diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,49909 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 8314,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-08,
+      "loss": 3.1785,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-08,
+      "loss": 3.2695,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-08,
+      "loss": 3.091,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-08,
+      "loss": 3.172,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 3.1976,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-07,
+      "loss": 3.2198,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4e-07,
+      "loss": 3.261,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6e-07,
+      "loss": 3.2313,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-07,
+      "loss": 3.1219,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 3.2319,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2e-07,
+      "loss": 3.2158,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4000000000000003e-07,
+      "loss": 3.0471,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6e-07,
+      "loss": 3.1218,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8e-07,
+      "loss": 3.1062,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 3.0907,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2e-07,
+      "loss": 3.2499,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": 3.1516,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-07,
+      "loss": 3.146,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8e-07,
+      "loss": 3.1946,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 3.2708,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2000000000000006e-07,
+      "loss": 3.2105,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4e-07,
+      "loss": 3.2715,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6000000000000004e-07,
+      "loss": 3.1529,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.800000000000001e-07,
+      "loss": 2.9737,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 3.1571,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.2e-07,
+      "loss": 3.1964,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4e-07,
+      "loss": 3.1716,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6e-07,
+      "loss": 3.0507,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.800000000000001e-07,
+      "loss": 3.1547,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 3.2577,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.200000000000001e-07,
+      "loss": 3.1158,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.4e-07,
+      "loss": 3.1809,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.6e-07,
+      "loss": 3.2016,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 3.186,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 3.1904,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2e-07,
+      "loss": 3.1372,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.4e-07,
+      "loss": 3.2298,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.6e-07,
+      "loss": 3.0846,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.8e-07,
+      "loss": 3.1781,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 3.1607,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.200000000000001e-07,
+      "loss": 3.1568,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 3.0852,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.6e-07,
+      "loss": 3.0359,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.8e-07,
+      "loss": 3.1305,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.000000000000001e-07,
+      "loss": 3.2682,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.200000000000001e-07,
+      "loss": 3.2945,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.400000000000001e-07,
+      "loss": 3.2281,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.600000000000001e-07,
+      "loss": 3.1374,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.800000000000001e-07,
+      "loss": 3.2815,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 3.113,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.02e-06,
+      "loss": 3.1565,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.04e-06,
+      "loss": 3.204,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.06e-06,
+      "loss": 3.3244,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.08e-06,
+      "loss": 3.2843,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1e-06,
+      "loss": 3.2232,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.12e-06,
+      "loss": 3.2925,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.14e-06,
+      "loss": 3.0409,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1600000000000001e-06,
+      "loss": 3.2141,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1800000000000001e-06,
+      "loss": 3.0975,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 3.0773,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2200000000000002e-06,
+      "loss": 3.1929,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 3.0544,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.26e-06,
+      "loss": 3.265,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.28e-06,
+      "loss": 3.2144,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3e-06,
+      "loss": 3.0359,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.32e-06,
+      "loss": 3.2082,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.34e-06,
+      "loss": 3.0978,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3600000000000001e-06,
+      "loss": 3.1745,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 3.1885,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 3.146,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.42e-06,
+      "loss": 3.244,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.44e-06,
+      "loss": 3.1626,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.46e-06,
+      "loss": 3.1942,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.48e-06,
+      "loss": 3.181,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5e-06,
+      "loss": 3.2878,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.52e-06,
+      "loss": 3.0503,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.54e-06,
+      "loss": 3.0305,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.56e-06,
+      "loss": 3.1652,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5800000000000001e-06,
+      "loss": 3.1214,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 3.0848,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6200000000000002e-06,
+      "loss": 3.1586,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6400000000000002e-06,
+      "loss": 3.1994,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6600000000000002e-06,
+      "loss": 2.8981,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6800000000000002e-06,
+      "loss": 3.1719,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 3.0866,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.72e-06,
+      "loss": 2.9877,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.74e-06,
+      "loss": 2.9919,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.76e-06,
+      "loss": 3.1237,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7800000000000001e-06,
+      "loss": 3.2316,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 3.1247,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8200000000000002e-06,
+      "loss": 3.1272,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8400000000000002e-06,
+      "loss": 3.1617,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8600000000000002e-06,
+      "loss": 3.1933,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8800000000000002e-06,
+      "loss": 3.2075,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9000000000000002e-06,
+      "loss": 3.0874,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9200000000000003e-06,
+      "loss": 3.178,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.94e-06,
+      "loss": 3.067,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9600000000000003e-06,
+      "loss": 3.1208,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.98e-06,
+      "loss": 3.053,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 3.075,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.02e-06,
+      "loss": 3.208,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.04e-06,
+      "loss": 3.0711,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.06e-06,
+      "loss": 3.0657,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.08e-06,
+      "loss": 3.0076,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 3.1029,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.12e-06,
+      "loss": 3.1759,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1400000000000003e-06,
+      "loss": 3.0021,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.16e-06,
+      "loss": 3.1169,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1800000000000003e-06,
+      "loss": 3.0419,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2e-06,
+      "loss": 3.1203,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2200000000000003e-06,
+      "loss": 3.0801,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.24e-06,
+      "loss": 3.2021,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2600000000000004e-06,
+      "loss": 3.1117,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.28e-06,
+      "loss": 3.2937,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3000000000000004e-06,
+      "loss": 3.1996,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3200000000000002e-06,
+      "loss": 2.9658,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3400000000000005e-06,
+      "loss": 3.0849,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3600000000000003e-06,
+      "loss": 3.0924,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.38e-06,
+      "loss": 3.1194,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 3.1017,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.42e-06,
+      "loss": 3.1338,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 3.2404,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.46e-06,
+      "loss": 3.0279,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4800000000000004e-06,
+      "loss": 3.079,
+      "step": 124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5e-06,
+      "loss": 3.0293,
+      "step": 125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.52e-06,
+      "loss": 3.103,
+      "step": 126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5400000000000002e-06,
+      "loss": 2.9997,
+      "step": 127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.56e-06,
+      "loss": 3.0742,
+      "step": 128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5800000000000003e-06,
+      "loss": 3.1797,
+      "step": 129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6e-06,
+      "loss": 3.1649,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6200000000000003e-06,
+      "loss": 3.1405,
+      "step": 131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.64e-06,
+      "loss": 3.286,
+      "step": 132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6600000000000004e-06,
+      "loss": 3.1416,
+      "step": 133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.68e-06,
+      "loss": 3.1222,
+      "step": 134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7000000000000004e-06,
+      "loss": 3.0737,
+      "step": 135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7200000000000002e-06,
+      "loss": 3.1448,
+      "step": 136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7400000000000004e-06,
+      "loss": 3.0493,
+      "step": 137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7600000000000003e-06,
+      "loss": 3.2033,
+      "step": 138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7800000000000005e-06,
+      "loss": 3.1203,
+      "step": 139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 3.1785,
+      "step": 140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.82e-06,
+      "loss": 3.0821,
+      "step": 141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.84e-06,
+      "loss": 3.0099,
+      "step": 142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.86e-06,
+      "loss": 3.0784,
+      "step": 143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.88e-06,
+      "loss": 3.0435,
+      "step": 144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9e-06,
+      "loss": 3.0802,
+      "step": 145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.92e-06,
+      "loss": 3.1208,
+      "step": 146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9400000000000002e-06,
+      "loss": 3.2314,
+      "step": 147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.96e-06,
+      "loss": 3.117,
+      "step": 148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9800000000000003e-06,
+      "loss": 3.1076,
+      "step": 149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3e-06,
+      "loss": 3.0511,
+      "step": 150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0200000000000003e-06,
+      "loss": 2.9885,
+      "step": 151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.04e-06,
+      "loss": 3.1262,
+      "step": 152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0600000000000003e-06,
+      "loss": 3.1413,
+      "step": 153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.08e-06,
+      "loss": 3.0831,
+      "step": 154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1000000000000004e-06,
+      "loss": 2.9659,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.12e-06,
+      "loss": 2.9846,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1400000000000004e-06,
+      "loss": 3.0918,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1600000000000002e-06,
+      "loss": 3.0398,
+      "step": 158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1800000000000005e-06,
+      "loss": 3.0502,
+      "step": 159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 2.987,
+      "step": 160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2200000000000005e-06,
+      "loss": 3.0629,
+      "step": 161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 2.9819,
+      "step": 162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2600000000000006e-06,
+      "loss": 2.9693,
+      "step": 163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2800000000000004e-06,
+      "loss": 3.1555,
+      "step": 164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3000000000000006e-06,
+      "loss": 3.1472,
+      "step": 165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3200000000000004e-06,
+      "loss": 3.3572,
+      "step": 166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3400000000000006e-06,
+      "loss": 2.979,
+      "step": 167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3600000000000004e-06,
+      "loss": 3.1623,
+      "step": 168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3800000000000007e-06,
+      "loss": 3.1813,
+      "step": 169
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 3.0842,
+      "step": 170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4200000000000007e-06,
+      "loss": 2.9356,
+      "step": 171
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.44e-06,
+      "loss": 3.0433,
+      "step": 172
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.46e-06,
+      "loss": 3.1004,
+      "step": 173
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.48e-06,
+      "loss": 3.114,
+      "step": 174
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5e-06,
+      "loss": 3.2494,
+      "step": 175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.52e-06,
+      "loss": 3.2037,
+      "step": 176
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.54e-06,
+      "loss": 3.1626,
+      "step": 177
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5600000000000002e-06,
+      "loss": 3.1276,
+      "step": 178
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.58e-06,
+      "loss": 3.0766,
+      "step": 179
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 3.2185,
+      "step": 180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.62e-06,
+      "loss": 3.082,
+      "step": 181
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6400000000000003e-06,
+      "loss": 3.0148,
+      "step": 182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.66e-06,
+      "loss": 3.0695,
+      "step": 183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6800000000000003e-06,
+      "loss": 3.0554,
+      "step": 184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7e-06,
+      "loss": 2.8859,
+      "step": 185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7200000000000004e-06,
+      "loss": 3.1443,
+      "step": 186
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.74e-06,
+      "loss": 3.0964,
+      "step": 187
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7600000000000004e-06,
+      "loss": 3.0712,
+      "step": 188
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7800000000000002e-06,
+      "loss": 3.1625,
+      "step": 189
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 3.0586,
+      "step": 190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.820000000000001e-06,
+      "loss": 3.1122,
+      "step": 191
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.8400000000000005e-06,
+      "loss": 3.0199,
+      "step": 192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.86e-06,
+      "loss": 3.1458,
+      "step": 193
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.88e-06,
+      "loss": 3.2163,
+      "step": 194
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.900000000000001e-06,
+      "loss": 2.8927,
+      "step": 195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.920000000000001e-06,
+      "loss": 3.1854,
+      "step": 196
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.94e-06,
+      "loss": 3.1954,
+      "step": 197
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.96e-06,
+      "loss": 3.1884,
+      "step": 198
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.980000000000001e-06,
+      "loss": 3.1828,
+      "step": 199
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 3.1569,
+      "step": 200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.0200000000000005e-06,
+      "loss": 3.0561,
+      "step": 201
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.04e-06,
+      "loss": 3.0922,
+      "step": 202
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.060000000000001e-06,
+      "loss": 3.0153,
+      "step": 203
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.08e-06,
+      "loss": 3.1802,
+      "step": 204
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.1e-06,
+      "loss": 3.0574,
+      "step": 205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.12e-06,
+      "loss": 3.0898,
+      "step": 206
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.14e-06,
+      "loss": 3.0306,
+      "step": 207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.16e-06,
+      "loss": 3.1174,
+      "step": 208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.18e-06,
+      "loss": 3.0552,
+      "step": 209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 3.0428,
+      "step": 210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.22e-06,
+      "loss": 3.0898,
+      "step": 211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.24e-06,
+      "loss": 3.0953,
+      "step": 212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.26e-06,
+      "loss": 3.1077,
+      "step": 213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.2800000000000005e-06,
+      "loss": 3.0957,
+      "step": 214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.3e-06,
+      "loss": 3.0291,
+      "step": 215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.32e-06,
+      "loss": 3.0486,
+      "step": 216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.34e-06,
+      "loss": 3.0932,
+      "step": 217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.360000000000001e-06,
+      "loss": 3.2047,
+      "step": 218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.38e-06,
+      "loss": 3.0911,
+      "step": 219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4e-06,
+      "loss": 3.2072,
+      "step": 220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.42e-06,
+      "loss": 2.9981,
+      "step": 221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.440000000000001e-06,
+      "loss": 3.0462,
+      "step": 222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4600000000000005e-06,
+      "loss": 3.0989,
+      "step": 223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.48e-06,
+      "loss": 3.1608,
+      "step": 224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.5e-06,
+      "loss": 2.9573,
+      "step": 225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.520000000000001e-06,
+      "loss": 3.0081,
+      "step": 226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.540000000000001e-06,
+      "loss": 2.9556,
+      "step": 227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.56e-06,
+      "loss": 2.909,
+      "step": 228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.58e-06,
+      "loss": 3.2459,
+      "step": 229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 3.1779,
+      "step": 230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.620000000000001e-06,
+      "loss": 2.9864,
+      "step": 231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.6400000000000005e-06,
+      "loss": 3.223,
+      "step": 232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.66e-06,
+      "loss": 3.0488,
+      "step": 233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.680000000000001e-06,
+      "loss": 3.2359,
+      "step": 234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.7e-06,
+      "loss": 3.0789,
+      "step": 235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.7200000000000005e-06,
+      "loss": 2.9805,
+      "step": 236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.74e-06,
+      "loss": 3.0342,
+      "step": 237
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.76e-06,
+      "loss": 3.0681,
+      "step": 238
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.78e-06,
+      "loss": 3.0441,
+      "step": 239
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 3.1118,
+      "step": 240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.8200000000000004e-06,
+      "loss": 3.0326,
+      "step": 241
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.84e-06,
+      "loss": 3.1635,
+      "step": 242
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.86e-06,
+      "loss": 3.0902,
+      "step": 243
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.880000000000001e-06,
+      "loss": 2.9848,
+      "step": 244
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 2.8768,
+      "step": 245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.92e-06,
+      "loss": 3.1545,
+      "step": 246
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.94e-06,
+      "loss": 3.1653,
+      "step": 247
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.960000000000001e-06,
+      "loss": 3.1175,
+      "step": 248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.980000000000001e-06,
+      "loss": 3.0581,
+      "step": 249
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5e-06,
+      "loss": 2.9291,
+      "step": 250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.02e-06,
+      "loss": 3.1148,
+      "step": 251
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.04e-06,
+      "loss": 3.1249,
+      "step": 252
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.060000000000001e-06,
+      "loss": 3.0876,
+      "step": 253
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.0800000000000005e-06,
+      "loss": 3.0444,
+      "step": 254
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.1e-06,
+      "loss": 3.0213,
+      "step": 255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.12e-06,
+      "loss": 3.1121,
+      "step": 256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.140000000000001e-06,
+      "loss": 3.0326,
+      "step": 257
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.1600000000000006e-06,
+      "loss": 3.144,
+      "step": 258
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.18e-06,
+      "loss": 3.1211,
+      "step": 259
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.2e-06,
+      "loss": 3.0328,
+      "step": 260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.220000000000001e-06,
+      "loss": 3.0771,
+      "step": 261
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.240000000000001e-06,
+      "loss": 3.109,
+      "step": 262
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.2600000000000005e-06,
+      "loss": 3.2358,
+      "step": 263
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.28e-06,
+      "loss": 3.1524,
+      "step": 264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.300000000000001e-06,
+      "loss": 3.0682,
+      "step": 265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.320000000000001e-06,
+      "loss": 3.1102,
+      "step": 266
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.3400000000000005e-06,
+      "loss": 3.0988,
+      "step": 267
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.36e-06,
+      "loss": 3.1489,
+      "step": 268
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.380000000000001e-06,
+      "loss": 3.1062,
+      "step": 269
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 2.9769,
+      "step": 270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.420000000000001e-06,
+      "loss": 3.0803,
+      "step": 271
