,
+      "step": 18586
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.871423828206551e-06,
+      "loss": 2.03,
+      "step": 18587
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8709005822468622e-06,
+      "loss": 2.0396,
+      "step": 18588
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8703777002504563e-06,
+      "loss": 2.0195,
+      "step": 18589
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8698551822310407e-06,
+      "loss": 2.0383,
+      "step": 18590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8693330282023191e-06,
+      "loss": 2.0762,
+      "step": 18591
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8688112381779893e-06,
+      "loss": 2.0894,
+      "step": 18592
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8682898121717284e-06,
+      "loss": 2.0398,
+      "step": 18593
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8677687501972141e-06,
+      "loss": 2.0686,
+      "step": 18594
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8672480522681068e-06,
+      "loss": 2.1016,
+      "step": 18595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8667277183980674e-06,
+      "loss": 2.0708,
+      "step": 18596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8662077486007367e-06,
+      "loss": 2.0408,
+      "step": 18597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8656881428897505e-06,
+      "loss": 2.0532,
+      "step": 18598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8651689012787343e-06,
+      "loss": 2.0396,
+      "step": 18599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8646500237813057e-06,
+      "loss": 2.0239,
+      "step": 18600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8641315104110723e-06,
+      "loss": 2.073,
+      "step": 18601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8636133611816313e-06,
+      "loss": 2.0659,
+      "step": 18602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8630955761065689e-06,
+      "loss": 2.043,
+      "step": 18603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8625781551994653e-06,
+      "loss": 2.0786,
+      "step": 18604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.86206109847389e-06,
+      "loss": 2.0942,
+      "step": 18605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8615444059433973e-06,
+      "loss": 2.093,
+      "step": 18606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8610280776215428e-06,
+      "loss": 2.0762,
+      "step": 18607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.860512113521864e-06,
+      "loss": 2.064,
+      "step": 18608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8599965136578934e-06,
+      "loss": 2.0422,
+      "step": 18609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.859481278043147e-06,
+      "loss": 2.113,
+      "step": 18610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8589664066911406e-06,
+      "loss": 2.1006,
+      "step": 18611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.858451899615375e-06,
+      "loss": 2.0867,
+      "step": 18612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8579377568293426e-06,
+      "loss": 2.0757,
+      "step": 18613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.857423978346528e-06,
+      "loss": 2.0681,
+      "step": 18614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8569105641804002e-06,
+      "loss": 2.085,
+      "step": 18615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8563975143444267e-06,
+      "loss": 2.1138,
+      "step": 18616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8558848288520602e-06,
+      "loss": 2.0271,
+      "step": 18617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8553725077167434e-06,
+      "loss": 2.0854,
+      "step": 18618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8548605509519152e-06,
+      "loss": 2.0403,
+      "step": 18619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8543489585709967e-06,
+      "loss": 2.0781,
+      "step": 18620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8538377305874089e-06,
+      "loss": 2.0439,
+      "step": 18621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8533268670145542e-06,
+      "loss": 2.063,
+      "step": 18622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8528163678658306e-06,
+      "loss": 2.0859,
+      "step": 18623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.852306233154627e-06,
+      "loss": 2.0901,
+      "step": 18624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.851796462894318e-06,
+      "loss": 2.0549,
+      "step": 18625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8512870570982744e-06,
+      "loss": 2.075,
+      "step": 18626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8507780157798504e-06,
+      "loss": 2.0376,
+      "step": 18627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8502693389524022e-06,
+      "loss": 2.0317,
+      "step": 18628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.849761026629264e-06,
+      "loss": 2.0439,
+      "step": 18629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8492530788237671e-06,
+      "loss": 2.0349,
+      "step": 18630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8487454955492322e-06,
+      "loss": 2.0417,
+      "step": 18631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.848238276818967e-06,
+      "loss": 2.0647,
+      "step": 18632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8477314226462793e-06,
+      "loss": 2.0884,
+      "step": 18633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8472249330444552e-06,
+      "loss": 2.0911,
+      "step": 18634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8467188080267771e-06,
+      "loss": 2.0854,
+      "step": 18635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8462130476065198e-06,
+      "loss": 2.0884,
+      "step": 18636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.845707651796944e-06,
+      "loss": 2.0713,
+      "step": 18637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8452026206113076e-06,
+      "loss": 2.0466,
+      "step": 18638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.84469795406285e-06,
+      "loss": 2.0684,
+      "step": 18639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.844193652164804e-06,
+      "loss": 2.0901,
+      "step": 18640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8436897149304005e-06,
+      "loss": 2.0464,
+      "step": 18641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8431861423728507e-06,
+      "loss": 2.0667,
+      "step": 18642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8426829345053626e-06,
+      "loss": 2.0552,
+      "step": 18643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8421800913411269e-06,
+      "loss": 2.0525,
+      "step": 18644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.841677612893335e-06,
+      "loss": 2.094,
+      "step": 18645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8411754991751628e-06,
+      "loss": 2.0701,
+      "step": 18646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8406737501997766e-06,
+      "loss": 2.1101,
+      "step": 18647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.840172365980336e-06,
+      "loss": 2.1165,
+      "step": 18648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8396713465299885e-06,
+      "loss": 1.9954,
+      "step": 18649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8391706918618706e-06,
+      "loss": 1.9775,
+      "step": 18650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8386704019891148e-06,
+      "loss": 2.0503,
+      "step": 18651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8381704769248375e-06,
+      "loss": 2.031,
+      "step": 18652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8376709166821516e-06,
+      "loss": 2.0928,
+      "step": 18653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8371717212741545e-06,
+      "loss": 2.0793,
+      "step": 18654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8366728907139396e-06,
+      "loss": 2.0139,
+      "step": 18655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8361744250145863e-06,
+      "loss": 2.0244,
+      "step": 18656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8356763241891672e-06,
+      "loss": 2.0725,
+      "step": 18657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8351785882507438e-06,
+      "loss": 2.0552,
+      "step": 18658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8346812172123706e-06,
+      "loss": 2.0957,
+      "step": 18659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8341842110870887e-06,
+      "loss": 2.0706,
+      "step": 18660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8336875698879313e-06,
+      "loss": 2.0876,
+      "step": 18661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8331912936279226e-06,
+      "loss": 2.0981,
+      "step": 18662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8326953823200794e-06,
+      "loss": 2.0557,
+      "step": 18663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.832199835977404e-06,
+      "loss": 2.0615,
+      "step": 18664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.831704654612888e-06,
+      "loss": 2.009,
+      "step": 18665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8312098382395225e-06,
+      "loss": 2.0715,
+      "step": 18666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8307153868702825e-06,
+      "loss": 2.0591,
+      "step": 18667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8302213005181322e-06,
+      "loss": 2.0984,
+      "step": 18668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8297275791960316e-06,
+      "loss": 2.0737,
+      "step": 18669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.829234222916925e-06,
+      "loss": 2.0754,
+      "step": 18670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8287412316937508e-06,
+      "loss": 2.0159,
+      "step": 18671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8282486055394399e-06,
+      "loss": 2.0361,
+      "step": 18672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8277563444669054e-06,
+      "loss": 2.0698,
+      "step": 18673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8272644484890606e-06,
+      "loss": 2.0615,
+      "step": 18674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8267729176188047e-06,
+      "loss": 2.0364,
+      "step": 18675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8262817518690244e-06,
+      "loss": 2.0886,
+      "step": 18676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8257909512526029e-06,
+      "loss": 2.0193,
+      "step": 18677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8253005157824094e-06,
+      "loss": 2.0605,
+      "step": 18678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.824810445471307e-06,
+      "loss": 2.0789,
+      "step": 18679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.824320740332146e-06,
+      "loss": 2.084,
+      "step": 18680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8238314003777674e-06,
+      "loss": 2.1008,
+      "step": 18681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8233424256210041e-06,
+      "loss": 2.1191,
+      "step": 18682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8228538160746811e-06,
+      "loss": 2.0579,
+      "step": 18683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.822365571751608e-06,
+      "loss": 2.0491,
+      "step": 18684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8218776926645914e-06,
+      "loss": 2.1062,
+      "step": 18685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8213901788264225e-06,
+      "loss": 2.1267,
+      "step": 18686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.820903030249888e-06,
+      "loss": 2.0847,
+      "step": 18687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8204162469477609e-06,
+      "loss": 2.103,
+      "step": 18688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8199298289328108e-06,
+      "loss": 2.0654,
+      "step": 18689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.819443776217788e-06,
+      "loss": 2.0894,
+      "step": 18690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8189580888154417e-06,
+      "loss": 2.062,
+      "step": 18691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.818472766738507e-06,
+      "loss": 2.0635,
+      "step": 18692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8179878099997137e-06,
+      "loss": 2.0242,
+      "step": 18693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8175032186117784e-06,
+      "loss": 2.0544,
+      "step": 18694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8170189925874042e-06,
+      "loss": 2.113,
+      "step": 18695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8165351319392974e-06,
+      "loss": 2.1206,
+      "step": 18696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8160516366801382e-06,
+      "loss": 1.9956,
+      "step": 18697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.815568506822613e-06,
+      "loss": 2.0217,
+      "step": 18698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.8150857423793864e-06,
+      "loss": 2.1077,
+      "step": 18699
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8146033433631204e-06,
+      "loss": 2.113,
+      "step": 18700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8141213097864644e-06,
+      "loss": 2.0654,
+      "step": 18701
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8136396416620621e-06,
+      "loss": 2.0691,
+      "step": 18702
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8131583390025396e-06,
+      "loss": 2.0098,
+      "step": 18703
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8126774018205256e-06,
+      "loss": 2.0332,
+      "step": 18704
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.812196830128623e-06,
+      "loss": 2.0862,
+      "step": 18705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.81171662393944e-06,
+      "loss": 2.0667,
+      "step": 18706
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8112367832655703e-06,
+      "loss": 2.0569,
+      "step": 18707
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.810757308119595e-06,
+      "loss": 2.0405,
+      "step": 18708
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8102781985140892e-06,
+      "loss": 2.1016,
+      "step": 18709
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.809799454461613e-06,
+      "loss": 2.1067,
+      "step": 18710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8093210759747264e-06,
+      "loss": 2.063,
+      "step": 18711
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8088430630659692e-06,
+      "loss": 2.0542,
+      "step": 18712
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.808365415747883e-06,
+      "loss": 2.0403,
+      "step": 18713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8078881340329877e-06,
+      "loss": 2.0764,
+      "step": 18714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8074112179338021e-06,
+      "loss": 2.0869,
+      "step": 18715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8069346674628322e-06,
+      "loss": 2.1091,
+      "step": 18716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8064584826325754e-06,
+      "loss": 2.0288,
+      "step": 18717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8059826634555176e-06,
+      "loss": 2.0481,
+      "step": 18718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8055072099441395e-06,
+      "loss": 1.9966,
+      "step": 18719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8050321221109055e-06,
+      "loss": 2.0139,
+      "step": 18720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8045573999682796e-06,
+      "loss": 2.0913,
+      "step": 18721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8040830435287032e-06,
+      "loss": 2.1138,
+      "step": 18722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8036090528046245e-06,
+      "loss": 2.0447,
+      "step": 18723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8031354278084671e-06,
+      "loss": 2.0632,
+      "step": 18724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8026621685526529e-06,
+      "loss": 2.0725,
+      "step": 18725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8021892750495934e-06,
+      "loss": 2.0913,
+      "step": 18726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8017167473116886e-06,
+      "loss": 2.0923,
+      "step": 18727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8012445853513305e-06,
+      "loss": 2.0737,
+      "step": 18728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8007727891809007e-06,
+      "loss": 2.0657,
+      "step": 18729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8003013588127726e-06,
+      "loss": 2.043,
+      "step": 18730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7998302942593065e-06,
+      "loss": 2.0703,
+      "step": 18731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.799359595532859e-06,
+      "loss": 2.0669,
+      "step": 18732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7988892626457704e-06,
+      "loss": 2.114,
+      "step": 18733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7984192956103773e-06,
+      "loss": 2.1033,
+      "step": 18734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7979496944390017e-06,
+      "loss": 2.0693,
+      "step": 18735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.797480459143962e-06,
+      "loss": 2.0776,
+      "step": 18736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7970115897375566e-06,
+      "loss": 2.071,
+      "step": 18737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7965430862320892e-06,
+      "loss": 2.042,
+      "step": 18738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7960749486398381e-06,
+      "loss": 2.103,
+      "step": 18739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.795607176973087e-06,
+      "loss": 2.0859,
+      "step": 18740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7951397712440975e-06,
+      "loss": 2.0701,
+      "step": 18741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7946727314651268e-06,
+      "loss": 2.1091,
+      "step": 18742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7942060576484249e-06,
+      "loss": 2.0669,
+      "step": 18743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7937397498062303e-06,
+      "loss": 2.063,
+      "step": 18744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7932738079507665e-06,
+      "loss": 2.0996,
+      "step": 18745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7928082320942591e-06,
+      "loss": 2.0776,
+      "step": 18746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7923430222489129e-06,
+      "loss": 2.0601,
+      "step": 18747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7918781784269266e-06,
+      "loss": 2.0688,
+      "step": 18748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.791413700640494e-06,
+      "loss": 2.0698,
+      "step": 18749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7909495889017936e-06,
+      "loss": 2.1118,
+      "step": 18750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7904858432229954e-06,
+      "loss": 1.9585,
+      "step": 18751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.79002246361626e-06,
+      "loss": 1.9792,
+      "step": 18752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.789559450093744e-06,
+      "loss": 2.0718,
+      "step": 18753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.789096802667583e-06,
+      "loss": 2.062,
+      "step": 18754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7886345213499138e-06,
+      "loss": 2.0674,
+      "step": 18755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7881726061528552e-06,
+      "loss": 2.0872,
+      "step": 18756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7877110570885258e-06,
+      "loss": 2.0979,
+      "step": 18757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7872498741690223e-06,
+      "loss": 2.0796,
+      "step": 18758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7867890574064439e-06,
+      "loss": 2.0459,
+      "step": 18759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7863286068128738e-06,
+      "loss": 2.0674,
+      "step": 18760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7858685224003873e-06,
+      "loss": 2.0493,
+      "step": 18761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7854088041810485e-06,
+      "loss": 2.0322,
+      "step": 18762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7849494521669124e-06,
+      "loss": 2.0649,
+      "step": 18763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7844904663700278e-06,
+      "loss": 2.0754,