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.4400000000000004e-06,
+      "loss": 3.0435,
+      "step": 272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.460000000000001e-06,
+      "loss": 2.9866,
+      "step": 273
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.480000000000001e-06,
+      "loss": 3.0422,
+      "step": 274
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 3.0597,
+      "step": 275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.5200000000000005e-06,
+      "loss": 3.0272,
+      "step": 276
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.540000000000001e-06,
+      "loss": 3.0163,
+      "step": 277
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.560000000000001e-06,
+      "loss": 3.129,
+      "step": 278
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.580000000000001e-06,
+      "loss": 3.1062,
+      "step": 279
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 3.2071,
+      "step": 280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.620000000000001e-06,
+      "loss": 3.1472,
+      "step": 281
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.64e-06,
+      "loss": 3.0662,
+      "step": 282
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.66e-06,
+      "loss": 3.1862,
+      "step": 283
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.68e-06,
+      "loss": 3.0458,
+      "step": 284
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.7e-06,
+      "loss": 3.0001,
+      "step": 285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.72e-06,
+      "loss": 3.1104,
+      "step": 286
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.74e-06,
+      "loss": 3.0196,
+      "step": 287
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.76e-06,
+      "loss": 3.1666,
+      "step": 288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.78e-06,
+      "loss": 3.0872,
+      "step": 289
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5.8e-06,
+      "loss": 3.2008,
+      "step": 290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.82e-06,
+      "loss": 2.9974,
+      "step": 291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.84e-06,
+      "loss": 3.1357,
+      "step": 292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.86e-06,
+      "loss": 3.0922,
+      "step": 293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.8800000000000005e-06,
+      "loss": 3.0677,
+      "step": 294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.9e-06,
+      "loss": 3.0126,
+      "step": 295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.92e-06,
+      "loss": 3.163,
+      "step": 296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.94e-06,
+      "loss": 3.0437,
+      "step": 297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.9600000000000005e-06,
+      "loss": 3.1113,
+      "step": 298
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.98e-06,
+      "loss": 2.9995,
+      "step": 299
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6e-06,
+      "loss": 3.0607,
+      "step": 300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.02e-06,
+      "loss": 2.9391,
+      "step": 301
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.040000000000001e-06,
+      "loss": 3.0333,
+      "step": 302
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.0600000000000004e-06,
+      "loss": 2.9697,
+      "step": 303
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.08e-06,
+      "loss": 3.0962,
+      "step": 304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.1e-06,
+      "loss": 3.1434,
+      "step": 305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.120000000000001e-06,
+      "loss": 3.1579,
+      "step": 306
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.1400000000000005e-06,
+      "loss": 2.9887,
+      "step": 307
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.16e-06,
+      "loss": 2.9773,
+      "step": 308
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.18e-06,
+      "loss": 3.1531,
+      "step": 309
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 2.891,
+      "step": 310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.220000000000001e-06,
+      "loss": 3.2376,
+      "step": 311
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.24e-06,
+      "loss": 3.0786,
+      "step": 312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.26e-06,
+      "loss": 3.1812,
+      "step": 313
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.280000000000001e-06,
+      "loss": 3.0049,
+      "step": 314
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.300000000000001e-06,
+      "loss": 3.0394,
+      "step": 315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.3200000000000005e-06,
+      "loss": 3.1509,
+      "step": 316
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.34e-06,
+      "loss": 2.9543,
+      "step": 317
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.360000000000001e-06,
+      "loss": 3.0603,
+      "step": 318
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.380000000000001e-06,
+      "loss": 2.8277,
+      "step": 319
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 3.0614,
+      "step": 320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.42e-06,
+      "loss": 3.0641,
+      "step": 321
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.440000000000001e-06,
+      "loss": 3.0315,
+      "step": 322
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.460000000000001e-06,
+      "loss": 3.1763,
+      "step": 323
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.480000000000001e-06,
+      "loss": 2.9266,
+      "step": 324
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 2.9134,
+      "step": 325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.520000000000001e-06,
+      "loss": 3.0819,
+      "step": 326
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.540000000000001e-06,
+      "loss": 3.0621,
+      "step": 327
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.560000000000001e-06,
+      "loss": 3.1963,
+      "step": 328
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.5800000000000005e-06,
+      "loss": 2.9981,
+      "step": 329
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 3.0877,
+      "step": 330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.620000000000001e-06,
+      "loss": 3.0905,
+      "step": 331
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.640000000000001e-06,
+      "loss": 3.126,
+      "step": 332
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.660000000000001e-06,
+      "loss": 3.1055,
+      "step": 333
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.680000000000001e-06,
+      "loss": 3.0913,
+      "step": 334
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.700000000000001e-06,
+      "loss": 3.2261,
+      "step": 335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.720000000000001e-06,
+      "loss": 3.1757,
+      "step": 336
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.740000000000001e-06,
+      "loss": 2.9439,
+      "step": 337
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.760000000000001e-06,
+      "loss": 3.0181,
+      "step": 338
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.780000000000001e-06,
+      "loss": 3.209,
+      "step": 339
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 3.0331,
+      "step": 340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.820000000000001e-06,
+      "loss": 3.2446,
+      "step": 341
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.8400000000000014e-06,
+      "loss": 2.9921,
+      "step": 342
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.860000000000001e-06,
+      "loss": 3.0697,
+      "step": 343
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.88e-06,
+      "loss": 3.1452,
+      "step": 344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.9e-06,
+      "loss": 3.0569,
+      "step": 345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.92e-06,
+      "loss": 3.0697,
+      "step": 346
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.9400000000000005e-06,
+      "loss": 3.2088,
+      "step": 347
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.96e-06,
+      "loss": 3.1416,
+      "step": 348
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.98e-06,
+      "loss": 2.8042,
+      "step": 349
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7e-06,
+      "loss": 3.1696,
+      "step": 350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.0200000000000006e-06,
+      "loss": 2.9337,
+      "step": 351
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.04e-06,
+      "loss": 2.9174,
+      "step": 352
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.06e-06,
+      "loss": 3.046,
+      "step": 353
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.08e-06,
+      "loss": 3.0631,
+      "step": 354
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.100000000000001e-06,
+      "loss": 3.0288,
+      "step": 355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.1200000000000004e-06,
+      "loss": 3.1073,
+      "step": 356
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.14e-06,
+      "loss": 3.051,
+      "step": 357
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.16e-06,
+      "loss": 3.0653,
+      "step": 358
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.180000000000001e-06,
+      "loss": 3.0813,
+      "step": 359
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 3.0182,
+      "step": 360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.22e-06,
+      "loss": 3.0579,
+      "step": 361
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.24e-06,
+      "loss": 3.1769,
+      "step": 362
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.260000000000001e-06,
+      "loss": 3.0508,
+      "step": 363
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.280000000000001e-06,
+      "loss": 3.1263,
+      "step": 364
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.3e-06,
+      "loss": 3.1253,
+      "step": 365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.32e-06,
+      "loss": 3.1308,
+      "step": 366
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.340000000000001e-06,
+      "loss": 3.1495,
+      "step": 367
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.360000000000001e-06,
+      "loss": 2.9547,
+      "step": 368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.3800000000000005e-06,
+      "loss": 2.8752,
+      "step": 369
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.4e-06,
+      "loss": 3.108,
+      "step": 370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.420000000000001e-06,
+      "loss": 3.1252,
+      "step": 371
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.440000000000001e-06,
+      "loss": 3.083,
+      "step": 372
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.4600000000000006e-06,
+      "loss": 2.9864,
+      "step": 373
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 7.48e-06,
+      "loss": 3.102,
+      "step": 374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 3.0058,
+      "step": 375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.520000000000001e-06,
+      "loss": 3.0131,
+      "step": 376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.540000000000001e-06,
+      "loss": 3.1352,
+      "step": 377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.5600000000000005e-06,
+      "loss": 3.18,
+      "step": 378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.58e-06,
+      "loss": 3.1698,
+      "step": 379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 3.1226,
+      "step": 380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.620000000000001e-06,
+      "loss": 3.2686,
+      "step": 381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.640000000000001e-06,
+      "loss": 3.0526,
+      "step": 382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.660000000000001e-06,
+      "loss": 3.3077,
+      "step": 383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.680000000000001e-06,
+      "loss": 3.043,
+      "step": 384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.7e-06,
+      "loss": 3.2137,
+      "step": 385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.72e-06,
+      "loss": 2.8803,
+      "step": 386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.74e-06,
+      "loss": 3.1119,
+      "step": 387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.76e-06,
+      "loss": 3.0924,
+      "step": 388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.78e-06,
+      "loss": 3.0619,
+      "step": 389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 3.1927,
+      "step": 390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.820000000000001e-06,
+      "loss": 3.1101,
+      "step": 391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.840000000000001e-06,
+      "loss": 3.0402,
+      "step": 392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.860000000000001e-06,
+      "loss": 3.1376,
+      "step": 393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.88e-06,
+      "loss": 3.0506,
+      "step": 394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.9e-06,
+      "loss": 3.1855,
+      "step": 395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.92e-06,
+      "loss": 3.0404,
+      "step": 396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.94e-06,
+      "loss": 3.1486,
+      "step": 397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.960000000000002e-06,
+      "loss": 3.126,
+      "step": 398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.980000000000002e-06,
+      "loss": 3.2166,
+      "step": 399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 3.0081,
+      "step": 400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.020000000000001e-06,
+      "loss": 3.0502,
+      "step": 401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.040000000000001e-06,
+      "loss": 3.0949,
+      "step": 402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.06e-06,
+      "loss": 3.0327,
+      "step": 403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.08e-06,
+      "loss": 3.1336,
+      "step": 404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.1e-06,
+      "loss": 3.1113,
+      "step": 405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.120000000000002e-06,
+      "loss": 3.0291,
+      "step": 406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.14e-06,
+      "loss": 2.9647,
+      "step": 407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.16e-06,
+      "loss": 2.9896,
+      "step": 408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.18e-06,
+      "loss": 2.9812,
+      "step": 409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.2e-06,
+      "loss": 2.9819,
+      "step": 410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.220000000000001e-06,
+      "loss": 3.0556,
+      "step": 411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.24e-06,
+      "loss": 2.9893,
+      "step": 412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.26e-06,
+      "loss": 3.1341,
+      "step": 413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.28e-06,
+      "loss": 2.9957,
+      "step": 414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.3e-06,
+      "loss": 3.0175,
+      "step": 415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.32e-06,
+      "loss": 3.1211,
+      "step": 416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.34e-06,
+      "loss": 3.186,
+      "step": 417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.36e-06,
+      "loss": 3.086,
+      "step": 418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.380000000000001e-06,
+      "loss": 2.8786,
+      "step": 419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 3.1794,
+      "step": 420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.42e-06,
+      "loss": 3.1308,
+      "step": 421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.44e-06,
+      "loss": 3.0581,
+      "step": 422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.46e-06,
+      "loss": 3.0657,
+      "step": 423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.48e-06,
+      "loss": 3.0328,
+      "step": 424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.5e-06,
+      "loss": 3.095,
+      "step": 425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.52e-06,
+      "loss": 3.0478,
+      "step": 426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.540000000000001e-06,
+      "loss": 2.977,
+      "step": 427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.560000000000001e-06,
+      "loss": 2.9719,
+      "step": 428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.580000000000001e-06,
+      "loss": 2.9309,
+      "step": 429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.6e-06,
+      "loss": 3.0869,
+      "step": 430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.62e-06,
+      "loss": 3.0892,
+      "step": 431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.64e-06,
+      "loss": 3.0276,
+      "step": 432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.66e-06,
+      "loss": 3.1213,
+      "step": 433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.68e-06,
+      "loss": 3.1562,
+      "step": 434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.700000000000001e-06,
+      "loss": 3.0373,
+      "step": 435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.720000000000001e-06,
+      "loss": 2.9365,
+      "step": 436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.740000000000001e-06,
+      "loss": 3.015,
+      "step": 437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.76e-06,
+      "loss": 3.1503,
+      "step": 438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.78e-06,
+      "loss": 3.1242,
+      "step": 439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.8e-06,
+      "loss": 3.1622,
+      "step": 440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.82e-06,
+      "loss": 3.1153,
+      "step": 441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.84e-06,
+      "loss": 3.0024,
+      "step": 442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.860000000000002e-06,
+      "loss": 2.9679,
+      "step": 443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.880000000000001e-06,
+      "loss": 3.0221,
+      "step": 444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.900000000000001e-06,
+      "loss": 2.9975,
+      "step": 445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.920000000000001e-06,
+      "loss": 3.1229,
+      "step": 446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.94e-06,
+      "loss": 3.0897,
+      "step": 447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.96e-06,
+      "loss": 3.2282,
+      "step": 448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.98e-06,
+      "loss": 2.9895,
+      "step": 449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9e-06,
+      "loss": 3.0429,
+      "step": 450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.020000000000002e-06,
+      "loss": 3.0743,
+      "step": 451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.040000000000002e-06,
+      "loss": 3.0832,
+      "step": 452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.060000000000001e-06,
+      "loss": 2.9677,
+      "step": 453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.080000000000001e-06,
+      "loss": 3.1529,
+      "step": 454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.100000000000001e-06,
+      "loss": 2.9651,
+      "step": 455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.12e-06,
+      "loss": 3.1278,
+      "step": 456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.14e-06,
+      "loss": 3.1068,
+      "step": 457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.16e-06,
+      "loss": 2.9124,
+      "step": 458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.180000000000002e-06,
+      "loss": 3.0377,
+      "step": 459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 3.1322,
+      "step": 460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.220000000000002e-06,
+      "loss": 3.061,
+      "step": 461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.240000000000001e-06,
+      "loss": 3.0479,
+      "step": 462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.260000000000001e-06,
+      "loss": 3.285,
+      "step": 463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.280000000000001e-06,
+      "loss": 3.0072,
+      "step": 464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.3e-06,
+      "loss": 2.8649,
+      "step": 465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.32e-06,
+      "loss": 3.0354,
+      "step": 466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.340000000000002e-06,
+      "loss": 3.0652,
+      "step": 467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.360000000000002e-06,
+      "loss": 3.0414,
+      "step": 468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.38e-06,
+      "loss": 2.9699,
+      "step": 469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.4e-06,
+      "loss": 3.0269,
+      "step": 470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.42e-06,
+      "loss": 3.2409,
+      "step": 471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.440000000000001e-06,
+      "loss": 3.0622,
+      "step": 472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.460000000000001e-06,
+      "loss": 3.1338,
+      "step": 473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.48e-06,
+      "loss": 3.0206,
+      "step": 474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.5e-06,
+      "loss": 3.0132,
+      "step": 475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.52e-06,
+      "loss": 3.0594,
+      "step": 476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.54e-06,
+      "loss": 2.9445,
+      "step": 477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.56e-06,
+      "loss": 3.1531,
+      "step": 478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.58e-06,
+      "loss": 3.0505,
+      "step": 479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 3.0427,
+      "step": 480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.620000000000001e-06,
+      "loss": 2.9886,
+      "step": 481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.640000000000001e-06,
+      "loss": 3.1803,
+      "step": 482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.66e-06,
+      "loss": 2.9807,
+      "step": 483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.68e-06,
+      "loss": 2.9869,
+      "step": 484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.7e-06,
+      "loss": 2.9625,
+      "step": 485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.72e-06,
+      "loss": 3.0103,
+      "step": 486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.74e-06,
+      "loss": 3.0629,
+      "step": 487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.760000000000001e-06,
+      "loss": 3.1274,
+      "step": 488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.780000000000001e-06,
+      "loss": 2.9809,
+      "step": 489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 3.2116,
+      "step": 490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.820000000000001e-06,
+      "loss": 3.0269,
+      "step": 491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.84e-06,
+      "loss": 3.165,
+      "step": 492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.86e-06,
+      "loss": 3.0022,
+      "step": 493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.88e-06,
+      "loss": 3.1392,
+      "step": 494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9e-06,
+      "loss": 2.8944,
+      "step": 495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920000000000002e-06,
+      "loss": 3.1758,
+      "step": 496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.940000000000001e-06,
+      "loss": 3.0123,
+      "step": 497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.960000000000001e-06,
+      "loss": 3.1335,
+      "step": 498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980000000000001e-06,
+      "loss": 3.0011,
+      "step": 499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1e-05,
+      "loss": 3.085,
+      "step": 500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999999595896201e-06,
+      "loss": 3.1384,
+      "step": 501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999998383584866e-06,
+      "loss": 3.1792,
+      "step": 502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999996363066192e-06,
+      "loss": 3.1981,
+      "step": 503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999993534340506e-06,
+      "loss": 2.9495,
+      "step": 504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999989897408264e-06,
+      "loss": 3.0814,
+      "step": 505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999985452270057e-06,
+      "loss": 2.8607,
+      "step": 506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9999801989266e-06,
+      "loss": 3.0874,
+      "step": 507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999974137378744e-06,
+      "loss": 2.9915,
+      "step": 508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999967267627468e-06,
+      "loss": 3.0348,
+      "step": 509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999959589673884e-06,
+      "loss": 3.0418,
+      "step": 510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99995110351923e-06,
+      "loss": 3.0336,
+      "step": 511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99994180916488e-06,
+      "loss": 3.0015,
+      "step": 512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999931706612336e-06,
+      "loss": 3.0864,
+      "step": 513
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999920795863232e-06,
+      "loss": 3.0927,