+      "step": 18764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.784031846802427e-06,
+      "loss": 2.0671,
+      "step": 18765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7835735934761416e-06,
+      "loss": 2.0925,
+      "step": 18766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7831157064031823e-06,
+      "loss": 2.0732,
+      "step": 18767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7826581855955643e-06,
+      "loss": 2.0684,
+      "step": 18768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7822010310652782e-06,
+      "loss": 2.0996,
+      "step": 18769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7817442428243174e-06,
+      "loss": 2.0374,
+      "step": 18770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7812878208846595e-06,
+      "loss": 2.1045,
+      "step": 18771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.780831765258273e-06,
+      "loss": 2.0811,
+      "step": 18772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.780376075957118e-06,
+      "loss": 2.0027,
+      "step": 18773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.779920752993142e-06,
+      "loss": 2.0002,
+      "step": 18774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.779465796378289e-06,
+      "loss": 2.0103,
+      "step": 18775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7790112061244889e-06,
+      "loss": 1.9705,
+      "step": 18776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.778556982243661e-06,
+      "loss": 2.0349,
+      "step": 18777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.778103124747717e-06,
+      "loss": 2.0513,
+      "step": 18778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.777649633648561e-06,
+      "loss": 2.1042,
+      "step": 18779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7771965089580817e-06,
+      "loss": 2.0994,
+      "step": 18780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7767437506881662e-06,
+      "loss": 2.0952,
+      "step": 18781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7762913588506852e-06,
+      "loss": 2.0828,
+      "step": 18782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7758393334575025e-06,
+      "loss": 2.0569,
+      "step": 18783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7753876745204684e-06,
+      "loss": 2.0952,
+      "step": 18784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7749363820514335e-06,
+      "loss": 2.0906,
+      "step": 18785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7744854560622286e-06,
+      "loss": 2.0859,
+      "step": 18786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.774034896564679e-06,
+      "loss": 2.0928,
+      "step": 18787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7735847035706005e-06,
+      "loss": 2.071,
+      "step": 18788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7731348770918e-06,
+      "loss": 2.0256,
+      "step": 18789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7726854171400714e-06,
+      "loss": 2.0364,
+      "step": 18790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7722363237272037e-06,
+      "loss": 2.0425,
+      "step": 18791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7717875968649693e-06,
+      "loss": 2.0479,
+      "step": 18792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.771339236565142e-06,
+      "loss": 2.0334,
+      "step": 18793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7708912428394758e-06,
+      "loss": 2.0234,
+      "step": 18794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7704436156997212e-06,
+      "loss": 2.1033,
+      "step": 18795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.769996355157612e-06,
+      "loss": 2.1099,
+      "step": 18796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7695494612248808e-06,
+      "loss": 2.051,
+      "step": 18797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7691029339132464e-06,
+      "loss": 2.0752,
+      "step": 18798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7686567732344177e-06,
+      "loss": 2.0698,
+      "step": 18799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7682109792000956e-06,
+      "loss": 2.053,
+      "step": 18800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7677655518219687e-06,
+      "loss": 2.0688,
+      "step": 18801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7673204911117214e-06,
+      "loss": 2.0659,
+      "step": 18802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7668757970810206e-06,
+      "loss": 2.1077,
+      "step": 18803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7664314697415306e-06,
+      "loss": 2.1011,
+      "step": 18804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7659875091049055e-06,
+      "loss": 2.0452,
+      "step": 18805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.765543915182784e-06,
+      "loss": 2.0459,
+      "step": 18806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.765100687986797e-06,
+      "loss": 2.0876,
+      "step": 18807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7646578275285719e-06,
+      "loss": 2.0938,
+      "step": 18808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7642153338197193e-06,
+      "loss": 2.1003,
+      "step": 18809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7637732068718482e-06,
+      "loss": 2.0825,
+      "step": 18810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7633314466965462e-06,
+      "loss": 2.0413,
+      "step": 18811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7628900533054004e-06,
+      "loss": 2.0645,
+      "step": 18812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7624490267099869e-06,
+      "loss": 2.1003,
+      "step": 18813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7620083669218712e-06,
+      "loss": 2.0649,
+      "step": 18814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7615680739526092e-06,
+      "loss": 2.042,
+      "step": 18815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7611281478137447e-06,
+      "loss": 2.0569,
+      "step": 18816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7606885885168155e-06,
+      "loss": 2.1094,
+      "step": 18817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7602493960733507e-06,
+      "loss": 2.0908,
+      "step": 18818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.759810570494864e-06,
+      "loss": 2.0056,
+      "step": 18819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7593721117928669e-06,
+      "loss": 2.0371,
+      "step": 18820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7589340199788544e-06,
+      "loss": 2.0229,
+      "step": 18821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7584962950643146e-06,
+      "loss": 2.0259,
+      "step": 18822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7580589370607314e-06,
+      "loss": 1.9844,
+      "step": 18823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7576219459795673e-06,
+      "loss": 2.0034,
+      "step": 18824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7571853218322847e-06,
+      "loss": 2.062,
+      "step": 18825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7567490646303365e-06,
+      "loss": 2.0657,
+      "step": 18826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.756313174385158e-06,
+      "loss": 2.0642,
+      "step": 18827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7558776511081836e-06,
+      "loss": 2.0581,
+      "step": 18828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7554424948108311e-06,
+      "loss": 2.0959,
+      "step": 18829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7550077055045128e-06,
+      "loss": 2.1111,
+      "step": 18830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7545732832006361e-06,
+      "loss": 2.0151,
+      "step": 18831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7541392279105856e-06,
+      "loss": 2.0205,
+      "step": 18832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.753705539645747e-06,
+      "loss": 2.0474,
+      "step": 18833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7532722184174943e-06,
+      "loss": 2.0884,
+      "step": 18834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7528392642371905e-06,
+      "loss": 2.1011,
+      "step": 18835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7524066771161878e-06,
+      "loss": 2.1055,
+      "step": 18836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7519744570658309e-06,
+      "loss": 2.0774,
+      "step": 18837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7515426040974552e-06,
+      "loss": 2.0376,
+      "step": 18838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7511111182223855e-06,
+      "loss": 2.0494,
+      "step": 18839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7506799994519357e-06,
+      "loss": 2.0667,
+      "step": 18840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7502492477974119e-06,
+      "loss": 2.0886,
+      "step": 18841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.74981886327011e-06,
+      "loss": 2.0874,
+      "step": 18842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7493888458813179e-06,
+      "loss": 2.064,
+      "step": 18843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7489591956423095e-06,
+      "loss": 2.0808,
+      "step": 18844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7485299125643545e-06,
+      "loss": 2.0842,
+      "step": 18845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7481009966587086e-06,
+      "loss": 2.0759,
+      "step": 18846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7476724479366197e-06,
+      "loss": 2.0977,
+      "step": 18847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7472442664093268e-06,
+      "loss": 2.1167,
+      "step": 18848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7468164520880598e-06,
+      "loss": 2.0332,
+      "step": 18849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7463890049840342e-06,
+      "loss": 2.0259,
+      "step": 18850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7459619251084612e-06,
+      "loss": 2.0054,
+      "step": 18851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7455352124725422e-06,
+      "loss": 1.9978,
+      "step": 18852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7451088670874643e-06,
+      "loss": 2.0496,
+      "step": 18853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.744682888964409e-06,
+      "loss": 2.0767,
+      "step": 18854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.744257278114549e-06,
+      "loss": 2.0918,
+      "step": 18855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7438320345490399e-06,
+      "loss": 2.0703,
+      "step": 18856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.74340715827904e-06,
+      "loss": 2.0544,
+      "step": 18857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.742982649315687e-06,
+      "loss": 2.0825,
+      "step": 18858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.742558507670115e-06,
+      "loss": 2.0254,
+      "step": 18859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7421347333534454e-06,
+      "loss": 2.0203,
+      "step": 18860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.741711326376791e-06,
+      "loss": 2.0828,
+      "step": 18861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7412882867512575e-06,
+      "loss": 2.0879,
+      "step": 18862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7408656144879364e-06,
+      "loss": 2.0801,
+      "step": 18863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7404433095979106e-06,
+      "loss": 2.0759,
+      "step": 18864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7400213720922591e-06,
+      "loss": 2.0881,
+      "step": 18865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7395998019820403e-06,
+      "loss": 2.0598,
+      "step": 18866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7391785992783166e-06,
+      "loss": 2.0481,
+      "step": 18867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7387577639921292e-06,
+      "loss": 2.0425,
+      "step": 18868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7383372961345145e-06,
+      "loss": 2.0364,
+      "step": 18869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7379171957164985e-06,
+      "loss": 2.0288,
+      "step": 18870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7374974627490975e-06,
+      "loss": 2.0852,
+      "step": 18871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7370780972433226e-06,
+      "loss": 2.064,
+      "step": 18872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.736659099210165e-06,
+      "loss": 2.0483,
+      "step": 18873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7362404686606176e-06,
+      "loss": 2.0559,
+      "step": 18874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.735822205605655e-06,
+      "loss": 2.0862,
+      "step": 18875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7354043100562467e-06,
+      "loss": 2.0779,
+      "step": 18876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7349867820233523e-06,
+      "loss": 2.0994,
+      "step": 18877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.734569621517923e-06,
+      "loss": 2.0864,
+      "step": 18878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7341528285508918e-06,
+      "loss": 2.0171,
+      "step": 18879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.733736403133195e-06,
+      "loss": 2.0017,
+      "step": 18880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7333203452757484e-06,
+      "loss": 2.0581,
+      "step": 18881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7329046549894689e-06,
+      "loss": 2.0559,
+      "step": 18882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7324893322852508e-06,
+      "loss": 2.0742,
+      "step": 18883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7320743771739872e-06,
+      "loss": 2.062,
+      "step": 18884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.731659789666561e-06,
+      "loss": 2.0996,
+      "step": 18885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7312455697738466e-06,
+      "loss": 2.1025,
+      "step": 18886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7308317175067006e-06,
+      "loss": 2.1162,
+      "step": 18887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.730418232875981e-06,
+      "loss": 2.095,
+      "step": 18888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7300051158925273e-06,
+      "loss": 2.073,
+      "step": 18889
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.729592366567176e-06,
+      "loss": 2.0586,
+      "step": 18890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7291799849107482e-06,
+      "loss": 2.0254,
+      "step": 18891
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7287679709340604e-06,
+      "loss": 2.0605,
+      "step": 18892
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7283563246479173e-06,
+      "loss": 2.0425,
+      "step": 18893
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7279450460631102e-06,
+      "loss": 2.0291,
+      "step": 18894
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7275341351904306e-06,
+      "loss": 2.0273,
+      "step": 18895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.727123592040648e-06,
+      "loss": 2.0371,
+      "step": 18896
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7267134166245325e-06,
+      "loss": 2.155,
+      "step": 18897
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7263036089528384e-06,
+      "loss": 2.1194,
+      "step": 18898
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.725894169036314e-06,
+      "loss": 2.0896,
+      "step": 18899
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7254850968856938e-06,
+      "loss": 2.0938,
+      "step": 18900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7250763925117063e-06,
+      "loss": 2.0178,
+      "step": 18901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7246680559250727e-06,
+      "loss": 2.0586,
+      "step": 18902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7242600871364978e-06,
+      "loss": 2.0728,
+      "step": 18903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7238524861566796e-06,
+      "loss": 2.0542,
+      "step": 18904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.723445252996308e-06,
+      "loss": 2.0596,
+      "step": 18905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7230383876660627e-06,
+      "loss": 2.0657,
+      "step": 18906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.722631890176612e-06,
+      "loss": 2.0735,
+      "step": 18907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7222257605386188e-06,
+      "loss": 2.0613,
+      "step": 18908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.721819998762728e-06,
+      "loss": 2.0864,
+      "step": 18909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7214146048595847e-06,
+      "loss": 2.0759,
+      "step": 18910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7210095788398183e-06,
+      "loss": 2.1033,
+      "step": 18911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7206049207140523e-06,
+      "loss": 2.1152,
+      "step": 18912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7202006304928946e-06,
+      "loss": 2.0754,
+      "step": 18913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7197967081869503e-06,
+      "loss": 2.0845,
+      "step": 18914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7193931538068108e-06,
+      "loss": 2.0073,
+      "step": 18915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7189899673630577e-06,
+      "loss": 2.0237,
+      "step": 18916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7185871488662675e-06,
+      "loss": 2.031,
+      "step": 18917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7181846983269982e-06,
+      "loss": 2.1033,
+      "step": 18918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7177826157558086e-06,
+      "loss": 2.0269,
+      "step": 18919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7173809011632414e-06,
+      "loss": 2.0562,
+      "step": 18920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7169795545598282e-06,
+      "loss": 2.0608,
+      "step": 18921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7165785759560977e-06,
+      "loss": 2.0598,
+      "step": 18922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7161779653625642e-06,
+      "loss": 2.0691,
+      "step": 18923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7157777227897314e-06,
+      "loss": 2.0652,
+      "step": 18924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7153778482480975e-06,
+      "loss": 2.0808,
+      "step": 18925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7149783417481473e-06,
+      "loss": 2.062,
+      "step": 18926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7145792033003593e-06,
+      "loss": 2.0063,
+      "step": 18927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7141804329151966e-06,
+      "loss": 2.0178,
+      "step": 18928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.713782030603121e-06,
+      "loss": 2.0884,
+      "step": 18929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7133839963745774e-06,
+      "loss": 2.0945,
+      "step": 18930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.712986330240004e-06,
+      "loss": 2.0767,
+      "step": 18931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7125890322098293e-06,
+      "loss": 2.0828,
+      "step": 18932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7121921022944732e-06,
+      "loss": 2.0034,
+      "step": 18933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7117955405043422e-06,
+      "loss": 2.0056,
+      "step": 18934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7113993468498402e-06,
+      "loss": 2.0525,
+      "step": 18935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.71100352134135e-06,
+      "loss": 2.0728,
+      "step": 18936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7106080639892603e-06,
+      "loss": 2.0806,
+      "step": 18937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7102129748039342e-06,
+      "loss": 2.073,
+      "step": 18938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7098182537957351e-06,
+      "loss": 2.1216,
+      "step": 18939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7094239009750135e-06,
+      "loss": 2.105,
+      "step": 18940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7090299163521105e-06,
+      "loss": 2.0747,
+      "step": 18941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.70863629993736e-06,
+      "loss": 2.084,
+      "step": 18942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7082430517410835e-06,
+      "loss": 2.0977,
+      "step": 18943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7078501717735928e-06,
+      "loss": 2.0854,
+      "step": 18944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7074576600451914e-06,
+      "loss": 2.0767,
+      "step": 18945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7070655165661708e-06,
+      "loss": 2.1104,
+      "step": 18946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7066737413468164e-06,
+      "loss": 2.0627,
+      "step": 18947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7062823343974015e-06,
+      "loss": 2.0952,
+      "step": 18948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7058912957281897e-06,
+      "loss": 2.1082,
+      "step": 18949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7055006253494359e-06,
+      "loss": 2.0806,
+      "step": 18950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.705110323271384e-06,
+      "loss": 2.0967,
+      "step": 18951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7047203895042719e-06,
+      "loss": 2.0913,
+      "step": 18952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7043308240583232e-06,
+      "loss": 2.0349,
+      "step": 18953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.703941626943752e-06,
+      "loss": 2.0347,
+      "step": 18954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7035527981707694e-06,
+      "loss": 2.0005,
+      "step": 18955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7031643377495676e-06,
+      "loss": 2.0366,
+      "step": 18956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7027762456903335e-06,
+      "loss": 2.0828,
+      "step": 18957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7023885220032487e-06,
+      "loss": 2.0872,
+      "step": 18958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7020011666984786e-06,
+      "loss": 2.0781,
+      "step": 18959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7016141797861783e-06,
+      "loss": 2.0732,
+      "step": 18960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7012275612764983e-06,
+      "loss": 2.0144,
+      "step": 18961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7008413111795785e-06,
+      "loss": 2.0342,
+      "step": 18962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7004554295055474e-06,
+      "loss": 2.0894,
+      "step": 18963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7000699162645222e-06,
+      "loss": 2.0813,
+      "step": 18964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6996847714666147e-06,
+      "loss": 2.1094,
+      "step": 18965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6992999951219235e-06,
+      "loss": 2.1047,
+      "step": 18966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6989155872405405e-06,
+      "loss": 2.1226,
+      "step": 18967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6985315478325458e-06,
+      "loss": 2.1077,
+      "step": 18968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6981478769080067e-06,
+      "loss": 2.0261,
+      "step": 18969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6977645744769915e-06,
+      "loss": 2.0427,
+      "step": 18970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6973816405495456e-06,
+      "loss": 2.0505,
+      "step": 18971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6969990751357162e-06,
+      "loss": 2.1035,
+      "step": 18972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6966168782455299e-06,
+      "loss": 2.0916,
+      "step": 18973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.696235049889014e-06,
+      "loss": 2.0886,
+      "step": 18974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6958535900761802e-06,
+      "loss": 2.031,
+      "step": 18975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6954724988170305e-06,
+      "loss": 2.0425,
+      "step": 18976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6950917761215606e-06,
+      "loss": 2.0542,
+      "step": 18977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.694711421999754e-06,
+      "loss": 2.0759,
+      "step": 18978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6943314364615842e-06,
+      "loss": 2.0718,
+      "step": 18979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.693951819517015e-06,
+      "loss": 2.083,
+      "step": 18980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6935725711760036e-06,
+      "loss": 2.0449,
+      "step": 18981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6931936914484953e-06,
+      "loss": 2.0459,
+      "step": 18982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6928151803444236e-06,
+      "loss": 2.0171,
+      "step": 18983
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6924370378737175e-06,
+      "loss": 1.9849,
+      "step": 18984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6920592640462907e-06,
+      "loss": 2.0652,
+      "step": 18985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6916818588720502e-06,
+      "loss": 2.0811,
+      "step": 18986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6913048223608946e-06,
+      "loss": 2.0308,
+      "step": 18987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6909281545227113e-06,
+      "loss": 2.054,
+      "step": 18988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6905518553673755e-06,
+      "loss": 2.0779,
+      "step": 18989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6901759249047577e-06,
+      "loss": 2.0596,
+      "step": 18990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.689800363144715e-06,
+      "loss": 2.1089,
+      "step": 18991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6894251700970947e-06,
+      "loss": 2.092,
+      "step": 18992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6890503457717387e-06,
+      "loss": 2.0391,
+      "step": 18993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6886758901784743e-06,
+      "loss": 2.0857,
+      "step": 18994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6883018033271253e-06,
+      "loss": 2.0554,
+      "step": 18995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6879280852274937e-06,
+      "loss": 2.0332,
+      "step": 18996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6875547358893884e-06,
+      "loss": 2.0884,
+      "step": 18997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6871817553225933e-06,
+      "loss": 2.0786,
+      "step": 18998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6868091435368953e-06,
+      "loss": 2.0842,
+      "step": 18999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6864369005420588e-06,
+      "loss": 2.0862,
+      "step": 19000
+    },
+    {
+      "Bleu_1": 0.7719618857385856,
+      "Bleu_2": 0.6110179523227632,
+      "Bleu_3": 0.4692711217783879,
+      "Bleu_4": 0.35540244363532475,
+      "CIDEr": 1.198063404705114,
+      "METEOR": 0.2774581078802418,
+      "ROUGE_L": 0.5577478350236404,
+      "epoch": 0.95,
+      "step": 19000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.686065026347852e-06,
+      "loss": 2.0845,
+      "step": 19001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6856935209640211e-06,
+      "loss": 2.0962,
+      "step": 19002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6853223844003146e-06,
+      "loss": 2.082,
+      "step": 19003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6849516166664614e-06,
+      "loss": 2.0215,
+      "step": 19004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6845812177721854e-06,
+      "loss": 2.0725,
+      "step": 19005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.684211187727197e-06,
+      "loss": 2.0537,
+      "step": 19006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6838415265412055e-06,
+      "loss": 2.0383,
+      "step": 19007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6834722342238978e-06,
+      "loss": 2.0547,
+      "step": 19008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6831033107849647e-06,
+      "loss": 2.0547,
+      "step": 19009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6827347562340768e-06,
+      "loss": 2.0886,
+      "step": 19010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6823665705808995e-06,
+      "loss": 2.0178,
+      "step": 19011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6819987538350917e-06,
+      "loss": 2.0535,
+      "step": 19012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6816313060062943e-06,
+      "loss": 2.0767,
+      "step": 19013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6812642271041444e-06,
+      "loss": 2.0935,
+      "step": 19014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6808975171382712e-06,
+      "loss": 2.0637,
+      "step": 19015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6805311761182867e-06,
+      "loss": 2.0308,
+      "step": 19016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6801652040538018e-06,
+      "loss": 2.0706,
+      "step": 19017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6797996009544103e-06,
+      "loss": 2.094,
+      "step": 19018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6794343668297013e-06,
+      "loss": 2.1143,
+      "step": 19019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6790695016892537e-06,
+      "loss": 2.0977,
+      "step": 19020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6787050055426334e-06,
+      "loss": 2.0867,
+      "step": 19021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.678340878399401e-06,
+      "loss": 2.0999,
+      "step": 19022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6779771202691018e-06,
+      "loss": 2.0776,
+      "step": 19023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6776137311612789e-06,
+      "loss": 2.0623,
+      "step": 19024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6772507110854623e-06,
+      "loss": 2.0581,
+      "step": 19025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6768880600511683e-06,
+      "loss": 2.053,
+      "step": 19026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6765257780679071e-06,
+      "loss": 2.0898,
+      "step": 19027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6761638651451832e-06,
+      "loss": 2.095,
+      "step": 19028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6758023212924838e-06,
+      "loss": 2.0593,
+      "step": 19029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.67544114651929e-06,
+      "loss": 2.0535,
+      "step": 19030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6750803408350734e-06,
+      "loss": 2.0603,
+      "step": 19031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6747199042492973e-06,
+      "loss": 2.0615,
+      "step": 19032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.674359836771412e-06,
+      "loss": 2.0605,
+      "step": 19033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6740001384108633e-06,
+      "loss": 2.0969,
+      "step": 19034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6736408091770789e-06,
+      "loss": 2.0537,
+      "step": 19035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6732818490794844e-06,
+      "loss": 2.0657,
+      "step": 19036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.672923258127494e-06,
+      "loss": 2.0935,
+      "step": 19037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6725650363305066e-06,
+      "loss": 2.0757,
+      "step": 19038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6722071836979236e-06,
+      "loss": 2.0464,
+      "step": 19039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6718497002391253e-06,
+      "loss": 2.0664,
+      "step": 19040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6714925859634844e-06,
+      "loss": 2.0439,
+      "step": 19041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6711358408803668e-06,
+      "loss": 2.0325,
+      "step": 19042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6707794649991315e-06,
+      "loss": 2.0947,
+      "step": 19043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6704234583291194e-06,
+      "loss": 2.1138,
+      "step": 19044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6700678208796698e-06,
+      "loss": 2.1101,
+      "step": 19045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6697125526601054e-06,
+      "loss": 2.1108,
+      "step": 19046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.669357653679745e-06,
+      "loss": 2.0105,
+      "step": 19047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6690031239478964e-06,
+      "loss": 2.0503,
+      "step": 19048
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6686489634738538e-06,
+      "loss": 2.0566,
+      "step": 19049
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6682951722669082e-06,
+      "loss": 2.0652,
+      "step": 19050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6679417503363337e-06,
+      "loss": 2.054,
+      "step": 19051
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6675886976913996e-06,
+      "loss": 2.0698,
+      "step": 19052
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6672360143413635e-06,
+      "loss": 2.1008,
+      "step": 19053
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6668837002954796e-06,
+      "loss": 2.0826,
+      "step": 19054
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6665317555629806e-06,
+      "loss": 2.0022,
+      "step": 19055
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6661801801530956e-06,
+      "loss": 2.0449,
+      "step": 19056
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6658289740750493e-06,
+      "loss": 2.0637,
+      "step": 19057
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6654781373380472e-06,
+      "loss": 2.0891,
+      "step": 19058
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6651276699512924e-06,
+      "loss": 2.0627,
+      "step": 19059
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6647775719239738e-06,
+      "loss": 2.0581,
+      "step": 19060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6644278432652744e-06,
+      "loss": 2.0942,
+      "step": 19061
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6640784839843614e-06,
+      "loss": 2.114,
+      "step": 19062
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6637294940903997e-06,
+      "loss": 2.1099,
+      "step": 19063
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6633808735925415e-06,
+      "loss": 2.1074,
+      "step": 19064
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6630326224999281e-06,
+      "loss": 2.0447,
+      "step": 19065
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6626847408216905e-06,
+      "loss": 2.0618,
+      "step": 19066
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.662337228566953e-06,
+      "loss": 2.0728,
+      "step": 19067
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6619900857448283e-06,
+      "loss": 2.0771,
+      "step": 19068
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6616433123644193e-06,
+      "loss": 2.0344,
+      "step": 19069
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.66129690843482e-06,
+      "loss": 2.0461,
+      "step": 19070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6609508739651153e-06,
+      "loss": 2.0776,
+      "step": 19071
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6606052089643808e-06,
+      "loss": 2.0818,
+      "step": 19072
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6602599134416779e-06,
+      "loss": 2.1116,
+      "step": 19073
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6599149874060625e-06,
+      "loss": 2.0989,
+      "step": 19074
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6595704308665827e-06,
+      "loss": 1.9783,
+      "step": 19075
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.659226243832269e-06,
+      "loss": 2.0188,
+      "step": 19076
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6588824263121533e-06,
+      "loss": 2.0347,
+      "step": 19077
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.658538978315246e-06,
+      "loss": 2.0371,
+      "step": 19078
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6581958998505552e-06,
+      "loss": 2.0042,
+      "step": 19079
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6578531909270823e-06,
+      "loss": 2.0198,
+      "step": 19080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6575108515538078e-06,
+      "loss": 2.0347,
+      "step": 19081
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6571688817397152e-06,
+      "loss": 2.0793,
+      "step": 19082
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6568272814937685e-06,
+      "loss": 2.0439,
+      "step": 19083
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6564860508249274e-06,
+      "loss": 2.0383,
+      "step": 19084
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6561451897421381e-06,
+      "loss": 2.082,
+      "step": 19085
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.65580469825434e-06,
+      "loss": 2.0706,
+      "step": 19086
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6554645763704646e-06,
+      "loss": 2.0586,
+      "step": 19087
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6551248240994276e-06,
+      "loss": 2.0649,
+      "step": 19088
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6547854414501407e-06,
+      "loss": 2.0964,
+      "step": 19089
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6544464284315064e-06,
+      "loss": 2.0913,
+      "step": 19090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6541077850524095e-06,
+      "loss": 2.0837,
+      "step": 19091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6537695113217344e-06,
+      "loss": 2.0901,
+      "step": 19092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6534316072483474e-06,
+      "loss": 2.0474,
+      "step": 19093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.653094072841118e-06,
+      "loss": 2.0532,
+      "step": 19094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6527569081088878e-06,
+      "loss": 2.041,
+      "step": 19095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.652420113060506e-06,
+      "loss": 2.0642,
+      "step": 19096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6520836877048007e-06,
+      "loss": 2.0959,
+      "step": 19097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6517476320505981e-06,
+      "loss": 2.092,
+      "step": 19098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6514119461067045e-06,
+      "loss": 2.0095,
+      "step": 19099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.651076629881928e-06,
+      "loss": 2.0256,
+      "step": 19100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6507416833850614e-06,
+      "loss": 2.0828,
+      "step": 19101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6504071066248861e-06,
+      "loss": 2.0999,
+      "step": 19102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6500728996101783e-06,
+      "loss": 2.0698,
+      "step": 19103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.649739062349701e-06,
+      "loss": 2.0647,
+      "step": 19104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6494055948522092e-06,
+      "loss": 2.0657,
+      "step": 19105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6490724971264454e-06,
+      "loss": 2.0903,
+      "step": 19106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6487397691811496e-06,
+      "loss": 2.0051,
+      "step": 19107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.648407411025043e-06,
+      "loss": 1.9954,
+      "step": 19108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6480754226668438e-06,
+      "loss": 2.0977,
+      "step": 19109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.647743804115255e-06,
+      "loss": 2.0793,
+      "step": 19110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6474125553789742e-06,
+      "loss": 2.0352,
+      "step": 19111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6470816764666917e-06,
+      "loss": 2.0398,
+      "step": 19112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6467511673870772e-06,
+      "loss": 1.9902,
+      "step": 19113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6464210281488051e-06,
+      "loss": 2.0103,
+      "step": 19114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6460912587605287e-06,
+      "loss": 2.0432,
+      "step": 19115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6457618592308976e-06,
+      "loss": 2.0698,
+      "step": 19116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6454328295685483e-06,
+      "loss": 2.0865,
+      "step": 19117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.645104169782109e-06,