+      "step": 514
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99990907691933e-06,
+      "loss": 3.0762,
+      "step": 515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999896549782524e-06,
+      "loss": 2.9998,
+      "step": 516
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99988321445484e-06,
+      "loss": 2.9911,
+      "step": 517
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999869070938432e-06,
+      "loss": 3.0873,
+      "step": 518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999854119235591e-06,
+      "loss": 2.9681,
+      "step": 519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999838359348728e-06,
+      "loss": 2.9906,
+      "step": 520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999821791280393e-06,
+      "loss": 2.9753,
+      "step": 521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999804415033262e-06,
+      "loss": 3.103,
+      "step": 522
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99978623061015e-06,
+      "loss": 3.1089,
+      "step": 523
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999767238013987e-06,
+      "loss": 3.0382,
+      "step": 524
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99974743724785e-06,
+      "loss": 3.0759,
+      "step": 525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999726828314937e-06,
+      "loss": 2.9768,
+      "step": 526
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99970541121858e-06,
+      "loss": 3.1144,
+      "step": 527
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999683185962242e-06,
+      "loss": 3.0532,
+      "step": 528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999660152549513e-06,
+      "loss": 3.0849,
+      "step": 529
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999636310984118e-06,
+      "loss": 2.975,
+      "step": 530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999611661269908e-06,
+      "loss": 3.2181,
+      "step": 531
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999586203410871e-06,
+      "loss": 3.0907,
+      "step": 532
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.99955993741112e-06,
+      "loss": 3.1782,
+      "step": 533
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999532863274903e-06,
+      "loss": 2.9281,
+      "step": 534
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999504981006594e-06,
+      "loss": 3.1632,
+      "step": 535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9994762906107e-06,
+      "loss": 3.0975,
+      "step": 536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999446792091858e-06,
+      "loss": 2.9594,
+      "step": 537
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999416485454838e-06,
+      "loss": 2.9955,
+      "step": 538
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999385370704537e-06,
+      "loss": 3.0661,
+      "step": 539
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999353447845988e-06,
+      "loss": 3.2558,
+      "step": 540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999320716884347e-06,
+      "loss": 3.1156,
+      "step": 541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999287177824906e-06,
+      "loss": 3.1056,
+      "step": 542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999252830673087e-06,
+      "loss": 3.1299,
+      "step": 543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999217675434442e-06,
+      "loss": 3.133,
+      "step": 544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999181712114651e-06,
+      "loss": 3.0824,
+      "step": 545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999144940719532e-06,
+      "loss": 3.1997,
+      "step": 546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999107361255023e-06,
+      "loss": 3.0006,
+      "step": 547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999068973727204e-06,
+      "loss": 3.0727,
+      "step": 548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.999029778142275e-06,
+      "loss": 3.0825,
+      "step": 549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998989774506576e-06,
+      "loss": 2.9561,
+      "step": 550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99894896282657e-06,
+      "loss": 3.051,
+      "step": 551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998907343108857e-06,
+      "loss": 3.0165,
+      "step": 552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998864915360161e-06,
+      "loss": 3.1546,
+      "step": 553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99882167958734e-06,
+      "loss": 3.1363,
+      "step": 554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998777635797388e-06,
+      "loss": 3.067,
+      "step": 555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998732783997418e-06,
+      "loss": 3.0972,
+      "step": 556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998687124194684e-06,
+      "loss": 3.0825,
+      "step": 557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998640656396563e-06,
+      "loss": 3.2009,
+      "step": 558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99859338061057e-06,
+      "loss": 3.1441,
+      "step": 559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998545296844345e-06,
+      "loss": 3.106,
+      "step": 560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99849640510566e-06,
+      "loss": 2.9718,
+      "step": 561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998446705402418e-06,
+      "loss": 3.0502,
+      "step": 562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998396197742655e-06,
+      "loss": 2.9192,
+      "step": 563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99834488213453e-06,
+      "loss": 3.0094,
+      "step": 564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998292758586341e-06,
+      "loss": 3.0426,
+      "step": 565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998239827106514e-06,
+      "loss": 3.0515,
+      "step": 566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998186087703603e-06,
+      "loss": 3.113,
+      "step": 567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998131540386295e-06,
+      "loss": 3.0648,
+      "step": 568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99807618516341e-06,
+      "loss": 3.0434,
+      "step": 569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.998020022043891e-06,
+      "loss": 2.9913,
+      "step": 570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997963051036819e-06,
+      "loss": 3.1746,
+      "step": 571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997905272151403e-06,
+      "loss": 3.1775,
+      "step": 572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99784668539698e-06,
+      "loss": 2.9915,
+      "step": 573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997787290783025e-06,
+      "loss": 3.0446,
+      "step": 574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997727088319134e-06,
+      "loss": 3.0562,
+      "step": 575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997666078015037e-06,
+      "loss": 3.11,
+      "step": 576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997604259880603e-06,
+      "loss": 3.0556,
+      "step": 577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997541633925818e-06,
+      "loss": 3.1821,
+      "step": 578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997478200160806e-06,
+      "loss": 3.028,
+      "step": 579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997413958595824e-06,
+      "loss": 3.0179,
+      "step": 580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99734890924125e-06,
+      "loss": 3.1073,
+      "step": 581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997283052107605e-06,
+      "loss": 2.8365,
+      "step": 582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99721638720553e-06,
+      "loss": 3.0667,
+      "step": 583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997148914545804e-06,
+      "loss": 3.0064,
+      "step": 584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99708063413933e-06,
+      "loss": 2.8926,
+      "step": 585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.997011545997148e-06,
+      "loss": 3.0958,
+      "step": 586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996941650130423e-06,
+      "loss": 3.0322,
+      "step": 587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996870946550454e-06,
+      "loss": 2.9456,
+      "step": 588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996799435268669e-06,
+      "loss": 2.8868,
+      "step": 589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99672711629663e-06,
+      "loss": 3.1072,
+      "step": 590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996653989646023e-06,
+      "loss": 3.121,
+      "step": 591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996580055328672e-06,
+      "loss": 3.0132,
+      "step": 592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996505313356526e-06,
+      "loss": 3.0281,
+      "step": 593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996429763741664e-06,
+      "loss": 3.2166,
+      "step": 594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996353406496301e-06,
+      "loss": 3.1112,
+      "step": 595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996276241632781e-06,
+      "loss": 3.0205,
+      "step": 596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996198269163573e-06,
+      "loss": 3.1034,
+      "step": 597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996119489101282e-06,
+      "loss": 3.1475,
+      "step": 598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.996039901458643e-06,
+      "loss": 3.1589,
+      "step": 599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995959506248521e-06,
+      "loss": 3.0428,
+      "step": 600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99587830348391e-06,
+      "loss": 3.1614,
+      "step": 601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995796293177935e-06,
+      "loss": 3.0222,
+      "step": 602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995713475343855e-06,
+      "loss": 3.1068,
+      "step": 603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995629849995054e-06,
+      "loss": 3.1134,
+      "step": 604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995545417145052e-06,
+      "loss": 3.1286,
+      "step": 605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995460176807494e-06,
+      "loss": 3.0675,
+      "step": 606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99537412899616e-06,
+      "loss": 3.0818,
+      "step": 607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99528727372496e-06,
+      "loss": 3.0119,
+      "step": 608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.99519961100793e-06,
+      "loss": 3.0262,
+      "step": 609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995111140859243e-06,
+      "loss": 3.1977,
+      "step": 610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995021863293199e-06,
+      "loss": 3.1242,
+      "step": 611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994931778324228e-06,
+      "loss": 3.0035,
+      "step": 612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994840885966892e-06,
+      "loss": 3.2184,
+      "step": 613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994749186235884e-06,
+      "loss": 3.1481,
+      "step": 614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994656679146024e-06,
+      "loss": 3.0971,
+      "step": 615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994563364712266e-06,
+      "loss": 3.0005,
+      "step": 616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994469242949694e-06,
+      "loss": 3.1278,
+      "step": 617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994374313873523e-06,
+      "loss": 2.9811,
+      "step": 618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994278577499096e-06,
+      "loss": 3.0465,
+      "step": 619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994182033841888e-06,
+      "loss": 3.0774,
+      "step": 620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994084682917504e-06,
+      "loss": 2.8815,
+      "step": 621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.993986524741683e-06,
+      "loss": 3.12,
+      "step": 622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.993887559330286e-06,
+      "loss": 2.9711,
+      "step": 623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993787786699316e-06,
+      "loss": 3.0134,
+      "step": 624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993687206864895e-06,
+      "loss": 2.9585,
+      "step": 625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993585819843285e-06,
+      "loss": 3.0559,
+      "step": 626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993483625650873e-06,
+      "loss": 3.0887,
+      "step": 627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993380624304177e-06,
+      "loss": 3.2346,
+      "step": 628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993276815819845e-06,
+      "loss": 3.0095,
+      "step": 629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.99317220021466e-06,
+      "loss": 3.0839,
+      "step": 630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993066777505531e-06,
+      "loss": 2.9096,
+      "step": 631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9929605477095e-06,
+      "loss": 3.0795,
+      "step": 632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992853510843734e-06,
+      "loss": 3.3009,
+      "step": 633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992745666925538e-06,
+      "loss": 2.9208,
+      "step": 634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992637015972344e-06,
+      "loss": 3.0151,
+      "step": 635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992527558001713e-06,
+      "loss": 3.0105,
+      "step": 636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.99241729303134e-06,
+      "loss": 3.0843,
+      "step": 637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992306221079046e-06,
+      "loss": 3.1093,
+      "step": 638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992194342162787e-06,
+      "loss": 2.8051,
+      "step": 639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.992081656300645e-06,
+      "loss": 3.0323,
+      "step": 640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991968163510836e-06,
+      "loss": 3.1345,
+      "step": 641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991853863811706e-06,
+      "loss": 3.1187,
+      "step": 642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991738757221729e-06,
+      "loss": 2.9906,
+      "step": 643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991622843759512e-06,
+      "loss": 2.9303,
+      "step": 644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.99150612344379e-06,
+      "loss": 3.0835,
+      "step": 645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991388596293434e-06,
+      "loss": 3.0459,
+      "step": 646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991270262327435e-06,
+      "loss": 3.047,
+      "step": 647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991151121564926e-06,
+      "loss": 3.0801,
+      "step": 648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.991031174025161e-06,
+      "loss": 3.0477,
+      "step": 649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990910419727533e-06,
+      "loss": 3.0471,
+      "step": 650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990788858691558e-06,
+      "loss": 3.0925,
+      "step": 651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990666490936884e-06,
+      "loss": 3.0155,
+      "step": 652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990543316483294e-06,
+      "loss": 2.9966,
+      "step": 653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990419335350697e-06,
+      "loss": 3.1135,
+      "step": 654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990294547559131e-06,
+      "loss": 3.0049,
+      "step": 655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990168953128771e-06,
+      "loss": 3.0227,
+      "step": 656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.990042552079914e-06,
+      "loss": 3.1753,
+      "step": 657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989915344432997e-06,
+      "loss": 3.1413,
+      "step": 658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989787330208576e-06,
+      "loss": 2.9051,
+      "step": 659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989658509427346e-06,
+      "loss": 3.0933,
+      "step": 660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989528882110133e-06,
+      "loss": 3.0184,
+      "step": 661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989398448277885e-06,
+      "loss": 3.1771,
+      "step": 662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98926720795169e-06,
+      "loss": 3.0918,
+      "step": 663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989135161152758e-06,
+      "loss": 3.1616,
+      "step": 664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.989002307902435e-06,
+      "loss": 2.9601,
+      "step": 665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988868648222196e-06,
+      "loss": 3.0288,
+      "step": 666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988734182133645e-06,
+      "loss": 3.0013,
+      "step": 667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988598909658519e-06,
+      "loss": 3.1021,
+      "step": 668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988462830818682e-06,
+      "loss": 3.04,
+      "step": 669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988325945636128e-06,
+      "loss": 3.0296,
+      "step": 670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988188254132989e-06,
+      "loss": 3.0771,
+      "step": 671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.988049756331516e-06,
+      "loss": 3.0781,
+      "step": 672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9879104522541e-06,
+      "loss": 3.0794,
+      "step": 673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.987770341923256e-06,
+      "loss": 3.0596,
+      "step": 674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.987629425361633e-06,
+      "loss": 3.1287,
+      "step": 675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.987487702592008e-06,
+      "loss": 2.8639,
+      "step": 676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98734517363729e-06,
+      "loss": 3.0636,
+      "step": 677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.987201838520518e-06,
+      "loss": 3.0334,
+      "step": 678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.987057697264858e-06,
+      "loss": 2.9556,
+      "step": 679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986912749893612e-06,
+      "loss": 3.0697,
+      "step": 680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986766996430209e-06,
+      "loss": 2.9794,
+      "step": 681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986620436898208e-06,
+      "loss": 3.0082,
+      "step": 682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986473071321302e-06,
+      "loss": 2.9403,
+      "step": 683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986324899723308e-06,
+      "loss": 3.1005,
+      "step": 684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986175922128177e-06,
+      "loss": 3.1877,
+      "step": 685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.986026138559992e-06,
+      "loss": 3.0305,
+      "step": 686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.985875549042962e-06,
+      "loss": 3.1194,
+      "step": 687
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98572415360143e-06,
+      "loss": 3.1055,
+      "step": 688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98557195225987e-06,
+      "loss": 3.0536,
+      "step": 689
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.985418945042877e-06,
+      "loss": 3.0767,
+      "step": 690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98526513197519e-06,
+      "loss": 3.0426,
+      "step": 691
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.985110513081669e-06,
+      "loss": 2.9332,
+      "step": 692
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984955088387307e-06,
+      "loss": 2.996,
+      "step": 693
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984798857917228e-06,
+      "loss": 2.9786,
+      "step": 694
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984641821696684e-06,
+      "loss": 2.9497,
+      "step": 695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984483979751061e-06,
+      "loss": 3.0489,
+      "step": 696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98432533210587e-06,
+      "loss": 3.047,
+      "step": 697
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.984165878786753e-06,
+      "loss": 3.0853,
+      "step": 698
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.98400561981949e-06,
+      "loss": 3.1989,
+      "step": 699
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983844555229983e-06,
+      "loss": 3.162,
+      "step": 700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983682685044266e-06,
+      "loss": 3.0468,
+      "step": 701
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983520009288504e-06,
+      "loss": 3.1305,
+      "step": 702
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983356527988992e-06,
+      "loss": 3.1339,
+      "step": 703
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983192241172157e-06,
+      "loss": 3.0291,
+      "step": 704
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.983027148864554e-06,
+      "loss": 3.019,
+      "step": 705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.982861251092869e-06,
+      "loss": 3.0899,
+      "step": 706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.982694547883914e-06,
+      "loss": 3.1739,
+      "step": 707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.982527039264641e-06,
+      "loss": 3.0316,
+      "step": 708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.982358725262123e-06,
+      "loss": 3.0335,
+      "step": 709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.982189605903568e-06,
+      "loss": 3.1161,
+      "step": 710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.982019681216311e-06,
+      "loss": 3.1473,
+      "step": 711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981848951227821e-06,
+      "loss": 2.9863,
+      "step": 712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981677415965693e-06,
+      "loss": 3.0296,
+      "step": 713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981505075457655e-06,
+      "loss": 3.0688,
+      "step": 714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981331929731568e-06,
+      "loss": 3.0503,
+      "step": 715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.981157978815412e-06,
+      "loss": 2.9345,
+      "step": 716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.98098322273731e-06,
+      "loss": 2.9892,
+      "step": 717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.980807661525509e-06,
+      "loss": 3.1567,
+      "step": 718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.980631295208387e-06,
+      "loss": 3.1333,
+      "step": 719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.980454123814451e-06,
+      "loss": 2.9725,
+      "step": 720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.980276147372341e-06,
+      "loss": 3.0354,
+      "step": 721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.980097365910823e-06,
+      "loss": 3.0154,
+      "step": 722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.979917779458799e-06,
+      "loss": 3.1373,
+      "step": 723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.979737388045293e-06,
+      "loss": 3.0914,
+      "step": 724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.979556191699468e-06,
+      "loss": 3.0833,
+      "step": 725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97937419045061e-06,
+      "loss": 3.0465,
+      "step": 726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97919138432814e-06,
+      "loss": 2.9854,
+      "step": 727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.979007773361606e-06,
+      "loss": 3.1242,
+      "step": 728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.978823357580688e-06,
+      "loss": 3.0025,
+      "step": 729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.978638137015193e-06,
+      "loss": 3.1815,
+      "step": 730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.978452111695062e-06,
+      "loss": 3.0868,
+      "step": 731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.978265281650364e-06,
+      "loss": 3.0731,
+      "step": 732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.978077646911299e-06,
+      "loss": 3.0398,
+      "step": 733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.977889207508198e-06,
+      "loss": 3.0123,
+      "step": 734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.977699963471517e-06,
+      "loss": 3.1148,
+      "step": 735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.977509914831848e-06,
+      "loss": 3.0975,
+      "step": 736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.977319061619913e-06,
+      "loss": 3.2347,
+      "step": 737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.977127403866556e-06,
+      "loss": 3.1119,
+      "step": 738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97693494160276e-06,
+      "loss": 2.9783,
+      "step": 739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.976741674859638e-06,
+      "loss": 2.9714,
+      "step": 740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.976547603668423e-06,
+      "loss": 3.1644,