+      "loss": 2.0874,
+      "step": 19118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.644775879880201e-06,
+      "loss": 2.0166,
+      "step": 19119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6444479598714325e-06,
+      "loss": 2.043,
+      "step": 19120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6441204097644014e-06,
+      "loss": 2.0647,
+      "step": 19121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6437932295676977e-06,
+      "loss": 2.0842,
+      "step": 19122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6434664192899013e-06,
+      "loss": 2.0305,
+      "step": 19123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.643139978939585e-06,
+      "loss": 2.0645,
+      "step": 19124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6428139085253057e-06,
+      "loss": 2.0581,
+      "step": 19125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6424882080556163e-06,
+      "loss": 2.0781,
+      "step": 19126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6421628775390569e-06,
+      "loss": 2.0649,
+      "step": 19127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6418379169841585e-06,
+      "loss": 2.0635,
+      "step": 19128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6415133263994448e-06,
+      "loss": 2.0527,
+      "step": 19129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.641189105793427e-06,
+      "loss": 2.0447,
+      "step": 19130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6408652551746051e-06,
+      "loss": 2.0908,
+      "step": 19131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6405417745514755e-06,
+      "loss": 2.0823,
+      "step": 19132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.64021866393252e-06,
+      "loss": 2.058,
+      "step": 19133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6398959233262084e-06,
+      "loss": 2.021,
+      "step": 19134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.639573552741009e-06,
+      "loss": 2.0857,
+      "step": 19135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6392515521853729e-06,
+      "loss": 2.0769,
+      "step": 19136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6389299216677456e-06,
+      "loss": 2.0564,
+      "step": 19137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6386086611965585e-06,
+      "loss": 2.0537,
+      "step": 19138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6382877707802414e-06,
+      "loss": 2.0933,
+      "step": 19139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6379672504272026e-06,
+      "loss": 2.0693,
+      "step": 19140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6376471001458554e-06,
+      "loss": 2.0354,
+      "step": 19141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6373273199445881e-06,
+      "loss": 2.0266,
+      "step": 19142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6370079098317908e-06,
+      "loss": 2.0627,
+      "step": 19143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6366888698158366e-06,
+      "loss": 2.0491,
+      "step": 19144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6363701999050957e-06,
+      "loss": 2.0403,
+      "step": 19145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.636051900107921e-06,
+      "loss": 2.0583,
+      "step": 19146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6357339704326647e-06,
+      "loss": 2.0747,
+      "step": 19147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.635416410887658e-06,
+      "loss": 2.0525,
+      "step": 19148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6350992214812327e-06,
+      "loss": 2.0469,
+      "step": 19149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.634782402221704e-06,
+      "loss": 2.0498,
+      "step": 19150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6344659531173816e-06,
+      "loss": 2.0562,
+      "step": 19151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.634149874176566e-06,
+      "loss": 2.053,
+      "step": 19152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6338341654075417e-06,
+      "loss": 2.0239,
+      "step": 19153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6335188268185923e-06,
+      "loss": 2.0154,
+      "step": 19154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6332038584179831e-06,
+      "loss": 2.0569,
+      "step": 19155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6328892602139771e-06,
+      "loss": 2.0679,
+      "step": 19156
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6325750322148214e-06,
+      "loss": 2.0408,
+      "step": 19157
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6322611744287575e-06,
+      "loss": 2.0613,
+      "step": 19158
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.631947686864017e-06,
+      "loss": 1.9939,
+      "step": 19159
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6316345695288204e-06,
+      "loss": 1.9723,
+      "step": 19160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6313218224313775e-06,
+      "loss": 2.052,
+      "step": 19161
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6310094455798919e-06,
+      "loss": 2.0278,
+      "step": 19162
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6306974389825518e-06,
+      "loss": 2.1001,
+      "step": 19163
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6303858026475428e-06,
+      "loss": 2.0959,
+      "step": 19164
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6300745365830327e-06,
+      "loss": 2.0884,
+      "step": 19165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6297636407971921e-06,
+      "loss": 2.1013,
+      "step": 19166
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6294531152981663e-06,
+      "loss": 2.0127,
+      "step": 19167
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6291429600940998e-06,
+      "loss": 2.0166,
+      "step": 19168
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6288331751931302e-06,
+      "loss": 2.0676,
+      "step": 19169
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.628523760603377e-06,
+      "loss": 2.0417,
+      "step": 19170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6282147163329559e-06,
+      "loss": 2.082,
+      "step": 19171
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6279060423899717e-06,
+      "loss": 2.052,
+      "step": 19172
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6275977387825148e-06,
+      "loss": 2.011,
+      "step": 19173
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6272898055186785e-06,
+      "loss": 2.0178,
+      "step": 19174
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.62698224260653e-06,
+      "loss": 2.0781,
+      "step": 19175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6266750500541393e-06,
+      "loss": 2.0879,
+      "step": 19176
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6263682278695603e-06,
+      "loss": 2.0952,
+      "step": 19177
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6260617760608375e-06,
+      "loss": 2.0989,
+      "step": 19178
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.625755694636012e-06,
+      "loss": 2.0803,
+      "step": 19179
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6254499836031066e-06,
+      "loss": 2.084,
+      "step": 19180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6251446429701389e-06,
+      "loss": 2.1135,
+      "step": 19181
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6248396727451186e-06,
+      "loss": 2.1121,
+      "step": 19182
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6245350729360398e-06,
+      "loss": 2.0784,
+      "step": 19183
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6242308435508908e-06,
+      "loss": 2.0452,
+      "step": 19184
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.62392698459765e-06,
+      "loss": 2.1025,
+      "step": 19185
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6236234960842884e-06,
+      "loss": 2.0957,
+      "step": 19186
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6233203780187605e-06,
+      "loss": 2.0979,
+      "step": 19187
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6230176304090171e-06,
+      "loss": 2.0928,
+      "step": 19188
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6227152532629985e-06,
+      "loss": 2.1401,
+      "step": 19189
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6224132465886332e-06,
+      "loss": 2.1006,
+      "step": 19190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.62211161039384e-06,
+      "loss": 2.0303,
+      "step": 19191
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6218103446865307e-06,
+      "loss": 2.032,
+      "step": 19192
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6215094494746024e-06,
+      "loss": 2.0393,
+      "step": 19193
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6212089247659522e-06,
+      "loss": 2.0403,
+      "step": 19194
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6209087705684538e-06,
+      "loss": 2.0342,
+      "step": 19195
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6206089868899843e-06,
+      "loss": 2.019,
+      "step": 19196
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6203095737384004e-06,
+      "loss": 2.0647,
+      "step": 19197
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6200105311215547e-06,
+      "loss": 2.0564,
+      "step": 19198
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6197118590472937e-06,
+      "loss": 2.0938,
+      "step": 19199
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6194135575234448e-06,
+      "loss": 2.0962,
+      "step": 19200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6191156265578316e-06,
+      "loss": 2.0874,
+      "step": 19201
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.618818066158268e-06,
+      "loss": 2.085,
+      "step": 19202
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.618520876332556e-06,
+      "loss": 2.0879,
+      "step": 19203
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.618224057088491e-06,
+      "loss": 2.0637,
+      "step": 19204
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6179276084338568e-06,
+      "loss": 2.0369,
+      "step": 19205
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6176315303764238e-06,
+      "loss": 2.0298,
+      "step": 19206
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.617335822923961e-06,
+      "loss": 2.053,
+      "step": 19207
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6170404860842186e-06,
+      "loss": 2.0234,
+      "step": 19208
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6167455198649455e-06,
+      "loss": 2.0173,
+      "step": 19209
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6164509242738724e-06,
+      "loss": 1.9963,
+      "step": 19210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.616156699318731e-06,
+      "loss": 2.0535,
+      "step": 19211
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6158628450072289e-06,
+      "loss": 2.0576,
+      "step": 19212
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6155693613470796e-06,
+      "loss": 2.0212,
+      "step": 19213
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.615276248345977e-06,
+      "loss": 2.0312,
+      "step": 19214
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6149835060116053e-06,
+      "loss": 2.0579,
+      "step": 19215
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.614691134351643e-06,
+      "loss": 2.0625,
+      "step": 19216
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6143991333737574e-06,
+      "loss": 2.0481,
+      "step": 19217
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6141075030856055e-06,
+      "loss": 2.051,
+      "step": 19218
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.613816243494835e-06,
+      "loss": 2.032,
+      "step": 19219
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6135253546090827e-06,
+      "loss": 2.0544,
+      "step": 19220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6132348364359795e-06,
+      "loss": 2.0989,
+      "step": 19221
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6129446889831439e-06,
+      "loss": 2.0869,
+      "step": 19222
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6126549122581803e-06,
+      "loss": 2.04,
+      "step": 19223
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6123655062686937e-06,
+      "loss": 2.0337,
+      "step": 19224
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6120764710222669e-06,
+      "loss": 2.031,
+      "step": 19225
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6117878065264869e-06,
+      "loss": 2.0469,
+      "step": 19226
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6114995127889143e-06,
+      "loss": 2.0989,
+      "step": 19227
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6112115898171214e-06,
+      "loss": 2.0647,
+      "step": 19228
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6109240376186474e-06,
+      "loss": 2.0793,
+      "step": 19229
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6106368562010392e-06,
+      "loss": 2.1072,
+      "step": 19230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6103500455718262e-06,
+      "loss": 2.094,
+      "step": 19231
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.610063605738527e-06,
+      "loss": 2.0974,
+      "step": 19232
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6097775367086592e-06,
+      "loss": 1.9792,
+      "step": 19233
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6094918384897217e-06,
+      "loss": 1.9827,
+      "step": 19234
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6092065110892019e-06,
+      "loss": 2.1025,
+      "step": 19235
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6089215545145903e-06,
+      "loss": 2.0959,
+      "step": 19236
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6086369687733542e-06,
+      "loss": 2.0054,
+      "step": 19237
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.608352753872958e-06,
+      "loss": 2.0132,
+      "step": 19238
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6080689098208557e-06,
+      "loss": 2.124,
+      "step": 19239
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.607785436624491e-06,
+      "loss": 2.1062,
+      "step": 19240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6075023342912932e-06,
+      "loss": 2.0696,
+      "step": 19241
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.607219602828691e-06,
+      "loss": 2.0764,
+      "step": 19242
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6069372422440992e-06,
+      "loss": 2.0813,
+      "step": 19243
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6066552525449197e-06,
+      "loss": 2.0862,
+      "step": 19244
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.606373633738548e-06,
+      "loss": 2.0298,
+      "step": 19245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6060923858323689e-06,
+      "loss": 2.0232,
+      "step": 19246
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6058115088337593e-06,
+      "loss": 2.0325,
+      "step": 19247
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6055310027500853e-06,
+      "loss": 2.0076,
+      "step": 19248
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.605250867588699e-06,
+      "loss": 2.0762,
+      "step": 19249
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6049711033569513e-06,
+      "loss": 2.0547,
+      "step": 19250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6046917100621767e-06,
+      "loss": 2.0535,
+      "step": 19251
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.604412687711699e-06,
+      "loss": 2.0679,
+      "step": 19252
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6041340363128421e-06,
+      "loss": 2.0676,
+      "step": 19253
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6038557558729055e-06,
+      "loss": 2.0952,
+      "step": 19254
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.603577846399192e-06,
+      "loss": 2.0564,
+      "step": 19255
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6033003078989884e-06,
+      "loss": 2.0713,
+      "step": 19256
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6030231403795726e-06,
+      "loss": 2.074,
+      "step": 19257
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.602746343848212e-06,
+      "loss": 2.0862,
+      "step": 19258
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.602469918312166e-06,
+      "loss": 2.1279,
+      "step": 19259
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6021938637786837e-06,
+      "loss": 2.134,
+      "step": 19260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6019181802550043e-06,
+      "loss": 2.0654,
+      "step": 19261
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6016428677483585e-06,
+      "loss": 2.0591,
+      "step": 19262
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6013679262659643e-06,
+      "loss": 2.104,
+      "step": 19263
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.601093355815032e-06,
+      "loss": 2.1343,
+      "step": 19264
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6008191564027614e-06,
+      "loss": 2.0645,
+      "step": 19265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6005453280363448e-06,
+      "loss": 2.033,
+      "step": 19266
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6002718707229597e-06,
+      "loss": 2.0703,
+      "step": 19267
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.599998784469784e-06,
+      "loss": 2.0583,
+      "step": 19268
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5997260692839715e-06,
+      "loss": 2.0732,
+      "step": 19269
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.599453725172677e-06,
+      "loss": 2.0842,
+      "step": 19270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.599181752143043e-06,
+      "loss": 2.123,
+      "step": 19271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5989101502022001e-06,
+      "loss": 2.1011,
+      "step": 19272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5986389193572714e-06,
+      "loss": 2.1047,
+      "step": 19273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.598368059615371e-06,
+      "loss": 2.0649,
+      "step": 19274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5980975709836e-06,
+      "loss": 2.1135,
+      "step": 19275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5978274534690523e-06,
+      "loss": 2.1309,
+      "step": 19276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5975577070788128e-06,
+      "loss": 2.1001,
+      "step": 19277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5972883318199538e-06,
+      "loss": 2.0845,
+      "step": 19278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.597019327699538e-06,
+      "loss": 2.0632,
+      "step": 19279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5967506947246201e-06,
+      "loss": 2.052,
+      "step": 19280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5964824329022492e-06,
+      "loss": 2.0029,
+      "step": 19281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5962145422394563e-06,
+      "loss": 2.021,
+      "step": 19282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.595947022743266e-06,
+      "loss": 2.0042,
+      "step": 19283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.595679874420694e-06,
+      "loss": 2.0271,
+      "step": 19284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5954130972787484e-06,
+      "loss": 2.0935,
+      "step": 19285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5951466913244217e-06,
+      "loss": 2.0774,
+      "step": 19286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5948806565647034e-06,
+      "loss": 2.0725,
+      "step": 19287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5946149930065662e-06,
+      "loss": 2.0549,
+      "step": 19288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5943497006569812e-06,
+      "loss": 2.0388,
+      "step": 19289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.594084779522903e-06,
+      "loss": 2.0396,
+      "step": 19290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5938202296112775e-06,
+      "loss": 2.0623,
+      "step": 19291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5935560509290443e-06,
+      "loss": 2.0925,
+      "step": 19292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.593292243483131e-06,
+      "loss": 2.0581,
+      "step": 19293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5930288072804556e-06,
+      "loss": 2.0796,
+      "step": 19294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.592765742327924e-06,
+      "loss": 2.0488,
+      "step": 19295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5925030486324393e-06,
+      "loss": 2.0457,
+      "step": 19296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5922407262008841e-06,
+      "loss": 2.0847,
+      "step": 19297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5919787750401466e-06,
+      "loss": 2.1038,
+      "step": 19298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5917171951570875e-06,
+      "loss": 2.0969,
+      "step": 19299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5914559865585699e-06,
+      "loss": 2.0986,
+      "step": 19300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5911951492514432e-06,
+      "loss": 2.0491,
+      "step": 19301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5909346832425487e-06,
+      "loss": 2.0417,
+      "step": 19302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.590674588538716e-06,
+      "loss": 2.0481,
+      "step": 19303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.590414865146766e-06,
+      "loss": 2.0815,
+      "step": 19304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5901555130735084e-06,
+      "loss": 2.0754,
+      "step": 19305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.589896532325748e-06,
+      "loss": 2.0876,
+      "step": 19306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5896379229102692e-06,
+      "loss": 2.0933,
+      "step": 19307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5893796848338633e-06,
+      "loss": 2.0569,
+      "step": 19308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.589121818103293e-06,
+      "loss": 2.0752,
+      "step": 19309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.588864322725328e-06,
+      "loss": 2.0706,
+      "step": 19310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5886071987067164e-06,
+      "loss": 2.0857,
+      "step": 19311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.588350446054201e-06,
+      "loss": 2.0627,
+      "step": 19312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5880940647745163e-06,
+      "loss": 2.0581,
+      "step": 19313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5878380548743854e-06,
+      "loss": 2.0513,
+      "step": 19314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5875824163605245e-06,
+      "loss": 2.1143,
+      "step": 19315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.587327149239632e-06,
+      "loss": 2.1135,
+      "step": 19316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5870722535184052e-06,
+      "loss": 2.0969,
+      "step": 19317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5868177292035291e-06,
+      "loss": 2.1057,
+      "step": 19318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.586563576301675e-06,
+      "loss": 2.0586,
+      "step": 19319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5863097948195094e-06,
+      "loss": 2.0981,
+      "step": 19320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5860563847636916e-06,
+      "loss": 2.0969,
+      "step": 19321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5858033461408599e-06,
+      "loss": 2.0732,
+      "step": 19322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5855506789576558e-06,
+      "loss": 2.0549,
+      "step": 19323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.585298383220702e-06,
+      "loss": 2.0662,
+      "step": 19324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5850464589366148e-06,
+      "loss": 2.0352,
+      "step": 19325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.584794906112001e-06,
+      "loss": 2.0151,
+      "step": 19326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5845437247534583e-06,
+      "loss": 2.0596,
+      "step": 19327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5842929148675734e-06,
+      "loss": 2.0364,
+      "step": 19328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5840424764609207e-06,
+      "loss": 2.0337,
+      "step": 19329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5837924095400736e-06,
+      "loss": 2.0767,
+      "step": 19330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5835427141115831e-06,
+      "loss": 2.104,
+      "step": 19331
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5832933901820012e-06,
+      "loss": 2.1125,
+      "step": 19332
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.583044437757864e-06,
+      "loss": 2.0715,
+      "step": 19333
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.582795856845703e-06,
+      "loss": 2.0601,
+      "step": 19334
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5825476474520344e-06,
+      "loss": 2.0916,
+      "step": 19335
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5822998095833684e-06,
+      "loss": 2.0837,
+      "step": 19336
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5820523432462031e-06,
+      "loss": 2.0945,
+      "step": 19337
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5818052484470297e-06,
+      "loss": 2.0774,
+      "step": 19338
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5815585251923248e-06,
+      "loss": 2.0625,
+      "step": 19339
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5813121734885616e-06,
+      "loss": 2.0977,
+      "step": 19340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5810661933422e-06,
+      "loss": 2.0486,
+      "step": 19341
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5808205847596895e-06,
+      "loss": 2.0266,
+      "step": 19342
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5805753477474706e-06,
+      "loss": 2.053,
+      "step": 19343
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5803304823119758e-06,
+      "loss": 2.0544,
+      "step": 19344
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5800859884596252e-06,
+      "loss": 2.0681,
+      "step": 19345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5798418661968305e-06,
+      "loss": 2.0527,
+      "step": 19346
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5795981155299965e-06,
+      "loss": 2.1106,
+      "step": 19347
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5793547364655098e-06,
+      "loss": 2.0881,
+      "step": 19348
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5791117290097567e-06,
+      "loss": 2.0591,
+      "step": 19349
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5788690931691074e-06,
+      "loss": 2.0916,
+      "step": 19350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5786268289499283e-06,
+      "loss": 2.0444,
+      "step": 19351
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5783849363585679e-06,
+      "loss": 2.0525,
+      "step": 19352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5781434154013725e-06,
+      "loss": 2.0422,
+      "step": 19353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5779022660846753e-06,
+      "loss": 2.031,
+      "step": 19354
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5776614884148016e-06,
+      "loss": 1.9919,
+      "step": 19355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5774210823980612e-06,
+      "loss": 2.0083,
+      "step": 19356
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5771810480407622e-06,
+      "loss": 2.0857,
+      "step": 19357
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5769413853491982e-06,
+      "loss": 2.0759,
+      "step": 19358
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.576702094329654e-06,
+      "loss": 2.1125,
+      "step": 19359
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5764631749884045e-06,
+      "loss": 2.0901,
+      "step": 19360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5762246273317166e-06,
+      "loss": 2.0403,
+      "step": 19361
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5759864513658434e-06,
+      "loss": 2.084,
+      "step": 19362
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5757486470970316e-06,
+      "loss": 2.052,
+      "step": 19363
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.575511214531518e-06,
+      "loss": 2.0327,
+      "step": 19364
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5752741536755294e-06,
+      "loss": 2.051,
+      "step": 19365
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5750374645352822e-06,
+      "loss": 2.0583,
+      "step": 19366
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5748011471169816e-06,
+      "loss": 2.0337,
+      "step": 19367
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5745652014268275e-06,
+      "loss": 2.0229,
+      "step": 19368
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.574329627471005e-06,
+      "loss": 2.0681,
+      "step": 19369
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5740944252556927e-06,
+      "loss": 2.0669,
+      "step": 19370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5738595947870588e-06,
+      "loss": 2.0884,
+      "step": 19371
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.57362513607126e-06,
+      "loss": 2.061,
+      "step": 19372
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5733910491144465e-06,
+      "loss": 2.0879,
+      "step": 19373
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5731573339227585e-06,
+      "loss": 2.0762,
+      "step": 19374
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5729239905023191e-06,
+      "loss": 2.0251,
+      "step": 19375
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5726910188592535e-06,
+      "loss": 2.0193,
+      "step": 19376
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5724584189996654e-06,
+      "loss": 2.0237,
+      "step": 19377
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5722261909296615e-06,
+      "loss": 2.0034,
+      "step": 19378
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.571994334655325e-06,
+      "loss": 2.0566,
+      "step": 19379
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5717628501827413e-06,
+      "loss": 2.0491,
+      "step": 19380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.571531737517977e-06,
+      "loss": 2.0458,
+      "step": 19381
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5713009966670945e-06,
+      "loss": 2.0288,
+      "step": 19382
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5710706276361465e-06,
+      "loss": 1.9866,
+      "step": 19383
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5708406304311684e-06,
+      "loss": 1.9663,
+      "step": 19384
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5706110050581974e-06,
+      "loss": 2.0884,
+      "step": 19385
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5703817515232533e-06,
+      "loss": 2.1003,
+      "step": 19386
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5701528698323465e-06,
+      "loss": 2.1086,
+      "step": 19387
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.569924359991482e-06,
+      "loss": 2.0745,
+      "step": 19388
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.56969622200665e-06,
+      "loss": 2.0903,
+      "step": 19389
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.569468455883833e-06,
+      "loss": 2.0818,
+      "step": 19390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.569241061629007e-06,
+      "loss": 2.0161,
+      "step": 19391
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5690140392481312e-06,
+      "loss": 2.0496,
+      "step": 19392
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5687873887471605e-06,
+      "loss": 2.0845,
+      "step": 19393
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5685611101320385e-06,
+      "loss": 2.093,
+      "step": 19394
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5683352034087006e-06,
+      "loss": 2.0269,
+      "step": 19395
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5681096685830687e-06,
+      "loss": 2.041,
+      "step": 19396
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.567884505661058e-06,
+      "loss": 2.0525,
+      "step": 19397
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.567659714648572e-06,
+      "loss": 2.074,
+      "step": 19398
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5674352955515098e-06,
+      "loss": 2.073,
+      "step": 19399
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5672112483757529e-06,
+      "loss": 2.0618,
+      "step": 19400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5669875731271766e-06,
+      "loss": 2.1182,
+      "step": 19401
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5667642698116463e-06,
+      "loss": 2.0779,
+      "step": 19402
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5665413384350209e-06,
+      "loss": 2.1108,
+      "step": 19403
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5663187790031418e-06,
+      "loss": 2.1138,
+      "step": 19404
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.56609659152185e-06,
+      "loss": 2.0776,
+      "step": 19405
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5658747759969722e-06,
+      "loss": 2.0723,
+      "step": 19406
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5656533324343187e-06,
+      "loss": 2.0325,
+      "step": 19407
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5654322608397048e-06,
+      "loss": 2.0135,
+      "step": 19408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5652115612189211e-06,
+      "loss": 2.0753,
+      "step": 19409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.564991233577761e-06,
+      "loss": 2.0763,
+      "step": 19410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5647712779219987e-06,
+      "loss": 2.062,
+      "step": 19411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5645516942574022e-06,
+      "loss": 2.0513,
+      "step": 19412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.564332482589731e-06,
+      "loss": 2.1118,
+      "step": 19413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5641136429247333e-06,
+      "loss": 2.1055,
+      "step": 19414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5638951752681496e-06,
+      "loss": 2.0671,
+      "step": 19415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5636770796257055e-06,
+      "loss": 2.0806,
+      "step": 19416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5634593560031246e-06,
+      "loss": 2.0139,
+      "step": 19417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5632420044061108e-06,
+      "loss": 2.0029,
+      "step": 19418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5630250248403694e-06,
+      "loss": 2.072,
+      "step": 19419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5628084173115875e-06,
+      "loss": 2.0933,
+      "step": 19420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5625921818254474e-06,
+      "loss": 2.1106,
+      "step": 19421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5623763183876162e-06,
+      "loss": 2.1218,
+      "step": 19422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5621608270037574e-06,
+      "loss": 2.1047,
+      "step": 19423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5619457076795217e-06,
+      "loss": 2.0862,
+      "step": 19424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5617309604205515e-06,
+      "loss": 2.0044,
+      "step": 19425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5615165852324753e-06,
+      "loss": 2.0127,
+      "step": 19426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5613025821209186e-06,
+      "loss": 2.1033,
+      "step": 19427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5610889510914872e-06,
+      "loss": 2.0811,
+      "step": 19428
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5608756921497896e-06,
+      "loss": 2.0613,
+      "step": 19429
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5606628053014167e-06,
+      "loss": 2.0728,
+      "step": 19430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5604502905519485e-06,
+      "loss": 2.0535,
+      "step": 19431
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.560238147906961e-06,
+      "loss": 2.0779,
+      "step": 19432
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5600263773720164e-06,
+      "loss": 2.0635,
+      "step": 19433
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.55981497895267e-06,
+      "loss": 2.0396,
+      "step": 19434
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5596039526544609e-06,
+      "loss": 2.0901,
+      "step": 19435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5593932984829245e-06,
+      "loss": 2.0793,
+      "step": 19436
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.55918301644359e-06,
+      "loss": 2.1182,
+      "step": 19437
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5589731065419644e-06,
+      "loss": 2.1162,
+      "step": 19438
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5587635687835567e-06,
+      "loss": 2.093,
+      "step": 19439
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5585544031738625e-06,
+      "loss": 2.0989,
+      "step": 19440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5583456097183641e-06,
+      "loss": 2.0732,
+      "step": 19441
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5581371884225388e-06,
+      "loss": 2.0632,
+      "step": 19442
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.557929139291852e-06,
+      "loss": 2.0532,
+      "step": 19443
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.557721462331758e-06,
+      "loss": 2.0491,
+      "step": 19444
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5575141575477041e-06,
+      "loss": 2.0637,
+      "step": 19445
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5573072249451275e-06,
+      "loss": 2.043,
+      "step": 19446
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5571006645294538e-06,
+      "loss": 2.0784,
+      "step": 19447
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5568944763060971e-06,
+      "loss": 2.0859,
+      "step": 19448
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5566886602804698e-06,
+      "loss": 2.116,
+      "step": 19449
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5564832164579657e-06,
+      "loss": 2.0706,
+      "step": 19450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5562781448439738e-06,
+      "loss": 1.9832,
+      "step": 19451
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5560734454438702e-06,
+      "loss": 1.9532,
+      "step": 19452
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5558691182630253e-06,
+      "loss": 2.0645,
+      "step": 19453
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5556651633067948e-06,
+      "loss": 2.0632,
+      "step": 19454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5554615805805294e-06,
+      "loss": 2.0374,
+      "step": 19455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.555258370089565e-06,
+      "loss": 2.0366,
+      "step": 19456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.555055531839234e-06,
+      "loss": 2.1055,
+      "step": 19457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.554853065834852e-06,
+      "loss": 2.0706,
+      "step": 19458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5546509720817314e-06,
+      "loss": 2.0391,
+      "step": 19459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5544492505851715e-06,
+      "loss": 2.0481,
+      "step": 19460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5542479013504613e-06,
+      "loss": 2.0796,
+      "step": 19461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5540469243828814e-06,
+      "loss": 2.0781,
+      "step": 19462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.553846319687703e-06,
+      "loss": 2.0566,
+      "step": 19463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.553646087270183e-06,
+      "loss": 2.054,
+      "step": 19464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.553446227135576e-06,
+      "loss": 2.0947,
+      "step": 19465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5532467392891225e-06,
+      "loss": 2.0857,
+      "step": 19466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5530476237360535e-06,
+      "loss": 2.0935,
+      "step": 19467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5528488804815916e-06,
+      "loss": 2.0679,
+      "step": 19468
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.552650509530944e-06,
+      "loss": 2.1096,
+      "step": 19469
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5524525108893184e-06,
+      "loss": 2.0869,
+      "step": 19470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5522548845619055e-06,
+      "loss": 1.9844,
+      "step": 19471
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5520576305538882e-06,
+      "loss": 1.9927,
+      "step": 19472
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.551860748870435e-06,
+      "loss": 2.0667,
+      "step": 19473