+      "step": 741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.976352728060492e-06,
+      "loss": 3.0465,
+      "step": 742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.976157048067341e-06,
+      "loss": 2.8952,
+      "step": 743
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.975960563720599e-06,
+      "loss": 2.9125,
+      "step": 744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97576327505203e-06,
+      "loss": 3.0703,
+      "step": 745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.975565182093523e-06,
+      "loss": 2.92,
+      "step": 746
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.975366284877095e-06,
+      "loss": 3.0396,
+      "step": 747
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.975166583434897e-06,
+      "loss": 3.1609,
+      "step": 748
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.974966077799213e-06,
+      "loss": 3.0453,
+      "step": 749
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.974764768002447e-06,
+      "loss": 2.9661,
+      "step": 750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.974562654077145e-06,
+      "loss": 3.1069,
+      "step": 751
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.974359736055971e-06,
+      "loss": 2.876,
+      "step": 752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97415601397173e-06,
+      "loss": 3.0941,
+      "step": 753
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.973951487857351e-06,
+      "loss": 3.1128,
+      "step": 754
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.973746157745892e-06,
+      "loss": 3.041,
+      "step": 755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.973540023670544e-06,
+      "loss": 3.0266,
+      "step": 756
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.973333085664626e-06,
+      "loss": 3.1391,
+      "step": 757
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.973125343761589e-06,
+      "loss": 3.0902,
+      "step": 758
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972916797995011e-06,
+      "loss": 3.2029,
+      "step": 759
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972707448398605e-06,
+      "loss": 3.1335,
+      "step": 760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972497295006207e-06,
+      "loss": 3.1273,
+      "step": 761
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972286337851789e-06,
+      "loss": 3.0657,
+      "step": 762
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.972074576969448e-06,
+      "loss": 3.0551,
+      "step": 763
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.971862012393415e-06,
+      "loss": 3.2121,
+      "step": 764
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.971648644158049e-06,
+      "loss": 3.0386,
+      "step": 765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.971434472297838e-06,
+      "loss": 3.0352,
+      "step": 766
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.971219496847405e-06,
+      "loss": 3.0944,
+      "step": 767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.971003717841495e-06,
+      "loss": 2.8741,
+      "step": 768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.970787135314986e-06,
+      "loss": 3.065,
+      "step": 769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.97056974930289e-06,
+      "loss": 3.0026,
+      "step": 770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.970351559840346e-06,
+      "loss": 3.0606,
+      "step": 771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.970132566962618e-06,
+      "loss": 3.1143,
+      "step": 772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.96991277070511e-06,
+      "loss": 3.1011,
+      "step": 773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.969692171103344e-06,
+      "loss": 3.0153,
+      "step": 774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.969470768192984e-06,
+      "loss": 2.8793,
+      "step": 775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.969248562009814e-06,
+      "loss": 2.9954,
+      "step": 776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.969025552589754e-06,
+      "loss": 3.0773,
+      "step": 777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.96880173996885e-06,
+      "loss": 3.0391,
+      "step": 778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.96857712418328e-06,
+      "loss": 2.9953,
+      "step": 779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.968351705269353e-06,
+      "loss": 3.0781,
+      "step": 780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.968125483263501e-06,
+      "loss": 3.0468,
+      "step": 781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.967898458202296e-06,
+      "loss": 2.9552,
+      "step": 782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.967670630122435e-06,
+      "loss": 2.9644,
+      "step": 783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.967441999060739e-06,
+      "loss": 3.0413,
+      "step": 784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.96721256505417e-06,
+      "loss": 3.1125,
+      "step": 785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.966982328139809e-06,
+      "loss": 3.0776,
+      "step": 786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.966751288354878e-06,
+      "loss": 2.9823,
+      "step": 787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.966519445736717e-06,
+      "loss": 3.0925,
+      "step": 788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.966286800322804e-06,
+      "loss": 3.0798,
+      "step": 789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.966053352150742e-06,
+      "loss": 3.0973,
+      "step": 790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.96581910125827e-06,
+      "loss": 3.1737,
+      "step": 791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.96558404768325e-06,
+      "loss": 2.9111,
+      "step": 792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.965348191463675e-06,
+      "loss": 3.1656,
+      "step": 793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.965111532637672e-06,
+      "loss": 3.0666,
+      "step": 794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.964874071243495e-06,
+      "loss": 3.0567,
+      "step": 795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.964635807319525e-06,
+      "loss": 2.9735,
+      "step": 796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.964396740904276e-06,
+      "loss": 3.0243,
+      "step": 797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.964156872036394e-06,
+      "loss": 3.2434,
+      "step": 798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.963916200754647e-06,
+      "loss": 2.8988,
+      "step": 799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.963674727097942e-06,
+      "loss": 3.13,
+      "step": 800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.963432451105308e-06,
+      "loss": 3.1027,
+      "step": 801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.963189372815908e-06,
+      "loss": 3.0234,
+      "step": 802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.962945492269036e-06,
+      "loss": 3.0524,
+      "step": 803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.962700809504108e-06,
+      "loss": 3.0457,
+      "step": 804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.962455324560677e-06,
+      "loss": 3.0279,
+      "step": 805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.962209037478427e-06,
+      "loss": 3.0119,
+      "step": 806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.961961948297165e-06,
+      "loss": 3.0769,
+      "step": 807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.96171405705683e-06,
+      "loss": 3.1582,
+      "step": 808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.961465363797495e-06,
+      "loss": 3.2181,
+      "step": 809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.961215868559356e-06,
+      "loss": 2.9296,
+      "step": 810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.960965571382742e-06,
+      "loss": 3.1463,
+      "step": 811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.960714472308113e-06,
+      "loss": 2.9696,
+      "step": 812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.960462571376054e-06,
+      "loss": 2.983,
+      "step": 813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.960209868627288e-06,
+      "loss": 2.8856,
+      "step": 814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.959956364102658e-06,
+      "loss": 3.0165,
+      "step": 815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.959702057843142e-06,
+      "loss": 3.0805,
+      "step": 816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.959446949889848e-06,
+      "loss": 3.0237,
+      "step": 817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.95919104028401e-06,
+      "loss": 3.0286,
+      "step": 818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.958934329066991e-06,
+      "loss": 3.0549,
+      "step": 819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.958676816280292e-06,
+      "loss": 3.0708,
+      "step": 820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.958418501965535e-06,
+      "loss": 2.8963,
+      "step": 821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.958159386164474e-06,
+      "loss": 3.1857,
+      "step": 822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.957899468918995e-06,
+      "loss": 3.0547,
+      "step": 823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.957638750271108e-06,
+      "loss": 2.9677,
+      "step": 824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.957377230262957e-06,
+      "loss": 3.1118,
+      "step": 825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.957114908936815e-06,
+      "loss": 3.0532,
+      "step": 826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.956851786335087e-06,
+      "loss": 3.0804,
+      "step": 827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.956587862500299e-06,
+      "loss": 3.0185,
+      "step": 828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.956323137475115e-06,
+      "loss": 3.0697,
+      "step": 829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.956057611302326e-06,
+      "loss": 2.9622,
+      "step": 830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.955791284024851e-06,
+      "loss": 2.934,
+      "step": 831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.95552415568574e-06,
+      "loss": 3.1197,
+      "step": 832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.955256226328173e-06,
+      "loss": 3.019,
+      "step": 833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.954987495995458e-06,
+      "loss": 3.1705,
+      "step": 834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.954717964731031e-06,
+      "loss": 3.0177,
+      "step": 835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.95444763257846e-06,
+      "loss": 3.1258,
+      "step": 836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.954176499581443e-06,
+      "loss": 2.9613,
+      "step": 837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.953904565783808e-06,
+      "loss": 3.0448,
+      "step": 838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.953631831229508e-06,
+      "loss": 2.9931,
+      "step": 839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.953358295962628e-06,
+      "loss": 3.0898,
+      "step": 840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.953083960027386e-06,
+      "loss": 3.0243,
+      "step": 841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.952808823468122e-06,
+      "loss": 3.004,
+      "step": 842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.952532886329312e-06,
+      "loss": 2.9974,
+      "step": 843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.952256148655559e-06,
+      "loss": 3.0905,
+      "step": 844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.951978610491592e-06,
+      "loss": 2.9404,
+      "step": 845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.951700271882279e-06,
+      "loss": 3.0859,
+      "step": 846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.951421132872603e-06,
+      "loss": 3.1591,
+      "step": 847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.951141193507691e-06,
+      "loss": 3.0024,
+      "step": 848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.950860453832793e-06,
+      "loss": 2.8694,
+      "step": 849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.950578913893284e-06,
+      "loss": 2.9637,
+      "step": 850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.950296573734673e-06,
+      "loss": 3.0198,
+      "step": 851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.9500134334026e-06,
+      "loss": 3.1413,
+      "step": 852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.949729492942833e-06,
+      "loss": 3.0774,
+      "step": 853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.949444752401265e-06,
+      "loss": 2.935,
+      "step": 854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.949159211823925e-06,
+      "loss": 3.1149,
+      "step": 855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.948872871256965e-06,
+      "loss": 2.8708,
+      "step": 856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.948585730746674e-06,
+      "loss": 3.0682,
+      "step": 857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.948297790339463e-06,
+      "loss": 3.0234,
+      "step": 858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.948009050081877e-06,
+      "loss": 3.022,
+      "step": 859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.947719510020584e-06,
+      "loss": 2.9049,
+      "step": 860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.94742917020239e-06,
+      "loss": 2.9559,
+      "step": 861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.947138030674226e-06,
+      "loss": 3.0604,
+      "step": 862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.94684609148315e-06,
+      "loss": 3.1379,
+      "step": 863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.946553352676352e-06,
+      "loss": 3.2759,
+      "step": 864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.94625981430115e-06,
+      "loss": 3.091,
+      "step": 865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.945965476404998e-06,
+      "loss": 3.0076,
+      "step": 866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.945670339035464e-06,
+      "loss": 3.3025,
+      "step": 867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.94537440224026e-06,
+      "loss": 3.0809,
+      "step": 868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.94507766606722e-06,
+      "loss": 3.025,
+      "step": 869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.944780130564312e-06,
+      "loss": 3.1889,
+      "step": 870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.944481795779623e-06,
+      "loss": 3.0629,
+      "step": 871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.944182661761383e-06,
+      "loss": 3.0276,
+      "step": 872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.943882728557943e-06,
+      "loss": 3.0906,
+      "step": 873
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.943581996217784e-06,
+      "loss": 2.9266,
+      "step": 874
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.943280464789515e-06,
+      "loss": 2.9891,
+      "step": 875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.94297813432188e-06,
+      "loss": 2.9637,
+      "step": 876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.942675004863745e-06,
+      "loss": 2.9898,
+      "step": 877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.94237107646411e-06,
+      "loss": 3.0557,
+      "step": 878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.9420663491721e-06,
+      "loss": 2.9745,
+      "step": 879
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.941760823036974e-06,
+      "loss": 2.863,
+      "step": 880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.941454498108116e-06,
+      "loss": 3.057,
+      "step": 881
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.941147374435043e-06,
+      "loss": 3.0397,
+      "step": 882
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.940839452067396e-06,
+      "loss": 3.1291,
+      "step": 883
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.940530731054953e-06,
+      "loss": 2.8529,
+      "step": 884
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.940221211447612e-06,
+      "loss": 3.0147,
+      "step": 885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.939910893295404e-06,
+      "loss": 3.0863,
+      "step": 886
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93959977664849e-06,
+      "loss": 2.948,
+      "step": 887
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93928786155716e-06,
+      "loss": 3.1602,
+      "step": 888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.938975148071833e-06,
+      "loss": 3.0157,
+      "step": 889
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.938661636243054e-06,
+      "loss": 3.0406,
+      "step": 890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.938347326121503e-06,
+      "loss": 3.127,
+      "step": 891
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.938032217757984e-06,
+      "loss": 3.1582,
+      "step": 892
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93771631120343e-06,
+      "loss": 3.0701,
+      "step": 893
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.937399606508905e-06,
+      "loss": 2.9834,
+      "step": 894
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.937082103725606e-06,
+      "loss": 3.053,
+      "step": 895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93676380290485e-06,
+      "loss": 3.0938,
+      "step": 896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.936444704098088e-06,
+      "loss": 3.0933,
+      "step": 897
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.936124807356901e-06,
+      "loss": 3.0946,
+      "step": 898
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.935804112732996e-06,
+      "loss": 2.9907,
+      "step": 899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.935482620278215e-06,
+      "loss": 3.0555,
+      "step": 900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.935160330044518e-06,
+      "loss": 3.0426,
+      "step": 901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.934837242084003e-06,
+      "loss": 3.0007,
+      "step": 902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.934513356448898e-06,
+      "loss": 3.1634,
+      "step": 903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.934188673191552e-06,
+      "loss": 3.0567,
+      "step": 904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.933863192364447e-06,
+      "loss": 3.029,
+      "step": 905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.933536914020198e-06,
+      "loss": 2.9192,
+      "step": 906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.933209838211544e-06,
+      "loss": 3.0183,
+      "step": 907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.932881964991353e-06,
+      "loss": 3.1298,
+      "step": 908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.932553294412622e-06,
+      "loss": 3.0478,
+      "step": 909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93222382652848e-06,
+      "loss": 3.0627,
+      "step": 910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.931893561392178e-06,
+      "loss": 2.9587,
+      "step": 911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.931562499057107e-06,
+      "loss": 3.0635,
+      "step": 912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.931230639576778e-06,
+      "loss": 3.1439,
+      "step": 913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.930897983004832e-06,
+      "loss": 3.0141,
+      "step": 914
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.93056452939504e-06,
+      "loss": 2.9071,
+      "step": 915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.930230278801305e-06,
+      "loss": 3.1398,
+      "step": 916
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.929895231277653e-06,
+      "loss": 3.1261,
+      "step": 917
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.929559386878243e-06,
+      "loss": 2.9629,
+      "step": 918
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.929222745657359e-06,
+      "loss": 3.1196,
+      "step": 919
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.928885307669419e-06,
+      "loss": 3.1108,
+      "step": 920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.928547072968963e-06,
+      "loss": 2.9113,
+      "step": 921
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.92820804161067e-06,
+      "loss": 2.884,
+      "step": 922
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.927868213649337e-06,
+      "loss": 3.2359,
+      "step": 923
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.927527589139895e-06,
+      "loss": 2.9205,
+      "step": 924
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.927186168137404e-06,
+      "loss": 2.9693,
+      "step": 925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.92684395069705e-06,
+      "loss": 3.0839,
+      "step": 926
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.926500936874153e-06,
+      "loss": 3.2413,
+      "step": 927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.926157126724154e-06,
+      "loss": 3.0192,
+      "step": 928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.925812520302631e-06,
+      "loss": 3.0169,
+      "step": 929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.925467117665284e-06,
+      "loss": 3.0213,
+      "step": 930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.925120918867945e-06,
+      "loss": 3.0865,
+      "step": 931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.924773923966574e-06,
+      "loss": 3.2969,
+      "step": 932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.924426133017262e-06,
+      "loss": 3.1419,
+      "step": 933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.924077546076225e-06,
+      "loss": 3.2279,
+      "step": 934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.923728163199806e-06,
+      "loss": 2.9645,
+      "step": 935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.923377984444484e-06,
+      "loss": 3.1172,
+      "step": 936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.923027009866862e-06,
+      "loss": 3.0601,
+      "step": 937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.92267523952367e-06,
+      "loss": 3.1713,
+      "step": 938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.922322673471768e-06,
+      "loss": 3.0872,
+      "step": 939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.921969311768148e-06,
+      "loss": 2.8793,
+      "step": 940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.921615154469929e-06,
+      "loss": 3.0286,
+      "step": 941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.921260201634353e-06,
+      "loss": 2.9806,
+      "step": 942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.920904453318799e-06,
+      "loss": 2.8709,
+      "step": 943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.920547909580767e-06,
+      "loss": 3.0444,
+      "step": 944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.920190570477894e-06,
+      "loss": 2.8463,
+      "step": 945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.919832436067937e-06,
+      "loss": 3.0,
+      "step": 946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.919473506408788e-06,
+      "loss": 3.0667,
+      "step": 947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.919113781558463e-06,
+      "loss": 3.1304,
+      "step": 948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.918753261575108e-06,
+      "loss": 2.966,
+      "step": 949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.918391946517001e-06,
+      "loss": 2.9743,
+      "step": 950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.918029836442542e-06,
+      "loss": 3.0041,
+      "step": 951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.917666931410265e-06,
+      "loss": 3.0056,
+      "step": 952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.917303231478831e-06,
+      "loss": 3.1634,
+      "step": 953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.916938736707028e-06,
+      "loss": 3.0905,
+      "step": 954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.916573447153774e-06,
+      "loss": 2.9986,
+      "step": 955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.916207362878114e-06,
+      "loss": 3.2144,
+      "step": 956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.915840483939224e-06,
+      "loss": 3.03,
+      "step": 957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.915472810396406e-06,
+      "loss": 2.917,
+      "step": 958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.915104342309089e-06,
+      "loss": 2.916,
+      "step": 959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.91473507973684e-06,
+      "loss": 2.8929,
+      "step": 960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.914365022739337e-06,
+      "loss": 2.9527,
+      "step": 961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.913994171376404e-06,
+      "loss": 3.0294,
+      "step": 962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.913622525707985e-06,
+      "loss": 3.3004,
+      "step": 963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.91325008579415e-06,
+      "loss": 3.114,
+      "step": 964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.912876851695102e-06,
+      "loss": 3.0661,
+      "step": 965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.912502823471173e-06,
+      "loss": 2.9922,
+      "step": 966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.912128001182822e-06,
+      "loss": 3.0984,
+      "step": 967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.911752384890634e-06,
+      "loss": 3.1024,
+      "step": 968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.911375974655323e-06,
+      "loss": 3.1788,
+      "step": 969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.910998770537734e-06,
+      "loss": 2.9762,
+      "step": 970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.91062077259884e-06,
+      "loss": 2.9075,
+      "step": 971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.91024198089974e-06,
+      "loss": 3.0715,
+      "step": 972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.909862395501663e-06,
+      "loss": 3.1446,
+      "step": 973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.909482016465963e-06,
+      "loss": 2.9516,
+      "step": 974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.909100843854129e-06,
+      "loss": 2.9835,
+      "step": 975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.908718877727773e-06,
+      "loss": 3.0114,
+      "step": 976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.908336118148636e-06,
+      "loss": 2.9667,
+      "step": 977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907952565178586e-06,
+      "loss": 3.0924,
+      "step": 978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907568218879626e-06,
+      "loss": 3.0293,
+      "step": 979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907183079313877e-06,
+      "loss": 3.0209,
+      "step": 980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.906797146543598e-06,
+      "loss": 2.9483,
+      "step": 981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.906410420631169e-06,
+      "loss": 3.0203,
+      "step": 982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.906022901639103e-06,
+      "loss": 2.9994,
+      "step": 983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.905634589630036e-06,
+      "loss": 3.0299,
+      "step": 984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.90524548466674e-06,
+      "loss": 3.0013,
+      "step": 985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.904855586812104e-06,
+      "loss": 3.0347,
+      "step": 986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.904464896129158e-06,
+      "loss": 3.0494,
+      "step": 987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.90407341268105e-06,
+      "loss": 3.0064,
+      "step": 988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.903681136531062e-06,
+      "loss": 2.9237,
+      "step": 989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.9032880677426e-06,
+      "loss": 3.0765,
+      "step": 990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.902894206379203e-06,
+      "loss": 2.8621,
+      "step": 991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.902499552504534e-06,
+      "loss": 3.0413,
+      "step": 992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.902104106182384e-06,
+      "loss": 2.9959,
+      "step": 993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.901707867476678e-06,
+      "loss": 3.159,
+      "step": 994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.90131083645146e-06,
+      "loss": 2.9825,
+      "step": 995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.900913013170906e-06,
+      "loss": 3.0111,
+      "step": 996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.900514397699326e-06,
+      "loss": 3.1151,
+      "step": 997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.900114990101149e-06,
+      "loss": 3.1145,
+      "step": 998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.899714790440936e-06,
+      "loss": 3.0575,
+      "step": 999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.899313798783376e-06,
+      "loss": 2.835,
+      "step": 1000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.898912015193286e-06,
+      "loss": 2.9191,
+      "step": 1001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.898509439735614e-06,
+      "loss": 3.1103,
+      "step": 1002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.898106072475428e-06,
+      "loss": 3.0199,
+      "step": 1003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.897701913477933e-06,
+      "loss": 2.9917,
+      "step": 1004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.897296962808456e-06,
+      "loss": 3.1255,
+      "step": 1005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.896891220532453e-06,
+      "loss": 3.0932,
+      "step": 1006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.896484686715511e-06,
+      "loss": 3.1326,
+      "step": 1007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.89607736142334e-06,
+      "loss": 2.9853,
+      "step": 1008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.895669244721782e-06,
+      "loss": 3.0195,
+      "step": 1009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.895260336676806e-06,
+      "loss": 2.9778,
+      "step": 1010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.894850637354509e-06,
+      "loss": 2.8028,
+      "step": 1011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.894440146821114e-06,
+      "loss": 2.992,
+      "step": 1012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.894028865142975e-06,
+      "loss": 3.0182,
+      "step": 1013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.89361679238657e-06,
+      "loss": 3.1794,
+      "step": 1014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.893203928618509e-06,
+      "loss": 3.0883,
+      "step": 1015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.892790273905526e-06,
+      "loss": 3.1053,
+      "step": 1016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.892375828314487e-06,
+      "loss": 3.0405,
+      "step": 1017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.891960591912382e-06,
+      "loss": 2.979,
+      "step": 1018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.891544564766331e-06,
+      "loss": 3.0969,
+      "step": 1019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.891127746943582e-06,
+      "loss": 2.9724,
+      "step": 1020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.890710138511509e-06,
+      "loss": 2.982,
+      "step": 1021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.890291739537615e-06,
+      "loss": 3.0831,
+      "step": 1022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.88987255008953e-06,
+      "loss": 2.9604,
+      "step": 1023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.889452570235016e-06,
+      "loss": 2.9231,
+      "step": 1024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.889031800041954e-06,
+      "loss": 3.0498,
+      "step": 1025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.888610239578361e-06,
+      "loss": 3.0009,
+      "step": 1026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.888187888912378e-06,
+      "loss": 3.0301,
+      "step": 1027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.887764748112276e-06,
+      "loss": 2.8966,
+      "step": 1028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.887340817246448e-06,
+      "loss": 3.0128,
+      "step": 1029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.886916096383424e-06,
+      "loss": 3.0876,
+      "step": 1030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.886490585591851e-06,
+      "loss": 2.93,
+      "step": 1031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.886064284940515e-06,
+      "loss": 3.0247,
+      "step": 1032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.88563719449832e-06,
+      "loss": 3.0029,
+      "step": 1033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.8852093143343e-06,
+      "loss": 3.1368,
+      "step": 1034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.884780644517625e-06,
+      "loss": 3.0249,
+      "step": 1035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.88435118511758e-06,
+      "loss": 2.9731,
+      "step": 1036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.883920936203584e-06,
+      "loss": 2.9687,
+      "step": 1037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.883489897845184e-06,
+      "loss": 2.946,
+      "step": 1038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.883058070112055e-06,
+      "loss": 2.8536,
+      "step": 1039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.882625453073998e-06,
+      "loss": 3.1881,
+      "step": 1040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.88219204680094e-06,
+      "loss": 3.1593,
+      "step": 1041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.881757851362939e-06,
+      "loss": 3.0901,
+      "step": 1042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.881322866830176e-06,
+      "loss": 3.1223,
+      "step": 1043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.880887093272968e-06,
+      "loss": 3.0249,
+      "step": 1044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.88045053076175e-06,
+      "loss": 3.0726,
+      "step": 1045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.880013179367091e-06,
+      "loss": 3.1414,
+      "step": 1046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.879575039159683e-06,
+      "loss": 3.0595,
+      "step": 1047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.879136110210347e-06,
+      "loss": 2.9892,
+      "step": 1048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.878696392590037e-06,
+      "loss": 2.9703,
+      "step": 1049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.878255886369825e-06,
+      "loss": 3.1445,
+      "step": 1050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.877814591620916e-06,
+      "loss": 3.0167,
+      "step": 1051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.877372508414642e-06,
+      "loss": 3.0074,
+      "step": 1052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.876929636822461e-06,
+      "loss": 3.0615,
+      "step": 1053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.876485976915961e-06,
+      "loss": 3.132,
+      "step": 1054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.876041528766857e-06,
+      "loss": 3.0951,
+      "step": 1055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.875596292446987e-06,
+      "loss": 2.9328,
+      "step": 1056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.875150268028323e-06,
+      "loss": 2.9273,
+      "step": 1057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.874703455582956e-06,
+      "loss": 3.0642,
+      "step": 1058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.874255855183115e-06,
+      "loss": 3.0611,
+      "step": 1059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.873807466901148e-06,
+      "loss": 2.8038,
+      "step": 1060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.873358290809536e-06,
+      "loss": 3.1251,
+      "step": 1061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.87290832698088e-06,
+      "loss": 3.1331,
+      "step": 1062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.872457575487915e-06,
+      "loss": 2.9806,
+      "step": 1063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.872006036403502e-06,
+      "loss": 3.0906,
+      "step": 1064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.871553709800627e-06,
+      "loss": 2.9631,
+      "step": 1065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.871100595752405e-06,
+      "loss": 3.1137,
+      "step": 1066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.87064669433208e-06,
+      "loss": 2.9801,
+      "step": 1067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.87019200561302e-06,
+      "loss": 3.0582,
+      "step": 1068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86973652966872e-06,
+      "loss": 2.9758,
+      "step": 1069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.869280266572804e-06,
+      "loss": 3.0156,
+      "step": 1070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.868823216399028e-06,
+      "loss": 2.9366,
+      "step": 1071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.868365379221263e-06,
+      "loss": 3.1418,
+      "step": 1072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86790675511352e-06,
+      "loss": 3.0193,
+      "step": 1073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86744734414993e-06,
+      "loss": 3.075,
+      "step": 1074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.866987146404752e-06,
+      "loss": 3.0047,
+      "step": 1075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.866526161952373e-06,
+      "loss": 2.9866,
+      "step": 1076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86606439086731e-06,
+      "loss": 3.057,
+      "step": 1077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.8656018332242e-06,
+      "loss": 2.9524,
+      "step": 1078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.865138489097814e-06,
+      "loss": 2.9235,
+      "step": 1079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86467435856305e-06,
+      "loss": 3.1695,
+      "step": 1080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.864209441694925e-06,
+      "loss": 2.8334,
+      "step": 1081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.863743738568594e-06,
+      "loss": 3.0342,
+      "step": 1082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.863277249259331e-06,
+      "loss": 2.9585,
+      "step": 1083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.862809973842542e-06,
+      "loss": 2.8906,
+      "step": 1084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.862341912393758e-06,
+      "loss": 3.0565,
+      "step": 1085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.861873064988636e-06,
+      "loss": 3.1162,
+      "step": 1086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.86140343170296e-06,
+      "loss": 3.077,
+      "step": 1087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.860933012612645e-06,
+      "loss": 2.9702,
+      "step": 1088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.860461807793729e-06,
+      "loss": 2.9916,
+      "step": 1089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.859989817322378e-06,
+      "loss": 3.104,
+      "step": 1090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.859517041274886e-06,
+      "loss": 3.0469,
+      "step": 1091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.859043479727673e-06,
+      "loss": 3.0673,
+      "step": 1092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.858569132757286e-06,
+      "loss": 3.0511,
+      "step": 1093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.8580940004404e-06,
+      "loss": 3.0843,
+      "step": 1094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.857618082853813e-06,
+      "loss": 2.9874,
+      "step": 1095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.857141380074457e-06,
+      "loss": 3.2768,
+      "step": 1096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.856663892179385e-06,
+      "loss": 2.98,
+      "step": 1097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.856185619245779e-06,
+      "loss": 2.9995,
+      "step": 1098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.855706561350948e-06,
+      "loss": 2.9916,
+      "step": 1099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.855226718572328e-06,
+      "loss": 3.0842,
+      "step": 1100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.854746090987481e-06,
+      "loss": 3.0118,
+      "step": 1101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.854264678674097e-06,
+      "loss": 3.0124,
+      "step": 1102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.853782481709991e-06,
+      "loss": 2.9574,
+      "step": 1103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.853299500173108e-06,
+      "loss": 2.9806,
+      "step": 1104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.852815734141515e-06,
+      "loss": 3.0212,
+      "step": 1105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.852331183693411e-06,
+      "loss": 2.924,
+      "step": 1106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.851845848907119e-06,
+      "loss": 3.0333,
+      "step": 1107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.851359729861089e-06,
+      "loss": 2.9998,
+      "step": 1108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.850872826633898e-06,
+      "loss": 2.9518,
+      "step": 1109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.85038513930425e-06,
+      "loss": 3.0684,
+      "step": 1110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.849896667950974e-06,
+      "loss": 3.0346,
+      "step": 1111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.84940741265303e-06,
+      "loss": 3.026,
+      "step": 1112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.848917373489501e-06,
+      "loss": 3.1037,
+      "step": 1113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.848426550539596e-06,
+      "loss": 3.1129,
+      "step": 1114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.847934943882654e-06,
+      "loss": 2.9091,
+      "step": 1115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.847442553598139e-06,
+      "loss": 3.0816,
+      "step": 1116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.846949379765643e-06,
+      "loss": 3.0964,
+      "step": 1117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.846455422464879e-06,
+      "loss": 3.0245,
+      "step": 1118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.845960681775695e-06,
+      "loss": 3.0414,
+      "step": 1119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.84546515777806e-06,
+      "loss": 3.0885,
+      "step": 1120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.844968850552072e-06,
+      "loss": 2.8888,
+      "step": 1121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.844471760177954e-06,
+      "loss": 2.9242,
+      "step": 1122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.843973886736059e-06,
+      "loss": 3.1354,
+      "step": 1123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.843475230306859e-06,
+      "loss": 3.1241,
+      "step": 1124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.84297579097096e-06,
+      "loss": 3.2431,
+      "step": 1125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.842475568809095e-06,
+      "loss": 3.0449,
+      "step": 1126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.841974563902119e-06,
+      "loss": 3.1438,
+      "step": 1127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.841472776331013e-06,
+      "loss": 3.0495,
+      "step": 1128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.84097020617689e-06,
+      "loss": 3.0482,
+      "step": 1129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.840466853520983e-06,
+      "loss": 2.968,
+      "step": 1130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.839962718444657e-06,
+      "loss": 3.0602,
+      "step": 1131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.839457801029399e-06,
+      "loss": 2.8751,
+      "step": 1132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.838952101356827e-06,
+      "loss": 3.0184,
+      "step": 1133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.838445619508682e-06,
+      "loss": 2.9994,
+      "step": 1134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.837938355566832e-06,
+      "loss": 2.9137,
+      "step": 1135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.837430309613272e-06,
+      "loss": 3.1162,
+      "step": 1136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.836921481730124e-06,
+      "loss": 3.0554,
+      "step": 1137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.836411871999636e-06,
+      "loss": 2.9393,
+      "step": 1138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.835901480504182e-06,
+      "loss": 2.9864,
+      "step": 1139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.83539030732626e-06,
+      "loss": 2.8483,
+      "step": 1140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.834878352548502e-06,
+      "loss": 2.9853,
+      "step": 1141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.834365616253657e-06,
+      "loss": 3.0669,
+      "step": 1142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.833852098524604e-06,
+      "loss": 3.0203,
+      "step": 1143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.833337799444348e-06,
+      "loss": 2.9687,
+      "step": 1144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.832822719096027e-06,
+      "loss": 3.0671,
+      "step": 1145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.832306857562894e-06,
+      "loss": 2.9103,
+      "step": 1146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.831790214928336e-06,
+      "loss": 3.0604,
+      "step": 1147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.831272791275864e-06,
+      "loss": 3.0367,
+      "step": 1148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.830754586689112e-06,
+      "loss": 3.1997,
+      "step": 1149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.830235601251849e-06,
+      "loss": 3.158,
+      "step": 1150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.829715835047959e-06,
+      "loss": 3.0552,
+      "step": 1151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.829195288161461e-06,
+      "loss": 3.1498,
+      "step": 1152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.828673960676496e-06,
+      "loss": 2.9696,
+      "step": 1153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.828151852677332e-06,
+      "loss": 3.0216,
+      "step": 1154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.827628964248364e-06,
+      "loss": 3.0506,
+      "step": 1155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.827105295474111e-06,
+      "loss": 3.013,
+      "step": 1156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.826580846439222e-06,
+      "loss": 3.0934,
+      "step": 1157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.826055617228468e-06,
+      "loss": 2.951,
+      "step": 1158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.825529607926748e-06,
+      "loss": 3.1868,
+      "step": 1159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.825002818619087e-06,
+      "loss": 3.1503,
+      "step": 1160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.824475249390637e-06,
+      "loss": 3.076,
+      "step": 1161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.823946900326675e-06,
+      "loss": 3.0021,
+      "step": 1162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.823417771512602e-06,
+      "loss": 2.9631,
+      "step": 1163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.82288786303395e-06,
+      "loss": 3.205,
+      "step": 1164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.822357174976372e-06,
+      "loss": 3.1665,
+      "step": 1165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.82182570742565e-06,
+      "loss": 2.959,
+      "step": 1166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.82129346046769e-06,
+      "loss": 3.1738,
+      "step": 1167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.82076043418853e-06,
+      "loss": 3.0354,
+      "step": 1168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.820226628674324e-06,
+      "loss": 2.9723,
+      "step": 1169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.81969204401136e-06,
+      "loss": 3.0001,
+      "step": 1170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.819156680286047e-06,
+      "loss": 2.9782,
+      "step": 1171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.818620537584921e-06,
+      "loss": 2.9501,
+      "step": 1172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.818083615994648e-06,
+      "loss": 3.1181,
+      "step": 1173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.817545915602016e-06,
+      "loss": 3.0427,
+      "step": 1174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.81700743649394e-06,
+      "loss": 3.0216,
+      "step": 1175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.816468178757461e-06,
+      "loss": 2.985,
+      "step": 1176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.815928142479743e-06,
+      "loss": 3.0425,
+      "step": 1177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.81538732774808e-06,
+      "loss": 3.0078,
+      "step": 1178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.814845734649889e-06,
+      "loss": 3.0619,
+      "step": 1179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.814303363272714e-06,
+      "loss": 3.0007,
+      "step": 1180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.813760213704229e-06,
+      "loss": 2.8803,
+      "step": 1181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.813216286032223e-06,
+      "loss": 2.9945,
+      "step": 1182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.812671580344621e-06,
+      "loss": 3.1688,
+      "step": 1183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.81212609672947e-06,
+      "loss": 3.0487,
+      "step": 1184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.811579835274941e-06,
+      "loss": 3.0227,
+      "step": 1185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.811032796069335e-06,
+      "loss": 2.9277,
+      "step": 1186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.810484979201073e-06,
+      "loss": 2.9923,
+      "step": 1187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.80993638475871e-06,
+      "loss": 2.8778,
+      "step": 1188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.809387012830917e-06,
+      "loss": 3.0722,
+      "step": 1189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.808836863506496e-06,
+      "loss": 2.9377,
+      "step": 1190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.808285936874376e-06,