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5516642395167141e-06,
+      "loss": 2.0352,
+      "step": 19474
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.551468102497876e-06,
+      "loss": 2.0276,
+      "step": 19475
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5512723378190669e-06,
+      "loss": 2.0315,
+      "step": 19476
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5510769454854174e-06,
+      "loss": 2.1357,
+      "step": 19477
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5508819255020517e-06,
+      "loss": 2.1299,
+      "step": 19478
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.550687277874086e-06,
+      "loss": 2.0752,
+      "step": 19479
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5504930026066245e-06,
+      "loss": 2.0781,
+      "step": 19480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5502990997047594e-06,
+      "loss": 2.0925,
+      "step": 19481
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5501055691735788e-06,
+      "loss": 2.0815,
+      "step": 19482
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5499124110181534e-06,
+      "loss": 2.0698,
+      "step": 19483
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5497196252435526e-06,
+      "loss": 2.0901,
+      "step": 19484
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.549527211854832e-06,
+      "loss": 2.0417,
+      "step": 19485
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5493351708570347e-06,
+      "loss": 2.0686,
+      "step": 19486
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5491435022551998e-06,
+      "loss": 2.0881,
+      "step": 19487
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5489522060543498e-06,
+      "loss": 2.0303,
+      "step": 19488
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5487612822595043e-06,
+      "loss": 2.0627,
+      "step": 19489
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5485707308756692e-06,
+      "loss": 2.061,
+      "step": 19490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.548380551907842e-06,
+      "loss": 2.1006,
+      "step": 19491
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5481907453610072e-06,
+      "loss": 2.137,
+      "step": 19492
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.548001311240146e-06,
+      "loss": 2.0601,
+      "step": 19493
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.547812249550224e-06,
+      "loss": 2.0291,
+      "step": 19494
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5476235602962008e-06,
+      "loss": 2.0356,
+      "step": 19495
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5474352434830225e-06,
+      "loss": 2.0413,
+      "step": 19496
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5472472991156284e-06,
+      "loss": 2.0981,
+      "step": 19497
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5470597271989461e-06,
+      "loss": 2.1069,
+      "step": 19498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.546872527737897e-06,
+      "loss": 2.0486,
+      "step": 19499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5466857007373885e-06,
+      "loss": 2.0386,
+      "step": 19500
+    },
+    {
+      "Bleu_1": 0.7730373409384637,
+      "Bleu_2": 0.6126843931343607,
+      "Bleu_3": 0.4712432870940136,
+      "Bleu_4": 0.35822423810586307,
+      "CIDEr": 1.2028906931826104,
+      "METEOR": 0.2779136560378059,
+      "ROUGE_L": 0.5593698575515516,
+      "epoch": 0.97,
+      "step": 19500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5464992462023186e-06,
+      "loss": 2.0056,
+      "step": 19501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5463131641375783e-06,
+      "loss": 2.0305,
+      "step": 19502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5461274545480487e-06,
+      "loss": 2.0591,
+      "step": 19503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5459421174385974e-06,
+      "loss": 2.0955,
+      "step": 19504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5457571528140873e-06,
+      "loss": 2.0691,
+      "step": 19505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5455725606793664e-06,
+      "loss": 2.0764,
+      "step": 19506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5453883410392754e-06,
+      "loss": 2.0706,
+      "step": 19507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5452044938986477e-06,
+      "loss": 2.0366,
+      "step": 19508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5450210192623021e-06,
+      "loss": 2.0774,
+      "step": 19509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5448379171350518e-06,
+      "loss": 2.0576,
+      "step": 19510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5446551875216962e-06,
+      "loss": 2.0764,
+      "step": 19511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.54447283042703e-06,
+      "loss": 2.0554,
+      "step": 19512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5442908458558325e-06,
+      "loss": 2.0803,
+      "step": 19513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5441092338128781e-06,
+      "loss": 2.0669,
+      "step": 19514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5439279943029283e-06,
+      "loss": 2.0618,
+      "step": 19515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5437471273307358e-06,
+      "loss": 2.0422,
+      "step": 19516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5435666329010434e-06,
+      "loss": 2.0562,
+      "step": 19517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5433865110185875e-06,
+      "loss": 2.0537,
+      "step": 19518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5432067616880874e-06,
+      "loss": 2.0835,
+      "step": 19519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5430273849142578e-06,
+      "loss": 2.0745,
+      "step": 19520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5428483807018052e-06,
+      "loss": 2.0566,
+      "step": 19521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5426697490554187e-06,
+      "loss": 1.999,
+      "step": 19522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.54249148997979e-06,
+      "loss": 2.0623,
+      "step": 19523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5423136034795864e-06,
+      "loss": 2.0718,
+      "step": 19524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5421360895594784e-06,
+      "loss": 2.0813,
+      "step": 19525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5419589482241164e-06,
+      "loss": 2.073,
+      "step": 19526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5417821794781489e-06,
+      "loss": 2.1238,
+      "step": 19527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.541605783326212e-06,
+      "loss": 2.1221,
+      "step": 19528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5414297597729288e-06,
+      "loss": 2.1133,
+      "step": 19529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5412541088229153e-06,
+      "loss": 2.1243,
+      "step": 19530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5410788304807816e-06,
+      "loss": 2.0757,
+      "step": 19531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5409039247511202e-06,
+      "loss": 2.0408,
+      "step": 19532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5407293916385213e-06,
+      "loss": 2.0837,
+      "step": 19533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5405552311475557e-06,
+      "loss": 2.0115,
+      "step": 19534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5403814432827968e-06,
+      "loss": 2.0544,
+      "step": 19535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.540208028048799e-06,
+      "loss": 2.0358,
+      "step": 19536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.540034985450114e-06,
+      "loss": 2.0745,
+      "step": 19537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5398623154912728e-06,
+      "loss": 2.0757,
+      "step": 19538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5396900181768087e-06,
+      "loss": 2.0398,
+      "step": 19539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5395180935112379e-06,
+      "loss": 2.0024,
+      "step": 19540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5393465414990688e-06,
+      "loss": 2.0168,
+      "step": 19541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.539175362144799e-06,
+      "loss": 2.0332,
+      "step": 19542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.539004555452922e-06,
+      "loss": 2.0781,
+      "step": 19543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5388341214279138e-06,
+      "loss": 2.0837,
+      "step": 19544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.538664060074243e-06,
+      "loss": 2.0449,
+      "step": 19545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5384943713963722e-06,
+      "loss": 2.0583,
+      "step": 19546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5383250553987466e-06,
+      "loss": 2.0479,
+      "step": 19547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5381561120858123e-06,
+      "loss": 2.0662,
+      "step": 19548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5379875414619944e-06,
+      "loss": 2.0493,
+      "step": 19549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5378193435317172e-06,
+      "loss": 2.0488,
+      "step": 19550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5376515182993895e-06,
+      "loss": 2.0898,
+      "step": 19551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5374840657694123e-06,
+      "loss": 2.0601,
+      "step": 19552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5373169859461773e-06,
+      "loss": 2.0862,
+      "step": 19553
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5371502788340675e-06,
+      "loss": 2.0933,
+      "step": 19554
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.536983944437451e-06,
+      "loss": 2.0725,
+      "step": 19555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.536817982760695e-06,
+      "loss": 2.0596,
+      "step": 19556
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.536652393808145e-06,
+      "loss": 2.0818,
+      "step": 19557
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5364871775841482e-06,
+      "loss": 2.0874,
+      "step": 19558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5363223340930361e-06,
+      "loss": 2.0527,
+      "step": 19559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.53615786333913e-06,
+      "loss": 2.0662,
+      "step": 19560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5359937653267465e-06,
+      "loss": 2.0818,
+      "step": 19561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5358300400601854e-06,
+      "loss": 2.1238,
+      "step": 19562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.535666687543742e-06,
+      "loss": 2.1123,
+      "step": 19563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5355037077816991e-06,
+      "loss": 2.0957,
+      "step": 19564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5353411007783303e-06,
+      "loss": 2.0417,
+      "step": 19565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5351788665379018e-06,
+      "loss": 2.0759,
+      "step": 19566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5350170050646655e-06,
+      "loss": 2.1182,
+      "step": 19567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.534855516362868e-06,
+      "loss": 2.0845,
+      "step": 19568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.534694400436744e-06,
+      "loss": 2.0452,
+      "step": 19569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5345336572905155e-06,
+      "loss": 1.9985,
+      "step": 19570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5343732869284038e-06,
+      "loss": 2.0791,
+      "step": 19571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5342132893546073e-06,
+      "loss": 2.0686,
+      "step": 19572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5340536645733277e-06,
+      "loss": 2.0908,
+      "step": 19573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5338944125887484e-06,
+      "loss": 2.0647,
+      "step": 19574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5337355334050457e-06,
+      "loss": 2.0991,
+      "step": 19575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5335770270263849e-06,
+      "loss": 2.0962,
+      "step": 19576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5334188934569243e-06,
+      "loss": 2.0823,
+      "step": 19577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5332611327008122e-06,
+      "loss": 2.0845,
+      "step": 19578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.533103744762182e-06,
+      "loss": 2.0786,
+      "step": 19579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5329467296451638e-06,
+      "loss": 2.1042,
+      "step": 19580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5327900873538742e-06,
+      "loss": 2.103,
+      "step": 19581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5326338178924215e-06,
+      "loss": 2.1101,
+      "step": 19582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5324779212649026e-06,
+      "loss": 2.0906,
+      "step": 19583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5323223974754059e-06,
+      "loss": 2.0664,
+      "step": 19584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5321672465280095e-06,
+      "loss": 2.075,
+      "step": 19585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.532012468426784e-06,
+      "loss": 2.0896,
+      "step": 19586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5318580631757872e-06,
+      "loss": 2.0615,
+      "step": 19587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5317040307790665e-06,
+      "loss": 2.064,
+      "step": 19588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5315503712406632e-06,
+      "loss": 2.1147,
+      "step": 19589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5313970845646075e-06,
+      "loss": 2.1155,
+      "step": 19590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5312441707549179e-06,
+      "loss": 2.0464,
+      "step": 19591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.531091629815603e-06,
+      "loss": 2.0408,
+      "step": 19592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5309394617506661e-06,
+      "loss": 2.05,
+      "step": 19593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5307876665640957e-06,
+      "loss": 2.0803,
+      "step": 19594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.530636244259872e-06,
+      "loss": 2.0728,
+      "step": 19595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5304851948419684e-06,
+      "loss": 2.0693,
+      "step": 19596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5303345183143435e-06,
+      "loss": 2.0862,
+      "step": 19597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5301842146809492e-06,
+      "loss": 2.1038,
+      "step": 19598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.530034283945727e-06,
+      "loss": 2.0942,
+      "step": 19599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.529884726112611e-06,
+      "loss": 2.0444,
+      "step": 19600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.529735541185519e-06,
+      "loss": 2.0251,
+      "step": 19601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5295867291683668e-06,
+      "loss": 1.9868,
+      "step": 19602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5294382900650527e-06,
+      "loss": 2.113,
+      "step": 19603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5292902238794738e-06,
+      "loss": 2.1152,
+      "step": 19604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5291425306155115e-06,
+      "loss": 2.0918,
+      "step": 19605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5289952102770383e-06,
+      "loss": 2.0369,
+      "step": 19606
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5288482628679157e-06,
+      "loss": 2.0312,
+      "step": 19607
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5287016883920008e-06,
+      "loss": 2.0454,
+      "step": 19608
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.528555486853137e-06,
+      "loss": 2.0784,
+      "step": 19609
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5284096582551548e-06,
+      "loss": 2.106,
+      "step": 19610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5282642026018828e-06,
+      "loss": 2.0532,
+      "step": 19611
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5281191198971312e-06,
+      "loss": 2.0649,
+      "step": 19612
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5279744101447085e-06,
+      "loss": 2.05,
+      "step": 19613
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5278300733484068e-06,
+      "loss": 2.0391,
+      "step": 19614
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5276861095120114e-06,
+      "loss": 2.0447,
+      "step": 19615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5275425186392994e-06,
+      "loss": 2.0737,
+      "step": 19616
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5273993007340343e-06,
+      "loss": 2.0591,
+      "step": 19617
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5272564557999746e-06,
+      "loss": 2.0342,
+      "step": 19618
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5271139838408627e-06,
+      "loss": 2.0737,
+      "step": 19619
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5269718848604368e-06,
+      "loss": 2.0491,
+      "step": 19620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5268301588624228e-06,
+      "loss": 2.0698,
+      "step": 19621
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5266888058505374e-06,
+      "loss": 2.0703,
+      "step": 19622
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5265478258284877e-06,
+      "loss": 2.0767,
+      "step": 19623
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5264072187999706e-06,
+      "loss": 2.0737,
+      "step": 19624
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5262669847686716e-06,
+      "loss": 2.0752,
+      "step": 19625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5261271237382713e-06,
+      "loss": 2.0487,
+      "step": 19626
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5259876357124348e-06,
+      "loss": 1.9995,
+      "step": 19627
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5258485206948208e-06,
+      "loss": 2.0352,
+      "step": 19628
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5257097786890801e-06,
+      "loss": 2.0515,
+      "step": 19629
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5255714096988477e-06,
+      "loss": 2.0596,
+      "step": 19630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5254334137277508e-06,
+      "loss": 2.083,
+      "step": 19631
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5252957907794135e-06,
+      "loss": 2.0862,
+      "step": 19632
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5251585408574408e-06,
+      "loss": 1.9944,
+      "step": 19633
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5250216639654318e-06,
+      "loss": 1.9727,
+      "step": 19634
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5248851601069785e-06,
+      "loss": 2.0801,
+      "step": 19635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5247490292856581e-06,
+      "loss": 2.0686,
+      "step": 19636
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5246132715050442e-06,
+      "loss": 2.0623,
+      "step": 19637
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5244778867686909e-06,
+      "loss": 2.0752,
+      "step": 19638
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.524342875080152e-06,
+      "loss": 2.0916,
+      "step": 19639
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.524208236442971e-06,
+      "loss": 2.0525,
+      "step": 19640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.524073970860672e-06,
+      "loss": 2.0586,
+      "step": 19641
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5239400783367808e-06,
+      "loss": 2.0847,
+      "step": 19642
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5238065588748057e-06,
+      "loss": 2.0811,
+      "step": 19643
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5236734124782526e-06,
+      "loss": 2.0588,
+      "step": 19644
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5235406391506084e-06,
+      "loss": 2.105,
+      "step": 19645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5234082388953557e-06,
+      "loss": 2.1016,
+      "step": 19646
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5232762117159682e-06,
+      "loss": 2.1348,
+      "step": 19647
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5231445576159082e-06,
+      "loss": 2.1145,
+      "step": 19648