+      "loss": 3.0071,
+      "step": 1191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.807734233023608e-06,
+      "loss": 3.1252,
+      "step": 1192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.80718175204337e-06,
+      "loss": 2.9364,
+      "step": 1193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.806628494022968e-06,
+      "loss": 3.1751,
+      "step": 1194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.806074459051828e-06,
+      "loss": 2.9888,
+      "step": 1195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.805519647219508e-06,
+      "loss": 3.0235,
+      "step": 1196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.80496405861569e-06,
+      "loss": 2.9833,
+      "step": 1197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.804407693330174e-06,
+      "loss": 3.0328,
+      "step": 1198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.803850551452899e-06,
+      "loss": 2.9986,
+      "step": 1199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.803292633073917e-06,
+      "loss": 2.9906,
+      "step": 1200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.802733938283412e-06,
+      "loss": 2.8929,
+      "step": 1201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.802174467171694e-06,
+      "loss": 3.1279,
+      "step": 1202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.801614219829196e-06,
+      "loss": 3.0355,
+      "step": 1203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.801053196346476e-06,
+      "loss": 3.0448,
+      "step": 1204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.80049139681422e-06,
+      "loss": 2.9782,
+      "step": 1205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.79992882132324e-06,
+      "loss": 2.9378,
+      "step": 1206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.799365469964466e-06,
+      "loss": 3.0138,
+      "step": 1207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.798801342828963e-06,
+      "loss": 3.0235,
+      "step": 1208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.798236440007918e-06,
+      "loss": 2.9271,
+      "step": 1209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.79767076159264e-06,
+      "loss": 3.0597,
+      "step": 1210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.797104307674568e-06,
+      "loss": 2.8409,
+      "step": 1211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.796537078345264e-06,
+      "loss": 3.1238,
+      "step": 1212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.795969073696416e-06,
+      "loss": 3.0188,
+      "step": 1213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.795400293819837e-06,
+      "loss": 3.1487,
+      "step": 1214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.794830738807463e-06,
+      "loss": 2.9224,
+      "step": 1215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.794260408751364e-06,
+      "loss": 3.076,
+      "step": 1216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.793689303743723e-06,
+      "loss": 3.0895,
+      "step": 1217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.793117423876857e-06,
+      "loss": 3.0329,
+      "step": 1218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.792544769243203e-06,
+      "loss": 3.0709,
+      "step": 1219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.791971339935331e-06,
+      "loss": 3.0924,
+      "step": 1220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.791397136045927e-06,
+      "loss": 2.9935,
+      "step": 1221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.790822157667805e-06,
+      "loss": 2.9682,
+      "step": 1222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.790246404893908e-06,
+      "loss": 3.1507,
+      "step": 1223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.7896698778173e-06,
+      "loss": 3.1014,
+      "step": 1224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.789092576531174e-06,
+      "loss": 3.0197,
+      "step": 1225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.788514501128843e-06,
+      "loss": 3.0822,
+      "step": 1226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.787935651703752e-06,
+      "loss": 2.9275,
+      "step": 1227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.787356028349461e-06,
+      "loss": 2.9052,
+      "step": 1228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.786775631159668e-06,
+      "loss": 3.1039,
+      "step": 1229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.786194460228184e-06,
+      "loss": 2.957,
+      "step": 1230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.785612515648951e-06,
+      "loss": 3.1701,
+      "step": 1231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.78502979751604e-06,
+      "loss": 2.9896,
+      "step": 1232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.784446305923638e-06,
+      "loss": 3.1228,
+      "step": 1233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.783862040966061e-06,
+      "loss": 2.9589,
+      "step": 1234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.783277002737755e-06,
+      "loss": 2.9456,
+      "step": 1235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.78269119133328e-06,
+      "loss": 2.9862,
+      "step": 1236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.782104606847334e-06,
+      "loss": 3.1522,
+      "step": 1237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.781517249374729e-06,
+      "loss": 3.1059,
+      "step": 1238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.780929119010407e-06,
+      "loss": 3.0922,
+      "step": 1239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.780340215849438e-06,
+      "loss": 3.0618,
+      "step": 1240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.779750539987008e-06,
+      "loss": 2.9651,
+      "step": 1241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.779160091518435e-06,
+      "loss": 3.0787,
+      "step": 1242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77856887053916e-06,
+      "loss": 3.0326,
+      "step": 1243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.777976877144751e-06,
+      "loss": 3.1002,
+      "step": 1244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.777384111430896e-06,
+      "loss": 3.0627,
+      "step": 1245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.776790573493412e-06,
+      "loss": 3.0912,
+      "step": 1246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77619626342824e-06,
+      "loss": 2.8674,
+      "step": 1247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77560118133144e-06,
+      "loss": 2.9987,
+      "step": 1248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77500532729921e-06,
+      "loss": 3.0451,
+      "step": 1249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77440870142786e-06,
+      "loss": 3.1149,
+      "step": 1250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77381130381383e-06,
+      "loss": 3.1353,
+      "step": 1251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.773213134553683e-06,
+      "loss": 3.1363,
+      "step": 1252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.77261419374411e-06,
+      "loss": 3.0057,
+      "step": 1253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.772014481481925e-06,
+      "loss": 2.9135,
+      "step": 1254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.771413997864065e-06,
+      "loss": 3.0186,
+      "step": 1255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.770812742987595e-06,
+      "loss": 2.9335,
+      "step": 1256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.770210716949702e-06,
+      "loss": 3.0488,
+      "step": 1257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.769607919847695e-06,
+      "loss": 3.0566,
+      "step": 1258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.769004351779017e-06,
+      "loss": 3.1002,
+      "step": 1259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.768400012841224e-06,
+      "loss": 3.0573,
+      "step": 1260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.767794903132006e-06,
+      "loss": 3.0723,
+      "step": 1261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.767189022749172e-06,
+      "loss": 3.0886,
+      "step": 1262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.766582371790658e-06,
+      "loss": 3.064,
+      "step": 1263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.765974950354526e-06,
+      "loss": 2.9507,
+      "step": 1264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.765366758538957e-06,
+      "loss": 3.0222,
+      "step": 1265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.764757796442262e-06,
+      "loss": 2.9904,
+      "step": 1266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.764148064162874e-06,
+      "loss": 3.0897,
+      "step": 1267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.763537561799353e-06,
+      "loss": 3.0459,
+      "step": 1268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.762926289450376e-06,
+      "loss": 3.0479,
+      "step": 1269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.762314247214756e-06,
+      "loss": 2.9444,
+      "step": 1270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.761701435191425e-06,
+      "loss": 3.0325,
+      "step": 1271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.761087853479432e-06,
+      "loss": 2.9623,
+      "step": 1272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.760473502177962e-06,
+      "loss": 2.9225,
+      "step": 1273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.75985838138632e-06,
+      "loss": 2.8645,
+      "step": 1274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.759242491203935e-06,
+      "loss": 3.0316,
+      "step": 1275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.758625831730359e-06,
+      "loss": 3.0793,
+      "step": 1276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.758008403065269e-06,
+      "loss": 2.9476,
+      "step": 1277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.757390205308469e-06,
+      "loss": 3.1278,
+      "step": 1278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.756771238559886e-06,
+      "loss": 3.094,
+      "step": 1279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.756151502919569e-06,
+      "loss": 3.1436,
+      "step": 1280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.755530998487692e-06,
+      "loss": 2.8675,
+      "step": 1281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.754909725364557e-06,
+      "loss": 3.118,
+      "step": 1282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.754287683650587e-06,
+      "loss": 2.9795,
+      "step": 1283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.753664873446329e-06,
+      "loss": 3.1584,
+      "step": 1284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.753041294852454e-06,
+      "loss": 3.074,
+      "step": 1285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.752416947969761e-06,
+      "loss": 3.0108,
+      "step": 1286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.751791832899168e-06,
+      "loss": 2.9632,
+      "step": 1287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.751165949741721e-06,
+      "loss": 3.0755,
+      "step": 1288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.750539298598586e-06,
+      "loss": 2.9684,
+      "step": 1289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.749911879571059e-06,
+      "loss": 3.0231,
+      "step": 1290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.749283692760557e-06,
+      "loss": 3.073,
+      "step": 1291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.748654738268619e-06,
+      "loss": 2.9648,
+      "step": 1292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.74802501619691e-06,
+      "loss": 2.9147,
+      "step": 1293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.74739452664722e-06,
+      "loss": 3.0198,
+      "step": 1294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.746763269721464e-06,
+      "loss": 2.9417,
+      "step": 1295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.746131245521678e-06,
+      "loss": 2.9799,
+      "step": 1296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.745498454150021e-06,
+      "loss": 3.0258,
+      "step": 1297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.744864895708784e-06,
+      "loss": 2.9755,
+      "step": 1298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.744230570300371e-06,
+      "loss": 3.0337,
+      "step": 1299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.743595478027319e-06,
+      "loss": 3.0456,
+      "step": 1300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.74295961899228e-06,
+      "loss": 3.077,
+      "step": 1301
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.742322993298041e-06,
+      "loss": 2.9482,
+      "step": 1302
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.741685601047506e-06,
+      "loss": 2.9943,
+      "step": 1303
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.741047442343701e-06,
+      "loss": 2.9672,
+      "step": 1304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.740408517289781e-06,
+      "loss": 3.0654,
+      "step": 1305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.739768825989024e-06,
+      "loss": 2.8814,
+      "step": 1306
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.739128368544828e-06,
+      "loss": 2.9594,
+      "step": 1307
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.73848714506072e-06,
+      "loss": 3.0237,
+      "step": 1308
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.737845155640346e-06,
+      "loss": 3.0872,
+      "step": 1309
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.73720240038748e-06,
+      "loss": 2.9277,
+      "step": 1310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.736558879406017e-06,
+      "loss": 3.2322,
+      "step": 1311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.735914592799978e-06,
+      "loss": 2.8855,
+      "step": 1312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.735269540673505e-06,
+      "loss": 3.0261,
+      "step": 1313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.734623723130866e-06,
+      "loss": 3.0761,
+      "step": 1314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.733977140276451e-06,
+      "loss": 3.1238,
+      "step": 1315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.733329792214775e-06,
+      "loss": 3.0821,
+      "step": 1316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.732681679050478e-06,
+      "loss": 3.0519,
+      "step": 1317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.73203280088832e-06,
+      "loss": 3.0393,
+      "step": 1318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.731383157833187e-06,
+      "loss": 2.9448,
+      "step": 1319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.730732749990089e-06,
+      "loss": 3.3216,
+      "step": 1320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.73008157746416e-06,
+      "loss": 3.0776,
+      "step": 1321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.72942964036065e-06,
+      "loss": 3.1205,
+      "step": 1322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.728776938784949e-06,
+      "loss": 3.0245,
+      "step": 1323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.728123472842554e-06,
+      "loss": 3.1123,
+      "step": 1324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.727469242639094e-06,
+      "loss": 2.8739,
+      "step": 1325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.726814248280319e-06,
+      "loss": 3.0802,
+      "step": 1326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.726158489872103e-06,
+      "loss": 2.9917,
+      "step": 1327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.725501967520448e-06,
+      "loss": 3.2078,
+      "step": 1328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.72484468133147e-06,
+      "loss": 3.0833,
+      "step": 1329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.724186631411416e-06,
+      "loss": 3.0117,
+      "step": 1330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.723527817866655e-06,
+      "loss": 3.0903,
+      "step": 1331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.722868240803676e-06,
+      "loss": 2.9953,
+      "step": 1332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.722207900329098e-06,
+      "loss": 2.8974,
+      "step": 1333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.721546796549654e-06,
+      "loss": 2.9382,
+      "step": 1334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.720884929572212e-06,
+      "loss": 3.1016,
+      "step": 1335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.720222299503753e-06,
+      "loss": 3.0821,
+      "step": 1336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.719558906451385e-06,
+      "loss": 3.3529,
+      "step": 1337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.718894750522343e-06,
+      "loss": 2.9668,
+      "step": 1338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.718229831823982e-06,
+      "loss": 3.135,
+      "step": 1339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.717564150463776e-06,
+      "loss": 3.0341,
+      "step": 1340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.716897706549333e-06,
+      "loss": 3.0442,
+      "step": 1341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.716230500188372e-06,
+      "loss": 3.0803,
+      "step": 1342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.715562531488748e-06,
+      "loss": 3.2066,
+      "step": 1343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.714893800558425e-06,
+      "loss": 3.0132,
+      "step": 1344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.714224307505503e-06,
+      "loss": 2.9792,
+      "step": 1345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.713554052438197e-06,
+      "loss": 2.9935,
+      "step": 1346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.712883035464851e-06,
+      "loss": 3.0966,
+      "step": 1347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.712211256693926e-06,
+      "loss": 3.0129,
+      "step": 1348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.711538716234012e-06,
+      "loss": 3.1753,
+      "step": 1349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.710865414193817e-06,
+      "loss": 3.2208,
+      "step": 1350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.710191350682177e-06,
+      "loss": 2.9789,
+      "step": 1351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.709516525808046e-06,
+      "loss": 3.0012,
+      "step": 1352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.708840939680507e-06,
+      "loss": 3.0388,
+      "step": 1353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.70816459240876e-06,
+      "loss": 2.9197,
+      "step": 1354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.70748748410213e-06,
+      "loss": 3.098,
+      "step": 1355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.70680961487007e-06,
+      "loss": 3.0434,
+      "step": 1356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.706130984822149e-06,
+      "loss": 2.9905,
+      "step": 1357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.70545159406806e-06,
+      "loss": 3.1164,
+      "step": 1358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.704771442717626e-06,
+      "loss": 2.9328,
+      "step": 1359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.704090530880783e-06,
+      "loss": 2.8819,
+      "step": 1360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.703408858667597e-06,
+      "loss": 3.1968,
+      "step": 1361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.702726426188253e-06,
+      "loss": 3.0231,
+      "step": 1362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.702043233553061e-06,
+      "loss": 3.101,
+      "step": 1363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.701359280872454e-06,
+      "loss": 3.0684,
+      "step": 1364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.700674568256987e-06,
+      "loss": 3.0556,
+      "step": 1365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.699989095817336e-06,
+      "loss": 3.0016,
+      "step": 1366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.699302863664306e-06,
+      "loss": 2.941,
+      "step": 1367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.698615871908816e-06,
+      "loss": 2.9874,
+      "step": 1368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.697928120661918e-06,
+      "loss": 2.9606,
+      "step": 1369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.697239610034773e-06,
+      "loss": 2.8934,
+      "step": 1370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.69655034013868e-06,
+      "loss": 3.0281,
+      "step": 1371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.695860311085051e-06,
+      "loss": 3.0644,
+      "step": 1372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.695169522985421e-06,
+      "loss": 3.0799,
+      "step": 1373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.694477975951456e-06,
+      "loss": 2.8527,
+      "step": 1374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.693785670094933e-06,
+      "loss": 3.0371,
+      "step": 1375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.69309260552776e-06,
+      "loss": 2.9896,
+      "step": 1376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.692398782361965e-06,
+      "loss": 3.0728,
+      "step": 1377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.691704200709698e-06,
+      "loss": 3.0919,
+      "step": 1378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.691008860683233e-06,
+      "loss": 2.9249,
+      "step": 1379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.690312762394964e-06,
+      "loss": 3.0365,
+      "step": 1380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.689615905957413e-06,
+      "loss": 3.1255,
+      "step": 1381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.688918291483216e-06,
+      "loss": 3.0841,
+      "step": 1382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.688219919085143e-06,
+      "loss": 3.0655,
+      "step": 1383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.687520788876073e-06,
+      "loss": 2.9725,
+      "step": 1384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.686820900969017e-06,
+      "loss": 3.0437,
+      "step": 1385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.686120255477107e-06,
+      "loss": 3.0405,
+      "step": 1386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.685418852513596e-06,
+      "loss": 3.0211,
+      "step": 1387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.68471669219186e-06,
+      "loss": 2.9827,
+      "step": 1388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.684013774625398e-06,
+      "loss": 3.1341,
+      "step": 1389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.683310099927828e-06,
+      "loss": 3.082,
+      "step": 1390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.682605668212896e-06,
+      "loss": 3.1066,
+      "step": 1391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.681900479594466e-06,
+      "loss": 3.0244,
+      "step": 1392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.681194534186524e-06,
+      "loss": 3.0861,
+      "step": 1393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.680487832103183e-06,
+      "loss": 2.9967,
+      "step": 1394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.679780373458675e-06,
+      "loss": 2.9672,
+      "step": 1395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.679072158367353e-06,
+      "loss": 3.1077,
+      "step": 1396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.678363186943696e-06,
+      "loss": 2.9632,
+      "step": 1397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.677653459302301e-06,
+      "loss": 2.9526,
+      "step": 1398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.676942975557893e-06,
+      "loss": 3.1179,
+      "step": 1399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.676231735825313e-06,
+      "loss": 2.8803,
+      "step": 1400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.675519740219524e-06,
+      "loss": 2.9446,
+      "step": 1401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.674806988855621e-06,
+      "loss": 3.0989,
+      "step": 1402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.674093481848807e-06,
+      "loss": 3.0355,
+      "step": 1403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.67337921931442e-06,
+      "loss": 3.0335,
+      "step": 1404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.672664201367914e-06,
+      "loss": 3.044,
+      "step": 1405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.67194842812486e-06,
+      "loss": 2.9152,
+      "step": 1406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.671231899700964e-06,
+      "loss": 2.8639,
+      "step": 1407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.67051461621204e-06,
+      "loss": 2.8059,
+      "step": 1408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.669796577774037e-06,
+      "loss": 3.012,
+      "step": 1409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.669077784503016e-06,
+      "loss": 2.9524,
+      "step": 1410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.668358236515166e-06,
+      "loss": 3.0371,
+      "step": 1411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.667637933926793e-06,
+      "loss": 2.9144,
+      "step": 1412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.666916876854329e-06,
+      "loss": 3.004,
+      "step": 1413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.666195065414327e-06,
+      "loss": 2.9723,
+      "step": 1414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.665472499723465e-06,
+      "loss": 3.0733,
+      "step": 1415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.664749179898534e-06,
+      "loss": 2.9669,
+      "step": 1416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.664025106056454e-06,
+      "loss": 3.0886,
+      "step": 1417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.663300278314269e-06,
+      "loss": 3.0204,
+      "step": 1418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.662574696789138e-06,
+      "loss": 3.018,
+      "step": 1419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.661848361598346e-06,
+      "loss": 2.9582,
+      "step": 1420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.661121272859296e-06,
+      "loss": 3.0454,
+      "step": 1421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.66039343068952e-06,
+      "loss": 2.9719,
+      "step": 1422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.659664835206666e-06,
+      "loss": 3.0532,
+      "step": 1423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.658935486528507e-06,
+      "loss": 2.9766,
+      "step": 1424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.658205384772931e-06,
+      "loss": 2.8264,
+      "step": 1425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.657474530057959e-06,
+      "loss": 2.9966,
+      "step": 1426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.656742922501723e-06,
+      "loss": 2.8842,
+      "step": 1427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.656010562222484e-06,
+      "loss": 3.0329,
+      "step": 1428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.655277449338618e-06,
+      "loss": 2.9803,
+      "step": 1429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.65454358396863e-06,
+      "loss": 3.0749,
+      "step": 1430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.653808966231145e-06,
+      "loss": 2.9765,
+      "step": 1431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.653073596244901e-06,
+      "loss": 2.9879,
+      "step": 1432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.65233747412877e-06,
+      "loss": 2.8767,
+      "step": 1433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.651600600001735e-06,
+      "loss": 3.0137,
+      "step": 1434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.650862973982912e-06,
+      "loss": 3.0533,
+      "step": 1435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.650124596191526e-06,
+      "loss": 3.0481,
+      "step": 1436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.649385466746935e-06,
+      "loss": 3.0737,
+      "step": 1437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.648645585768609e-06,
+      "loss": 3.1124,
+      "step": 1438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.647904953376142e-06,
+      "loss": 3.1963,
+      "step": 1439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.647163569689254e-06,
+      "loss": 3.1796,
+      "step": 1440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.646421434827784e-06,
+      "loss": 2.9696,
+      "step": 1441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.64567854891169e-06,
+      "loss": 2.9615,
+      "step": 1442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.644934912061055e-06,
+      "loss": 2.9712,
+      "step": 1443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.64419052439608e-06,
+      "loss": 2.9875,
+      "step": 1444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.643445386037088e-06,
+      "loss": 2.9879,
+      "step": 1445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.642699497104527e-06,
+      "loss": 2.7813,
+      "step": 1446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.641952857718961e-06,
+      "loss": 3.2147,
+      "step": 1447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.641205468001081e-06,
+      "loss": 3.0958,
+      "step": 1448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.640457328071693e-06,
+      "loss": 2.9965,
+      "step": 1449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.639708438051729e-06,
+      "loss": 3.0527,
+      "step": 1450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.638958798062241e-06,
+      "loss": 2.9576,
+      "step": 1451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.6382084082244e-06,
+      "loss": 3.0429,
+      "step": 1452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.637457268659504e-06,
+      "loss": 3.0429,
+      "step": 1453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.636705379488966e-06,
+      "loss": 2.8648,
+      "step": 1454
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.63595274083432e-06,
+      "loss": 3.0577,
+      "step": 1455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.635199352817228e-06,
+      "loss": 2.9208,
+      "step": 1456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.634445215559467e-06,
+      "loss": 3.0529,
+      "step": 1457
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.633690329182936e-06,
+      "loss": 3.159,
+      "step": 1458
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.63293469380966e-06,
+      "loss": 3.0245,
+      "step": 1459
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.632178309561774e-06,
+      "loss": 3.047,
+      "step": 1460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.631421176561546e-06,
+      "loss": 2.9491,
+      "step": 1461
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.63066329493136e-06,
+      "loss": 3.0314,
+      "step": 1462
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.629904664793719e-06,
+      "loss": 3.1285,
+      "step": 1463
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.629145286271251e-06,
+      "loss": 3.0169,
+      "step": 1464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.628385159486703e-06,
+      "loss": 3.1307,
+      "step": 1465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.627624284562943e-06,
+      "loss": 2.8592,
+      "step": 1466
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.626862661622958e-06,
+      "loss": 2.9803,
+      "step": 1467
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.62610029078986e-06,
+      "loss": 3.031,
+      "step": 1468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.62533717218688e-06,
+      "loss": 3.0588,
+      "step": 1469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.624573305937368e-06,
+      "loss": 2.9949,
+      "step": 1470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.623808692164799e-06,
+      "loss": 3.0648,
+      "step": 1471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.623043330992763e-06,
+      "loss": 2.936,
+      "step": 1472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.622277222544976e-06,
+      "loss": 3.098,
+      "step": 1473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.621510366945272e-06,
+      "loss": 3.0062,
+      "step": 1474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.62074276431761e-06,
+      "loss": 2.9237,
+      "step": 1475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.619974414786063e-06,
+      "loss": 3.1432,
+      "step": 1476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.619205318474827e-06,
+      "loss": 2.8681,
+      "step": 1477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.618435475508223e-06,
+      "loss": 2.9401,
+      "step": 1478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.617664886010689e-06,
+      "loss": 3.0289,
+      "step": 1479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.616893550106784e-06,
+      "loss": 2.9725,
+      "step": 1480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.616121467921188e-06,
+      "loss": 3.087,
+      "step": 1481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.6153486395787e-06,
+      "loss": 2.9633,
+      "step": 1482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.614575065204243e-06,
+      "loss": 3.191,
+      "step": 1483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.61380074492286e-06,
+      "loss": 3.0582,
+      "step": 1484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.61302567885971e-06,
+      "loss": 2.9421,
+      "step": 1485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.612249867140078e-06,
+      "loss": 3.0789,
+      "step": 1486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.611473309889366e-06,
+      "loss": 3.0879,
+      "step": 1487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.610696007233097e-06,
+      "loss": 3.0719,
+      "step": 1488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.60991795929692e-06,
+      "loss": 3.212,
+      "step": 1489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.609139166206596e-06,
+      "loss": 3.1033,
+      "step": 1490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.608359628088009e-06,
+      "loss": 2.9448,
+      "step": 1491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.60757934506717e-06,
+      "loss": 3.0682,
+      "step": 1492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.6067983172702e-06,
+      "loss": 3.0004,
+      "step": 1493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.606016544823348e-06,
+      "loss": 2.9587,
+      "step": 1494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.605234027852981e-06,
+      "loss": 3.0752,
+      "step": 1495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.604450766485585e-06,
+      "loss": 3.0536,
+      "step": 1496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.60366676084777e-06,
+      "loss": 2.988,
+      "step": 1497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.602882011066258e-06,
+      "loss": 3.1118,
+      "step": 1498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.602096517267905e-06,
+      "loss": 3.0088,
+      "step": 1499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.601310279579675e-06,
+      "loss": 3.1295,
+      "step": 1500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.600523298128657e-06,
+      "loss": 3.0396,
+      "step": 1501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.59973557304206e-06,
+      "loss": 3.1903,
+      "step": 1502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.598947104447213e-06,
+      "loss": 2.8672,
+      "step": 1503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.598157892471565e-06,
+      "loss": 3.1817,
+      "step": 1504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.597367937242688e-06,
+      "loss": 3.0343,
+      "step": 1505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.596577238888268e-06,
+      "loss": 3.0585,
+      "step": 1506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.595785797536116e-06,
+      "loss": 2.9586,
+      "step": 1507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.594993613314161e-06,
+      "loss": 2.9883,
+      "step": 1508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.594200686350457e-06,
+      "loss": 3.0826,
+      "step": 1509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.593407016773168e-06,
+      "loss": 3.0441,
+      "step": 1510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.592612604710587e-06,
+      "loss": 3.0537,
+      "step": 1511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.591817450291126e-06,
+      "loss": 3.1494,
+      "step": 1512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.591021553643309e-06,
+      "loss": 3.044,
+      "step": 1513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.590224914895791e-06,
+      "loss": 2.967,
+      "step": 1514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.589427534177343e-06,
+      "loss": 3.022,
+      "step": 1515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.58862941161685e-06,
+      "loss": 3.0092,
+      "step": 1516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.587830547343324e-06,
+      "loss": 3.1654,
+      "step": 1517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.587030941485895e-06,
+      "loss": 3.0078,
+      "step": 1518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.586230594173812e-06,
+      "loss": 2.9677,
+      "step": 1519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.585429505536445e-06,
+      "loss": 3.0922,
+      "step": 1520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.584627675703282e-06,
+      "loss": 2.9109,
+      "step": 1521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.583825104803935e-06,
+      "loss": 2.9044,
+      "step": 1522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.583021792968127e-06,
+      "loss": 3.0037,
+      "step": 1523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.582217740325713e-06,
+      "loss": 3.0543,
+      "step": 1524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.581412947006656e-06,
+      "loss": 3.2457,
+      "step": 1525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.580607413141049e-06,
+      "loss": 3.0186,
+      "step": 1526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.579801138859095e-06,
+      "loss": 3.1224,
+      "step": 1527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.578994124291126e-06,
+      "loss": 2.8948,
+      "step": 1528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.578186369567584e-06,
+      "loss": 3.0387,
+      "step": 1529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.577377874819038e-06,
+      "loss": 3.0711,
+      "step": 1530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.576568640176177e-06,
+      "loss": 3.0187,
+      "step": 1531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.575758665769803e-06,
+      "loss": 3.1909,
+      "step": 1532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.574947951730842e-06,
+      "loss": 3.0356,
+      "step": 1533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.574136498190341e-06,
+      "loss": 2.9652,
+      "step": 1534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.573324305279464e-06,
+      "loss": 3.0151,
+      "step": 1535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.572511373129494e-06,
+      "loss": 3.1383,
+      "step": 1536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.571697701871838e-06,
+      "loss": 2.9068,
+      "step": 1537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.570883291638013e-06,
+      "loss": 2.9502,
+      "step": 1538
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.570068142559666e-06,
+      "loss": 3.124,
+      "step": 1539
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.569252254768558e-06,
+      "loss": 3.0895,
+      "step": 1540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.568435628396571e-06,
+      "loss": 2.9797,
+      "step": 1541
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.567618263575703e-06,
+      "loss": 2.918,
+      "step": 1542
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.566800160438077e-06,
+      "loss": 3.0449,
+      "step": 1543
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.565981319115932e-06,
+      "loss": 2.99,
+      "step": 1544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.565161739741624e-06,
+      "loss": 2.9664,
+      "step": 1545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.564341422447636e-06,
+      "loss": 2.9635,
+      "step": 1546
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.563520367366562e-06,
+      "loss": 3.1096,
+      "step": 1547
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.562698574631118e-06,
+      "loss": 3.0735,
+      "step": 1548
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.56187604437414e-06,
+      "loss": 2.9469,
+      "step": 1549
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.561052776728585e-06,
+      "loss": 3.0186,
+      "step": 1550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.560228771827526e-06,
+      "loss": 3.0905,
+      "step": 1551
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.559404029804158e-06,
+      "loss": 3.1431,
+      "step": 1552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.558578550791792e-06,
+      "loss": 3.1236,
+      "step": 1553
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.55775233492386e-06,
+      "loss": 2.9732,
+      "step": 1554
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.556925382333911e-06,
+      "loss": 2.9145,
+      "step": 1555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.556097693155618e-06,
+      "loss": 3.0159,
+      "step": 1556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.55526926752277e-06,
+      "loss": 3.0497,
+      "step": 1557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.55444010556927e-06,
+      "loss": 3.0901,
+      "step": 1558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.55361020742915e-06,
+      "loss": 2.9573,
+      "step": 1559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.552779573236554e-06,
+      "loss": 3.0468,
+      "step": 1560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.551948203125747e-06,
+      "loss": 3.0894,
+      "step": 1561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.551116097231115e-06,
+      "loss": 3.0583,
+      "step": 1562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.550283255687159e-06,
+      "loss": 3.0717,
+      "step": 1563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.549449678628499e-06,
+      "loss": 3.0683,
+      "step": 1564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.54861536618988e-06,
+      "loss": 3.0457,
+      "step": 1565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.54778031850616e-06,
+      "loss": 2.9543,
+      "step": 1566
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.546944535712314e-06,
+      "loss": 3.0787,
+      "step": 1567
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.546108017943443e-06,
+      "loss": 3.0024,
+      "step": 1568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.54527076533476e-06,
+      "loss": 3.2173,
+      "step": 1569
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.544432778021604e-06,
+      "loss": 2.9722,
+      "step": 1570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.543594056139424e-06,
+      "loss": 3.1676,
+      "step": 1571
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.542754599823795e-06,
+      "loss": 3.0585,
+      "step": 1572
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.541914409210408e-06,
+      "loss": 2.9992,
+      "step": 1573
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.541073484435073e-06,
+      "loss": 3.0321,
+      "step": 1574
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.540231825633717e-06,
+      "loss": 2.977,
+      "step": 1575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.539389432942385e-06,
+      "loss": 3.0718,
+      "step": 1576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.538546306497247e-06,
+      "loss": 3.0377,
+      "step": 1577
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.537702446434587e-06,
+      "loss": 2.9075,
+      "step": 1578
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.536857852890803e-06,
+      "loss": 2.9982,
+      "step": 1579
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.53601252600242e-06,
+      "loss": 2.9836,
+      "step": 1580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.535166465906078e-06,
+      "loss": 3.0833,
+      "step": 1581
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.534319672738537e-06,
+      "loss": 3.0325,
+      "step": 1582
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.533472146636668e-06,
+      "loss": 3.0105,
+      "step": 1583
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.532623887737473e-06,
+      "loss": 2.8896,
+      "step": 1584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.531774896178063e-06,
+      "loss": 2.9983,
+      "step": 1585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.530925172095669e-06,
+      "loss": 3.0598,
+      "step": 1586
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.530074715627643e-06,
+      "loss": 3.0503,
+      "step": 1587
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.529223526911456e-06,
+      "loss": 3.0609,
+      "step": 1588
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.528371606084692e-06,
+      "loss": 3.0233,
+      "step": 1589
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.527518953285059e-06,
+      "loss": 3.0422,
+      "step": 1590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.52666556865038e-06,
+      "loss": 3.1274,
+      "step": 1591
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.525811452318597e-06,
+      "loss": 3.0028,
+      "step": 1592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.524956604427774e-06,
+      "loss": 3.0945,
+      "step": 1593
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.524101025116084e-06,
+      "loss": 3.1471,
+      "step": 1594
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.523244714521832e-06,
+      "loss": 2.9009,
+      "step": 1595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.522387672783425e-06,
+      "loss": 3.0241,
+      "step": 1596
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.521529900039401e-06,
+      "loss": 3.0637,
+      "step": 1597
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.520671396428412e-06,
+      "loss": 3.0487,
+      "step": 1598
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.519812162089225e-06,
+      "loss": 3.0398,
+      "step": 1599
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.518952197160734e-06,
+      "loss": 2.9402,
+      "step": 1600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.518091501781937e-06,
+      "loss": 3.1419,
+      "step": 1601
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.517230076091964e-06,
+      "loss": 3.0213,
+      "step": 1602
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.516367920230056e-06,
+      "loss": 3.0533,
+      "step": 1603
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.51550503433557e-06,
+      "loss": 2.9897,
+      "step": 1604
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.514641418547988e-06,
+      "loss": 2.9533,
+      "step": 1605
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.513777073006904e-06,
+      "loss": 3.0205,
+      "step": 1606
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.512911997852034e-06,
+      "loss": 3.0863,
+      "step": 1607
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.512046193223208e-06,
+      "loss": 2.9939,
+      "step": 1608
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.511179659260377e-06,
+      "loss": 3.0789,
+      "step": 1609
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.51031239610361e-06,
+      "loss": 2.9886,
+      "step": 1610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.50944440389309e-06,
+      "loss": 2.9397,
+      "step": 1611
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.50857568276912e-06,
+      "loss": 3.1136,
+      "step": 1612
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.507706232872127e-06,
+      "loss": 2.8964,
+      "step": 1613
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.506836054342646e-06,
+      "loss": 3.111,
+      "step": 1614
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.505965147321335e-06,
+      "loss": 3.0239,
+      "step": 1615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.505093511948967e-06,
+      "loss": 3.0463,
+      "step": 1616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.504221148366437e-06,
+      "loss": 3.0148,
+      "step": 1617
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.503348056714753e-06,
+      "loss": 3.0817,
+      "step": 1618
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.502474237135044e-06,
+      "loss": 3.1731,
+      "step": 1619
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.501599689768558e-06,
+      "loss": 3.0376,
+      "step": 1620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.500724414756653e-06,
+      "loss": 2.9932,
+      "step": 1621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.499848412240813e-06,
+      "loss": 3.0866,
+      "step": 1622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.498971682362635e-06,
+      "loss": 3.0296,
+      "step": 1623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.498094225263837e-06,
+      "loss": 2.9307,
+      "step": 1624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.49721604108625e-06,
+      "loss": 3.1247,
+      "step": 1625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.496337129971826e-06,
+      "loss": 3.0717,
+      "step": 1626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.495457492062635e-06,
+      "loss": 3.077,
+      "step": 1627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.494577127500862e-06,
+      "loss": 2.9615,
+      "step": 1628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.49369603642881e-06,
+      "loss": 2.8714,
+      "step": 1629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.492814218988898e-06,
+      "loss": 2.8904,
+      "step": 1630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.49193167532367e-06,
+      "loss": 3.0231,
+      "step": 1631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.491048405575776e-06,
+      "loss": 2.9624,
+      "step": 1632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.49016440988799e-06,
+      "loss": 3.0424,
+      "step": 1633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.489279688403205e-06,
+      "loss": 2.986,
+      "step": 1634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.488394241264427e-06,
+      "loss": 2.9658,
+      "step": 1635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.487508068614782e-06,
+      "loss": 3.1047,
+      "step": 1636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.486621170597512e-06,
+      "loss": 3.0896,
+      "step": 1637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.485733547355975e-06,
+      "loss": 3.0132,