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.523013276598626e-06,
+      "loss": 2.0305,
+      "step": 19649
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5228823686675692e-06,
+      "loss": 2.0442,
+      "step": 19650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522751833826163e-06,
+      "loss": 2.0269,
+      "step": 19651
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522621672077837e-06,
+      "loss": 1.9969,
+      "step": 19652
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5224918834259993e-06,
+      "loss": 2.0681,
+      "step": 19653
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522362467874061e-06,
+      "loss": 2.0725,
+      "step": 19654
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5222334254254077e-06,
+      "loss": 2.0908,
+      "step": 19655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5221047560834284e-06,
+      "loss": 2.0588,
+      "step": 19656
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5219764598514952e-06,
+      "loss": 2.0972,
+      "step": 19657
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5218485367329754e-06,
+      "loss": 2.0891,
+      "step": 19658
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5217209867312182e-06,
+      "loss": 2.072,
+      "step": 19659
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5215938098495741e-06,
+      "loss": 2.0562,
+      "step": 19660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5214670060913755e-06,
+      "loss": 2.1113,
+      "step": 19661
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.521340575459948e-06,
+      "loss": 2.1052,
+      "step": 19662
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5212145179586056e-06,
+      "loss": 2.0796,
+      "step": 19663
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.521088833590656e-06,
+      "loss": 2.064,
+      "step": 19664
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.520963522359396e-06,
+      "loss": 2.0366,
+      "step": 19665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5208385842681083e-06,
+      "loss": 2.0239,
+      "step": 19666
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.520714019320072e-06,
+      "loss": 2.0391,
+      "step": 19667
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5205898275185527e-06,
+      "loss": 2.0298,
+      "step": 19668
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5204660088668076e-06,
+      "loss": 2.0815,
+      "step": 19669
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5203425633680845e-06,
+      "loss": 2.0466,
+      "step": 19670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5202194910256189e-06,
+      "loss": 2.0706,
+      "step": 19671
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.52009679184264e-06,
+      "loss": 2.0774,
+      "step": 19672
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.519974465822362e-06,
+      "loss": 2.0264,
+      "step": 19673
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5198525129679968e-06,
+      "loss": 2.0127,
+      "step": 19674
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5197309332827425e-06,
+      "loss": 2.0129,
+      "step": 19675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5196097267697829e-06,
+      "loss": 2.0447,
+      "step": 19676
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5194888934323003e-06,
+      "loss": 2.0562,
+      "step": 19677
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5193684332734624e-06,
+      "loss": 2.0686,
+      "step": 19678
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5192483462964302e-06,
+      "loss": 2.0918,
+      "step": 19679
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5191286325043474e-06,
+      "loss": 2.1057,
+      "step": 19680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5190092919003585e-06,
+      "loss": 2.0378,
+      "step": 19681
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5188903244875907e-06,
+      "loss": 2.0447,
+      "step": 19682
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.518771730269165e-06,
+      "loss": 2.0559,
+      "step": 19683
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5186535092481906e-06,
+      "loss": 2.0713,
+      "step": 19684
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5185356614277684e-06,
+      "loss": 2.0342,
+      "step": 19685
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5184181868109875e-06,
+      "loss": 1.9795,
+      "step": 19686
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5183010854009305e-06,
+      "loss": 2.0574,
+      "step": 19687
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5181843572006666e-06,
+      "loss": 2.0337,
+      "step": 19688
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5180680022132582e-06,
+      "loss": 2.0601,
+      "step": 19689
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5179520204417547e-06,
+      "loss": 2.0737,
+      "step": 19690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5178364118891986e-06,
+      "loss": 2.0366,
+      "step": 19691
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5177211765586225e-06,
+      "loss": 2.0157,
+      "step": 19692
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5176063144530471e-06,
+      "loss": 1.9868,
+      "step": 19693
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5174918255754839e-06,
+      "loss": 1.9993,
+      "step": 19694
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5173777099289366e-06,
+      "loss": 2.1133,
+      "step": 19695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5172639675163997e-06,
+      "loss": 2.0955,
+      "step": 19696
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5171505983408507e-06,
+      "loss": 2.0864,
+      "step": 19697
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5170376024052674e-06,
+      "loss": 2.0923,
+      "step": 19698
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5169249797126106e-06,
+      "loss": 2.0513,
+      "step": 19699
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5168127302658348e-06,
+      "loss": 2.0828,
+      "step": 19700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5167008540678805e-06,
+      "loss": 2.031,
+      "step": 19701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.516589351121686e-06,
+      "loss": 2.0537,
+      "step": 19702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5164782214301736e-06,
+      "loss": 2.0759,
+      "step": 19703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5163674649962561e-06,
+      "loss": 2.0828,
+      "step": 19704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.516257081822841e-06,
+      "loss": 2.0764,
+      "step": 19705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5161470719128196e-06,
+      "loss": 2.0889,
+      "step": 19706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5160374352690792e-06,
+      "loss": 2.0325,
+      "step": 19707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5159281718944928e-06,
+      "loss": 2.0469,
+      "step": 19708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5158192817919298e-06,
+      "loss": 2.0881,
+      "step": 19709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5157107649642393e-06,
+      "loss": 2.0933,
+      "step": 19710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5156026214142743e-06,
+      "loss": 2.0811,
+      "step": 19711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.515494851144864e-06,
+      "loss": 2.0903,
+      "step": 19712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.515387454158838e-06,
+      "loss": 2.0552,
+      "step": 19713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.515280430459014e-06,
+      "loss": 2.062,
+      "step": 19714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5151737800481946e-06,
+      "loss": 2.125,
+      "step": 19715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5150675029291811e-06,
+      "loss": 2.1321,
+      "step": 19716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5149615991047577e-06,
+      "loss": 2.0911,
+      "step": 19717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5148560685777008e-06,
+      "loss": 2.1018,
+      "step": 19718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5147509113507796e-06,
+      "loss": 2.0542,
+      "step": 19719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.514646127426752e-06,
+      "loss": 2.0803,
+      "step": 19720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5145417168083625e-06,
+      "loss": 2.0627,
+      "step": 19721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5144376794983538e-06,
+      "loss": 2.1008,
+      "step": 19722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5143340154994522e-06,
+      "loss": 2.061,
+      "step": 19723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5142307248143755e-06,
+      "loss": 2.033,
+      "step": 19724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.514127807445833e-06,
+      "loss": 2.0667,
+      "step": 19725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5140252633965246e-06,
+      "loss": 2.0645,
+      "step": 19726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.513923092669136e-06,
+      "loss": 2.0688,
+      "step": 19727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5138212952663521e-06,
+      "loss": 2.0759,
+      "step": 19728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5137198711908374e-06,
+      "loss": 2.0859,
+      "step": 19729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5136188204452527e-06,
+      "loss": 2.0562,
+      "step": 19730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.513518143032248e-06,
+      "loss": 2.0769,
+      "step": 19731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5134178389544663e-06,
+      "loss": 2.0554,
+      "step": 19732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.513317908214535e-06,
+      "loss": 2.0864,
+      "step": 19733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5132183508150755e-06,
+      "loss": 2.127,
+      "step": 19734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5131191667586974e-06,
+      "loss": 2.075,
+      "step": 19735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5130203560480038e-06,
+      "loss": 2.0857,
+      "step": 19736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512921918685584e-06,
+      "loss": 2.0432,
+      "step": 19737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5128238546740198e-06,
+      "loss": 2.0381,
+      "step": 19738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5127261640158852e-06,
+      "loss": 2.0632,
+      "step": 19739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5126288467137383e-06,
+      "loss": 2.0488,
+      "step": 19740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512531902770134e-06,
+      "loss": 2.095,
+      "step": 19741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5124353321876133e-06,
+      "loss": 2.0808,
+      "step": 19742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5123391349687077e-06,
+      "loss": 2.0974,
+      "step": 19743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5122433111159416e-06,
+      "loss": 2.0752,
+      "step": 19744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5121478606318267e-06,
+      "loss": 2.1016,
+      "step": 19745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5120527835188691e-06,
+      "loss": 2.074,
+      "step": 19746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5119580797795585e-06,
+      "loss": 2.0291,
+      "step": 19747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5118637494163779e-06,
+      "loss": 2.0247,
+      "step": 19748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5117697924318053e-06,
+      "loss": 2.0308,
+      "step": 19749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5116762088282986e-06,
+      "loss": 2.0669,
+      "step": 19750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.511582998608318e-06,
+      "loss": 2.0642,
+      "step": 19751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5114901617743044e-06,
+      "loss": 2.0732,
+      "step": 19752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5113976983286909e-06,
+      "loss": 2.0598,
+      "step": 19753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5113056082739074e-06,
+      "loss": 2.0747,
+      "step": 19754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5112138916123637e-06,
+      "loss": 2.0715,
+      "step": 19755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5111225483464675e-06,
+      "loss": 2.0811,
+      "step": 19756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5110315784786139e-06,
+      "loss": 2.0369,
+      "step": 19757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5109409820111891e-06,
+      "loss": 2.0781,
+      "step": 19758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5108507589465681e-06,
+      "loss": 2.0569,
+      "step": 19759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5107609092871172e-06,
+      "loss": 2.033,
+      "step": 19760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5106714330351913e-06,
+      "loss": 2.0996,
+      "step": 19761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5105823301931365e-06,
+      "loss": 2.1223,
+      "step": 19762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.510493600763293e-06,
+      "loss": 2.0513,
+      "step": 19763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5104052447479838e-06,
+      "loss": 2.0588,
+      "step": 19764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5103172621495287e-06,
+      "loss": 2.0847,
+      "step": 19765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5102296529702343e-06,
+      "loss": 2.053,
+      "step": 19766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5101424172123967e-06,
+      "loss": 2.0432,
+      "step": 19767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.510055554878303e-06,
+      "loss": 2.0415,
+      "step": 19768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5099690659702342e-06,
+      "loss": 2.0093,
+      "step": 19769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5098829504904555e-06,
+      "loss": 1.9927,
+      "step": 19770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5097972084412264e-06,
+      "loss": 2.0754,
+      "step": 19771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5097118398247956e-06,
+      "loss": 2.0171,
+      "step": 19772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5096268446434008e-06,
+      "loss": 2.0557,
+      "step": 19773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5095422228992739e-06,
+      "loss": 2.0681,
+      "step": 19774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5094579745946295e-06,
+      "loss": 2.0818,
+      "step": 19775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5093740997316796e-06,
+      "loss": 2.0806,
+      "step": 19776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5092905983126236e-06,
+      "loss": 2.0798,
+      "step": 19777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5092074703396501e-06,
+      "loss": 2.0745,
+      "step": 19778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5091247158149391e-06,
+      "loss": 2.0686,
+      "step": 19779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5090423347406635e-06,
+      "loss": 2.0535,
+      "step": 19780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.50896032711898e-06,
+      "loss": 2.0952,
+      "step": 19781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5088786929520403e-06,
+      "loss": 2.0957,
+      "step": 19782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5087974322419862e-06,
+      "loss": 2.0361,
+      "step": 19783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.508716544990949e-06,
+      "loss": 2.0387,
+      "step": 19784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5086360312010473e-06,
+      "loss": 2.0459,
+      "step": 19785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5085558908743956e-06,
+      "loss": 2.0359,
+      "step": 19786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5084761240130928e-06,
+      "loss": 2.0859,
+      "step": 19787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.508396730619232e-06,
+      "loss": 2.0862,
+      "step": 19788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5083177106948964e-06,
+      "loss": 2.0408,
+      "step": 19789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5082390642421545e-06,
+      "loss": 2.0348,
+      "step": 19790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.508160791263073e-06,
+      "loss": 2.1113,
+      "step": 19791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5080828917597016e-06,
+      "loss": 2.0656,
+      "step": 19792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5080053657340842e-06,
+      "loss": 2.0669,
+      "step": 19793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5079282131882539e-06,
+      "loss": 2.05,
+      "step": 19794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507851434124234e-06,
+      "loss": 2.009,
+      "step": 19795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5077750285440362e-06,
+      "loss": 1.967,
+      "step": 19796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507698996449669e-06,
+      "loss": 2.0369,
+      "step": 19797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5076233378431193e-06,
+      "loss": 2.0154,
+      "step": 19798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5075480527263786e-06,
+      "loss": 2.0688,
+      "step": 19799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5074731411014136e-06,
+      "loss": 2.051,
+      "step": 19800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5073986029701947e-06,
+      "loss": 2.0723,
+      "step": 19801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5073244383346734e-06,
+      "loss": 2.0869,
+      "step": 19802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5072506471967952e-06,
+      "loss": 2.0957,
+      "step": 19803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507177229558498e-06,
+      "loss": 2.1084,
+      "step": 19804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507104185421702e-06,
+      "loss": 2.0728,
+      "step": 19805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5070315147883263e-06,
+      "loss": 2.0923,
+      "step": 19806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5069592176602754e-06,
+      "loss": 1.9578,
+      "step": 19807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5068872940394463e-06,
+      "loss": 2.0251,
+      "step": 19808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5068157439277242e-06,
+      "loss": 2.0203,
+      "step": 19809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5067445673269857e-06,
+      "loss": 2.0449,
+      "step": 19810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5066737642390948e-06,
+      "loss": 2.0442,
+      "step": 19811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.506603334665913e-06,
+      "loss": 2.0522,
+      "step": 19812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5065332786092838e-06,
+      "loss": 2.0811,
+      "step": 19813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.506463596071046e-06,
+      "loss": 2.0837,
+      "step": 19814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5063942870530245e-06,
+      "loss": 2.0923,
+      "step": 19815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5063253515570413e-06,
+      "loss": 2.0723,
+      "step": 19816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5062567895849e-06,
+      "loss": 2.0583,
+      "step": 19817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5061886011383992e-06,
+      "loss": 2.0642,
+      "step": 19818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5061207862193289e-06,
+      "loss": 2.124,
+      "step": 19819
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5060533448294661e-06,
+      "loss": 2.1138,
+      "step": 19820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5059862769705812e-06,
+      "loss": 2.0669,
+      "step": 19821
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5059195826444294e-06,
+      "loss": 2.0387,
+      "step": 19822
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5058532618527642e-06,
+      "loss": 2.0737,
+      "step": 19823
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505787314597321e-06,
+      "loss": 2.0896,
+      "step": 19824
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5057217408798302e-06,
+      "loss": 2.1096,
+      "step": 19825
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5056565407020117e-06,
+      "loss": 2.0952,
+      "step": 19826
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505591714065576e-06,
+      "loss": 2.084,
+      "step": 19827
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5055272609722221e-06,
+      "loss": 2.0701,
+      "step": 19828
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5054631814236397e-06,
+      "loss": 2.0488,
+      "step": 19829
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505399475421511e-06,
+      "loss": 2.062,
+      "step": 19830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505336142967505e-06,
+      "loss": 2.0508,
+      "step": 19831
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505273184063285e-06,
+      "loss": 2.0457,
+      "step": 19832
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5052105987104996e-06,
+      "loss": 2.0618,
+      "step": 19833
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5051483869107895e-06,
+      "loss": 2.0862,
+      "step": 19834
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5050865486657883e-06,
+      "loss": 2.0737,
+      "step": 19835
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5050250839771163e-06,
+      "loss": 2.0867,
+      "step": 19836
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5049639928463854e-06,
+      "loss": 2.0217,
+      "step": 19837
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504903275275198e-06,
+      "loss": 1.9932,
+      "step": 19838
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5048429312651473e-06,
+      "loss": 2.033,
+      "step": 19839
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5047829608178143e-06,
+      "loss": 2.0269,
+      "step": 19840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5047233639347706e-06,
+      "loss": 2.1028,
+      "step": 19841
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504664140617582e-06,
+      "loss": 2.0645,
+      "step": 19842
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5046052908678004e-06,
+      "loss": 2.0598,
+      "step": 19843
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5045468146869678e-06,
+      "loss": 2.0532,
+      "step": 19844
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5044887120766182e-06,
+      "loss": 2.1067,
+      "step": 19845
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5044309830382785e-06,
+      "loss": 2.1135,
+      "step": 19846
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5043736275734577e-06,
+      "loss": 2.094,
+      "step": 19847
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504316645683661e-06,
+      "loss": 2.0776,
+      "step": 19848
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5042600373703841e-06,
+      "loss": 2.0618,
+      "step": 19849
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5042038026351125e-06,
+      "loss": 1.9991,
+      "step": 19850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504147941479318e-06,
+      "loss": 2.0818,
+      "step": 19851
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504092453904468e-06,
+      "loss": 2.0688,
+      "step": 19852
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.504037339912015e-06,
+      "loss": 2.0801,
+      "step": 19853
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5039825995034074e-06,
+      "loss": 2.0901,
+      "step": 19854
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503928232680079e-06,
+      "loss": 2.1055,
+      "step": 19855
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5038742394434558e-06,
+      "loss": 2.1023,
+      "step": 19856
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503820619794951e-06,
+      "loss": 2.0479,
+      "step": 19857
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503767373735976e-06,
+      "loss": 2.0481,
+      "step": 19858
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5037145012679238e-06,
+      "loss": 2.0359,
+      "step": 19859
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503662002392182e-06,
+      "loss": 2.025,
+      "step": 19860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5036098771101247e-06,
+      "loss": 2.0422,
+      "step": 19861
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5035581254231223e-06,
+      "loss": 2.05,
+      "step": 19862
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5035067473325304e-06,
+      "loss": 2.0918,
+      "step": 19863
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5034557428396962e-06,
+      "loss": 2.0579,
+      "step": 19864
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5034051119459584e-06,
+      "loss": 2.0325,
+      "step": 19865
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5033548546526431e-06,
+      "loss": 2.0808,
+      "step": 19866
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503304970961069e-06,
+      "loss": 2.1011,
+      "step": 19867
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5032554608725452e-06,
+      "loss": 2.104,
+      "step": 19868
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5032063243883688e-06,
+      "loss": 2.075,
+      "step": 19869
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503157561509827e-06,
+      "loss": 2.0718,
+      "step": 19870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.503109172238204e-06,
+      "loss": 2.0405,
+      "step": 19871
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5030611565747603e-06,
+      "loss": 2.051,
+      "step": 19872
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5030135145207634e-06,
+      "loss": 2.1008,
+      "step": 19873
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5029662460774574e-06,
+      "loss": 2.0925,
+      "step": 19874
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5029193512460826e-06,
+      "loss": 2.1074,
+      "step": 19875
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5028728300278699e-06,
+      "loss": 2.0724,
+      "step": 19876
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.50282668242404e-06,
+      "loss": 2.0457,
+      "step": 19877
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5027809084358002e-06,
+      "loss": 2.0833,
+      "step": 19878
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5027355080643545e-06,
+      "loss": 2.0312,
+      "step": 19879
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5026904813108921e-06,
+      "loss": 2.0183,
+      "step": 19880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.502645828176592e-06,
+      "loss": 2.0376,
+      "step": 19881
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5026015486626266e-06,
+      "loss": 2.0308,
+      "step": 19882
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5025576427701582e-06,
+      "loss": 2.0916,
+      "step": 19883
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5025141105003343e-06,
+      "loss": 2.1035,
+      "step": 19884
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5024709518543022e-06,
+      "loss": 2.0547,
+      "step": 19885
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5024281668331879e-06,
+      "loss": 2.0449,
+      "step": 19886
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5023857554381188e-06,
+      "loss": 2.0793,
+      "step": 19887
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.502343717670202e-06,
+      "loss": 2.0779,
+      "step": 19888
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5023020535305437e-06,
+      "loss": 2.1121,
+      "step": 19889
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5022607630202344e-06,
+      "loss": 2.0898,
+      "step": 19890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5022198461403586e-06,
+      "loss": 2.0886,
+      "step": 19891
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5021793028919868e-06,
+      "loss": 2.0669,
+      "step": 19892
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5021391332761848e-06,
+      "loss": 2.0439,
+      "step": 19893
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5020993372940052e-06,
+      "loss": 2.0317,
+      "step": 19894
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5020599149464903e-06,
+      "loss": 2.0977,
+      "step": 19895
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5020208662346746e-06,
+      "loss": 2.062,
+      "step": 19896
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5019821911595821e-06,
+      "loss": 2.0652,
+      "step": 19897
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5019438897222303e-06,
+      "loss": 2.0542,
+      "step": 19898
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5019059619236186e-06,
+      "loss": 2.1277,
+      "step": 19899
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5018684077647427e-06,
+      "loss": 2.0903,
+      "step": 19900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.50183122724659e-06,
+      "loss": 2.1243,
+      "step": 19901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5017944203701333e-06,
+      "loss": 2.1128,
+      "step": 19902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5017579871363386e-06,
+      "loss": 1.9739,
+      "step": 19903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5017219275461596e-06,
+      "loss": 2.0061,
+      "step": 19904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.501686241600546e-06,
+      "loss": 2.1106,
+      "step": 19905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5016509293004288e-06,
+      "loss": 2.0803,
+      "step": 19906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.501615990646737e-06,
+      "loss": 2.095,
+      "step": 19907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5015814256403866e-06,
+      "loss": 2.0793,
+      "step": 19908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5015472342822817e-06,
+      "loss": 2.0356,
+      "step": 19909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5015134165733235e-06,
+      "loss": 2.0532,
+      "step": 19910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5014799725143946e-06,
+      "loss": 2.0862,
+      "step": 19911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5014469021063742e-06,
+      "loss": 2.0923,
+      "step": 19912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5014142053501267e-06,
+      "loss": 2.054,
+      "step": 19913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5013818822465131e-06,
+      "loss": 2.0376,
+      "step": 19914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5013499327963808e-06,
+      "loss": 2.0327,
+      "step": 19915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5013183570005662e-06,
+      "loss": 2.0574,
+      "step": 19916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5012871548598967e-06,
+      "loss": 2.0684,
+      "step": 19917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5012563263751917e-06,
+      "loss": 2.0808,
+      "step": 19918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5012258715472587e-06,
+      "loss": 2.0742,
+      "step": 19919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5011957903768974e-06,
+      "loss": 2.0681,
+      "step": 19920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.501166082864897e-06,
+      "loss": 2.041,
+      "step": 19921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5011367490120354e-06,
+      "loss": 2.0518,
+      "step": 19922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5011077888190817e-06,
+      "loss": 2.0254,
+      "step": 19923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.501079202286797e-06,
+      "loss": 2.0518,
+      "step": 19924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5010509894159274e-06,
+      "loss": 2.0667,
+      "step": 19925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5010231502072192e-06,
+      "loss": 2.071,
+      "step": 19926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5009956846613948e-06,
+      "loss": 2.1086,
+      "step": 19927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500968592779179e-06,
+      "loss": 2.0808,
+      "step": 19928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5009418745612823e-06,
+      "loss": 2.0576,
+      "step": 19929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500915530008403e-06,
+      "loss": 2.0305,
+      "step": 19930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5008895591212356e-06,
+      "loss": 2.0627,
+      "step": 19931
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500863961900456e-06,
+      "loss": 2.0891,
+      "step": 19932
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5008387383467404e-06,
+      "loss": 2.0618,
+      "step": 19933
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5008138884607498e-06,
+      "loss": 2.0581,
+      "step": 19934
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5007894122431304e-06,
+      "loss": 2.0095,
+      "step": 19935
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5007653096945301e-06,
+      "loss": 2.0183,
+      "step": 19936
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5007415808155783e-06,
+      "loss": 2.0352,
+      "step": 19937
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500718225606898e-06,
+      "loss": 2.0811,
+      "step": 19938
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5006952440691035e-06,
+      "loss": 2.0266,
+      "step": 19939
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5006726362027927e-06,
+      "loss": 2.021,
+      "step": 19940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500650402008562e-06,
+      "loss": 2.0571,
+      "step": 19941
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5006285414869941e-06,
+      "loss": 2.0471,
+      "step": 19942
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5006070546386601e-06,
+      "loss": 2.0703,
+      "step": 19943
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005859414641266e-06,
+      "loss": 2.0554,
+      "step": 19944
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005652019639459e-06,
+      "loss": 2.1062,
+      "step": 19945
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005448361386615e-06,
+      "loss": 2.0486,
+      "step": 19946
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005248439888092e-06,
+      "loss": 2.0403,
+      "step": 19947
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005052255149089e-06,
+      "loss": 2.0222,
+      "step": 19948
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004859807174797e-06,
+      "loss": 2.0854,
+      "step": 19949
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004671095970234e-06,
+      "loss": 2.0945,
+      "step": 19950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004486121540378e-06,
+      "loss": 2.0544,
+      "step": 19951
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004304883890057e-06,
+      "loss": 2.0493,
+      "step": 19952
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004127383024036e-06,
+      "loss": 2.0986,
+      "step": 19953
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5003953618946958e-06,
+      "loss": 2.0308,
+      "step": 19954
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5003783591663372e-06,
+      "loss": 2.0911,
+      "step": 19955
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5003617301177759e-06,
+      "loss": 2.0674,
+      "step": 19956
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5003454747494461e-06,
+      "loss": 2.0803,
+      "step": 19957
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5003295930617762e-06,
+      "loss": 2.0681,
+      "step": 19958
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500314085055182e-06,
+      "loss": 2.0518,
+      "step": 19959
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500298950730069e-06,
+      "loss": 2.0496,
+      "step": 19960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002841900868347e-06,
+      "loss": 2.1003,
+      "step": 19961
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002698031258654e-06,
+      "loss": 2.1157,
+      "step": 19962
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500255789847541e-06,
+      "loss": 2.1194,
+      "step": 19963
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002421502522261e-06,
+      "loss": 2.1511,
+      "step": 19964
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500228884340279e-06,
+      "loss": 2.0745,
+      "step": 19965
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500215992112049e-06,
+      "loss": 2.0967,
+      "step": 19966
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002034735678728e-06,
+      "loss": 2.0295,
+      "step": 19967
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001913287080782e-06,
+      "loss": 2.0391,
+      "step": 19968
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001795575329852e-06,
+      "loss": 2.0132,
+      "step": 19969
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001681600429035e-06,
+      "loss": 2.0095,
+      "step": 19970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001571362381278e-06,
+      "loss": 2.0876,
+      "step": 19971
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500146486118953e-06,
+      "loss": 2.1045,
+      "step": 19972
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500136209685652e-06,
+      "loss": 2.0439,
+      "step": 19973
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001263069384979e-06,
+      "loss": 2.0464,
+      "step": 19974
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001167778777508e-06,
+      "loss": 2.0938,
+      "step": 19975
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001076225036587e-06,
+      "loss": 2.0801,
+      "step": 19976
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000988408164644e-06,
+      "loss": 2.1145,
+      "step": 19977
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000904328163967e-06,
+      "loss": 2.092,
+      "step": 19978
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000823985036733e-06,
+      "loss": 2.0469,
+      "step": 19979
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000747378785106e-06,
+      "loss": 2.043,
+      "step": 19980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500067450941104e-06,
+      "loss": 2.0557,
+      "step": 19981
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000605376916477e-06,
+      "loss": 2.0392,
+      "step": 19982
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000539981303218e-06,
+      "loss": 2.0093,
+      "step": 19983
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500047832257298e-06,
+      "loss": 2.0256,
+      "step": 19984
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000420400727411e-06,
+      "loss": 2.0435,
+      "step": 19985
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000366215767958e-06,
+      "loss": 2.0354,
+      "step": 19986
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000315767696103e-06,
+      "loss": 2.0532,
+      "step": 19987
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000269056513166e-06,
+      "loss": 2.0439,
+      "step": 19988
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000226082220341e-06,
+      "loss": 2.0781,
+      "step": 19989
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000186844818782e-06,
+      "loss": 2.0461,
+      "step": 19990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.50001513443095e-06,
+      "loss": 2.0066,
+      "step": 19991
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000119580693428e-06,
+      "loss": 2.019,
+      "step": 19992
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000091553971402e-06,
+      "loss": 2.0278,
+      "step": 19993
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000067264144169e-06,
+      "loss": 2.0371,
+      "step": 19994
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000046711212362e-06,
+      "loss": 2.073,
+      "step": 19995
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000029895176515e-06,
+      "loss": 2.0586,
+      "step": 19996
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000016816037044e-06,
+      "loss": 2.1045,
+      "step": 19997
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500000747379433e-06,
+      "loss": 2.1023,
+      "step": 19998
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500000186844861e-06,
+      "loss": 2.0532,
+      "step": 19999
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000000000000013e-06,
+      "loss": 2.0642,
+      "step": 20000
+    },
+    {
+      "Bleu_1": 0.772391398497911,
+      "Bleu_2": 0.6111344295059981,
+      "Bleu_3": 0.4690967332761042,
+      "Bleu_4": 0.3555337692837351,
+      "CIDEr": 1.1978637713904554,
+      "METEOR": 0.2776644388346762,
+      "ROUGE_L": 0.5578328489998249,
+      "epoch": 1.0,
+      "step": 20000
+    }
+  ],
+  "max_steps": 20000,
+  "num_train_epochs": 9223372036854775807,
+  "total_flos": 6.235587647768248e+20,
+  "trial_name": null,
+  "trial_params": null
+}