+      "step": 1638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.48484519903365e-06,
+      "loss": 2.9828,
+      "step": 1639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.483956125774129e-06,
+      "loss": 2.951,
+      "step": 1640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.483066327721126e-06,
+      "loss": 3.089,
+      "step": 1641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.482175805018465e-06,
+      "loss": 2.8746,
+      "step": 1642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.481284557810095e-06,
+      "loss": 3.0316,
+      "step": 1643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.480392586240076e-06,
+      "loss": 2.8195,
+      "step": 1644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.479499890452593e-06,
+      "loss": 3.0346,
+      "step": 1645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.478606470591936e-06,
+      "loss": 3.0023,
+      "step": 1646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.477712326802521e-06,
+      "loss": 3.1129,
+      "step": 1647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.47681745922888e-06,
+      "loss": 3.0484,
+      "step": 1648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.47592186801566e-06,
+      "loss": 3.064,
+      "step": 1649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.475025553307625e-06,
+      "loss": 2.9874,
+      "step": 1650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.474128515249659e-06,
+      "loss": 2.9471,
+      "step": 1651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.473230753986757e-06,
+      "loss": 3.1144,
+      "step": 1652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.472332269664038e-06,
+      "loss": 3.0308,
+      "step": 1653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.471433062426732e-06,
+      "loss": 3.0507,
+      "step": 1654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.47053313242019e-06,
+      "loss": 2.8715,
+      "step": 1655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.469632479789876e-06,
+      "loss": 3.0731,
+      "step": 1656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.468731104681374e-06,
+      "loss": 3.1113,
+      "step": 1657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.467829007240384e-06,
+      "loss": 3.0598,
+      "step": 1658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.466926187612724e-06,
+      "loss": 2.8503,
+      "step": 1659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.466022645944321e-06,
+      "loss": 3.0208,
+      "step": 1660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.465118382381233e-06,
+      "loss": 3.0146,
+      "step": 1661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.464213397069622e-06,
+      "loss": 3.0353,
+      "step": 1662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.46330769015577e-06,
+      "loss": 3.2085,
+      "step": 1663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.46240126178608e-06,
+      "loss": 3.0817,
+      "step": 1664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.461494112107066e-06,
+      "loss": 2.9064,
+      "step": 1665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.460586241265362e-06,
+      "loss": 2.949,
+      "step": 1666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.459677649407718e-06,
+      "loss": 3.0407,
+      "step": 1667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.458768336681001e-06,
+      "loss": 3.0719,
+      "step": 1668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.457858303232193e-06,
+      "loss": 3.0061,
+      "step": 1669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.45694754920839e-06,
+      "loss": 3.0349,
+      "step": 1670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.456036074756813e-06,
+      "loss": 3.0683,
+      "step": 1671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.45512388002479e-06,
+      "loss": 3.11,
+      "step": 1672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.454210965159773e-06,
+      "loss": 3.0402,
+      "step": 1673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.453297330309324e-06,
+      "loss": 3.008,
+      "step": 1674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.452382975621126e-06,
+      "loss": 3.0086,
+      "step": 1675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.451467901242975e-06,
+      "loss": 3.1082,
+      "step": 1676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.450552107322785e-06,
+      "loss": 3.046,
+      "step": 1677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.449635594008591e-06,
+      "loss": 2.9697,
+      "step": 1678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.448718361448533e-06,
+      "loss": 3.1816,
+      "step": 1679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.447800409790879e-06,
+      "loss": 2.8555,
+      "step": 1680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.446881739184004e-06,
+      "loss": 3.1746,
+      "step": 1681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.445962349776406e-06,
+      "loss": 3.0933,
+      "step": 1682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.445042241716696e-06,
+      "loss": 3.1131,
+      "step": 1683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.444121415153603e-06,
+      "loss": 2.9981,
+      "step": 1684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.443199870235967e-06,
+      "loss": 3.1865,
+      "step": 1685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.44227760711275e-06,
+      "loss": 3.0327,
+      "step": 1686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.44135462593303e-06,
+      "loss": 3.1324,
+      "step": 1687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.440430926845996e-06,
+      "loss": 3.0302,
+      "step": 1688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.439506510000958e-06,
+      "loss": 2.9274,
+      "step": 1689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.438581375547339e-06,
+      "loss": 2.8983,
+      "step": 1690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.43765552363468e-06,
+      "loss": 3.1121,
+      "step": 1691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.436728954412638e-06,
+      "loss": 2.9129,
+      "step": 1692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.435801668030982e-06,
+      "loss": 2.8599,
+      "step": 1693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.434873664639602e-06,
+      "loss": 3.0445,
+      "step": 1694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.433944944388502e-06,
+      "loss": 2.8806,
+      "step": 1695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.433015507427801e-06,
+      "loss": 3.0382,
+      "step": 1696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.432085353907736e-06,
+      "loss": 2.9902,
+      "step": 1697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.431154483978659e-06,
+      "loss": 3.0106,
+      "step": 1698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.430222897791033e-06,
+      "loss": 3.0141,
+      "step": 1699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.429290595495445e-06,
+      "loss": 3.0884,
+      "step": 1700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.428357577242593e-06,
+      "loss": 3.1417,
+      "step": 1701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.42742384318329e-06,
+      "loss": 3.0977,
+      "step": 1702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.426489393468468e-06,
+      "loss": 2.9282,
+      "step": 1703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.425554228249171e-06,
+      "loss": 2.9618,
+      "step": 1704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.424618347676563e-06,
+      "loss": 2.8963,
+      "step": 1705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.423681751901919e-06,
+      "loss": 2.9651,
+      "step": 1706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.422744441076631e-06,
+      "loss": 3.119,
+      "step": 1707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.421806415352211e-06,
+      "loss": 2.9372,
+      "step": 1708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.42086767488028e-06,
+      "loss": 2.8761,
+      "step": 1709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.419928219812578e-06,
+      "loss": 3.028,
+      "step": 1710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.41898805030096e-06,
+      "loss": 2.984,
+      "step": 1711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.418047166497398e-06,
+      "loss": 3.0917,
+      "step": 1712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.417105568553975e-06,
+      "loss": 2.9958,
+      "step": 1713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.416163256622894e-06,
+      "loss": 3.1657,
+      "step": 1714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.41522023085647e-06,
+      "loss": 3.0703,
+      "step": 1715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.41427649140714e-06,
+      "loss": 3.1109,
+      "step": 1716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.413332038427445e-06,
+      "loss": 3.0243,
+      "step": 1717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.412386872070054e-06,
+      "loss": 3.0687,
+      "step": 1718
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.41144099248774e-06,
+      "loss": 2.8785,
+      "step": 1719
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.410494399833398e-06,
+      "loss": 3.1474,
+      "step": 1720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.40954709426004e-06,
+      "loss": 3.0817,
+      "step": 1721
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.408599075920783e-06,
+      "loss": 3.0492,
+      "step": 1722
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.407650344968874e-06,
+      "loss": 2.9502,
+      "step": 1723
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.40670090155766e-06,
+      "loss": 2.9708,
+      "step": 1724
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.405750745840617e-06,
+      "loss": 3.023,
+      "step": 1725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.404799877971324e-06,
+      "loss": 3.0877,
+      "step": 1726
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.403848298103484e-06,
+      "loss": 3.0404,
+      "step": 1727
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.402896006390913e-06,
+      "loss": 3.1829,
+      "step": 1728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.401943002987538e-06,
+      "loss": 2.9362,
+      "step": 1729
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.400989288047405e-06,
+      "loss": 3.1844,
+      "step": 1730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.400034861724672e-06,
+      "loss": 3.0432,
+      "step": 1731
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.399079724173617e-06,
+      "loss": 3.0042,
+      "step": 1732
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.398123875548628e-06,
+      "loss": 2.9174,
+      "step": 1733
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.397167316004211e-06,
+      "loss": 3.0374,
+      "step": 1734
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.396210045694987e-06,
+      "loss": 3.093,
+      "step": 1735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.395252064775688e-06,
+      "loss": 2.9495,
+      "step": 1736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.394293373401165e-06,
+      "loss": 3.0475,
+      "step": 1737
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.393333971726379e-06,
+      "loss": 2.9569,
+      "step": 1738
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.392373859906414e-06,
+      "loss": 2.8808,
+      "step": 1739
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.39141303809646e-06,
+      "loss": 2.9552,
+      "step": 1740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.39045150645183e-06,
+      "loss": 2.953,
+      "step": 1741
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.389489265127942e-06,
+      "loss": 3.1043,
+      "step": 1742
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.38852631428034e-06,
+      "loss": 3.0874,
+      "step": 1743
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.387562654064672e-06,
+      "loss": 3.0593,
+      "step": 1744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.386598284636708e-06,
+      "loss": 3.0385,
+      "step": 1745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.38563320615233e-06,
+      "loss": 2.9285,
+      "step": 1746
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.384667418767532e-06,
+      "loss": 2.7526,
+      "step": 1747
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.38370092263843e-06,
+      "loss": 3.0134,
+      "step": 1748
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.382733717921246e-06,
+      "loss": 3.1353,
+      "step": 1749
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.381765804772322e-06,
+      "loss": 3.0977,
+      "step": 1750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.380797183348115e-06,
+      "loss": 3.0009,
+      "step": 1751
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.379827853805191e-06,
+      "loss": 3.1041,
+      "step": 1752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.378857816300235e-06,
+      "loss": 3.0271,
+      "step": 1753
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.377887070990046e-06,
+      "loss": 3.0387,
+      "step": 1754
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.376915618031536e-06,
+      "loss": 3.1645,
+      "step": 1755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.375943457581732e-06,
+      "loss": 2.9461,
+      "step": 1756
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.374970589797778e-06,
+      "loss": 3.1072,
+      "step": 1757
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.373997014836926e-06,
+      "loss": 2.9264,
+      "step": 1758
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.373022732856547e-06,
+      "loss": 3.1542,
+      "step": 1759
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.372047744014128e-06,
+      "loss": 3.1165,
+      "step": 1760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.371072048467265e-06,
+      "loss": 2.9363,
+      "step": 1761
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.370095646373674e-06,
+      "loss": 3.0369,
+      "step": 1762
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.369118537891177e-06,
+      "loss": 2.9688,
+      "step": 1763
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.368140723177722e-06,
+      "loss": 2.9668,
+      "step": 1764
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.367162202391357e-06,
+      "loss": 3.0377,
+      "step": 1765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.366182975690258e-06,
+      "loss": 2.9793,
+      "step": 1766
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.365203043232705e-06,
+      "loss": 3.0017,
+      "step": 1767
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.364222405177097e-06,
+      "loss": 3.0339,
+      "step": 1768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.363241061681946e-06,
+      "loss": 2.9455,
+      "step": 1769
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.362259012905878e-06,
+      "loss": 3.1011,
+      "step": 1770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.361276259007633e-06,
+      "loss": 2.9904,
+      "step": 1771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.360292800146062e-06,
+      "loss": 2.916,
+      "step": 1772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.359308636480135e-06,
+      "loss": 3.0489,
+      "step": 1773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.358323768168936e-06,
+      "loss": 3.1292,
+      "step": 1774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.357338195371658e-06,
+      "loss": 3.0585,
+      "step": 1775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.35635191824761e-06,
+      "loss": 3.0834,
+      "step": 1776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.355364936956215e-06,
+      "loss": 3.0171,
+      "step": 1777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.354377251657012e-06,
+      "loss": 2.9846,
+      "step": 1778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.353388862509652e-06,
+      "loss": 3.1387,
+      "step": 1779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.352399769673898e-06,
+      "loss": 2.9786,
+      "step": 1780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.351409973309632e-06,
+      "loss": 3.1277,
+      "step": 1781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.350419473576842e-06,
+      "loss": 3.0015,
+      "step": 1782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.349428270635635e-06,
+      "loss": 3.1335,
+      "step": 1783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.348436364646232e-06,
+      "loss": 3.0409,
+      "step": 1784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.347443755768963e-06,
+      "loss": 3.2042,
+      "step": 1785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.34645044416428e-06,
+      "loss": 2.9873,
+      "step": 1786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.345456429992739e-06,
+      "loss": 2.9855,
+      "step": 1787
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.344461713415016e-06,
+      "loss": 3.0531,
+      "step": 1788
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.343466294591898e-06,
+      "loss": 3.0204,
+      "step": 1789
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.342470173684286e-06,
+      "loss": 3.0725,
+      "step": 1790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.341473350853195e-06,
+      "loss": 2.9311,
+      "step": 1791
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.34047582625975e-06,
+      "loss": 2.975,
+      "step": 1792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.339477600065197e-06,
+      "loss": 3.1544,
+      "step": 1793
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.338478672430887e-06,
+      "loss": 2.956,
+      "step": 1794
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.337479043518292e-06,
+      "loss": 3.0205,
+      "step": 1795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.33647871348899e-06,
+      "loss": 2.9331,
+      "step": 1796
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.335477682504677e-06,
+      "loss": 2.9501,
+      "step": 1797
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.334475950727161e-06,
+      "loss": 2.9699,
+      "step": 1798
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.333473518318364e-06,
+      "loss": 2.9871,
+      "step": 1799
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.332470385440322e-06,
+      "loss": 2.9874,
+      "step": 1800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.331466552255179e-06,
+      "loss": 2.9284,
+      "step": 1801
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.3304620189252e-06,
+      "loss": 2.9306,
+      "step": 1802
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.329456785612758e-06,
+      "loss": 3.0429,
+      "step": 1803
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.32845085248034e-06,
+      "loss": 3.1321,
+      "step": 1804
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.327444219690546e-06,
+      "loss": 3.0297,
+      "step": 1805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.326436887406093e-06,
+      "loss": 2.9655,
+      "step": 1806
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.325428855789803e-06,
+      "loss": 3.036,
+      "step": 1807
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.324420125004619e-06,
+      "loss": 3.0207,
+      "step": 1808
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.323410695213593e-06,
+      "loss": 2.9274,
+      "step": 1809
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.322400566579889e-06,
+      "loss": 3.0092,
+      "step": 1810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.321389739266789e-06,
+      "loss": 2.9038,
+      "step": 1811
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.320378213437682e-06,
+      "loss": 3.123,
+      "step": 1812
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.319365989256073e-06,
+      "loss": 2.9504,
+      "step": 1813
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.318353066885579e-06,
+      "loss": 3.0355,
+      "step": 1814
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.317339446489935e-06,
+      "loss": 2.7502,
+      "step": 1815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.316325128232976e-06,
+      "loss": 2.8929,
+      "step": 1816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.315310112278665e-06,
+      "loss": 3.2012,
+      "step": 1817
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.314294398791069e-06,
+      "loss": 2.8366,
+      "step": 1818
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.313277987934366e-06,
+      "loss": 3.1215,
+      "step": 1819
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.312260879872855e-06,
+      "loss": 3.0512,
+      "step": 1820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.311243074770938e-06,
+      "loss": 3.1645,
+      "step": 1821
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.310224572793139e-06,
+      "loss": 2.8821,
+      "step": 1822
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.309205374104088e-06,
+      "loss": 2.9235,
+      "step": 1823
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.30818547886853e-06,
+      "loss": 3.0434,
+      "step": 1824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.307164887251323e-06,
+      "loss": 2.8668,
+      "step": 1825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.306143599417437e-06,
+      "loss": 3.0415,
+      "step": 1826
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.305121615531954e-06,
+      "loss": 2.9138,
+      "step": 1827
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.30409893576007e-06,
+      "loss": 3.0174,
+      "step": 1828
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.303075560267087e-06,
+      "loss": 3.082,
+      "step": 1829
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.302051489218434e-06,
+      "loss": 3.0636,
+      "step": 1830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.301026722779638e-06,
+      "loss": 2.9242,
+      "step": 1831
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.300001261116344e-06,
+      "loss": 2.7931,
+      "step": 1832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.298975104394312e-06,
+      "loss": 3.0713,
+      "step": 1833
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29794825277941e-06,
+      "loss": 2.9691,
+      "step": 1834
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.296920706437618e-06,
+      "loss": 3.0346,
+      "step": 1835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29589246553503e-06,
+      "loss": 2.9991,
+      "step": 1836
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.294863530237856e-06,
+      "loss": 3.1113,
+      "step": 1837
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.293833900712413e-06,
+      "loss": 3.2065,
+      "step": 1838
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.29280357712513e-06,
+      "loss": 2.9533,
+      "step": 1839
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.291772559642554e-06,
+      "loss": 3.0898,
+      "step": 1840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.290740848431334e-06,
+      "loss": 3.1112,
+      "step": 1841
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.289708443658245e-06,
+      "loss": 3.0512,
+      "step": 1842
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.288675345490161e-06,
+      "loss": 2.9263,
+      "step": 1843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.287641554094076e-06,
+      "loss": 2.9203,
+      "step": 1844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.286607069637092e-06,
+      "loss": 2.9432,
+      "step": 1845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.285571892286425e-06,
+      "loss": 3.0052,
+      "step": 1846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.284536022209405e-06,
+      "loss": 3.1011,
+      "step": 1847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.28349945957347e-06,
+      "loss": 2.9628,
+      "step": 1848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.282462204546167e-06,
+      "loss": 2.9957,
+      "step": 1849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.281424257295168e-06,
+      "loss": 3.115,
+      "step": 1850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.280385617988244e-06,
+      "loss": 3.0171,
+      "step": 1851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.279346286793281e-06,
+      "loss": 3.1512,
+      "step": 1852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27830626387828e-06,
+      "loss": 2.9821,
+      "step": 1853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.277265549411351e-06,
+      "loss": 3.1138,
+      "step": 1854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.276224143560717e-06,
+      "loss": 2.8577,
+      "step": 1855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.275182046494712e-06,
+      "loss": 3.0468,
+      "step": 1856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.274139258381785e-06,
+      "loss": 3.1302,
+      "step": 1857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.27309577939049e-06,
+      "loss": 3.0983,
+      "step": 1858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.272051609689496e-06,
+      "loss": 3.0304,
+      "step": 1859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.271006749447589e-06,
+      "loss": 2.8764,
+      "step": 1860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.269961198833657e-06,
+      "loss": 3.2078,
+      "step": 1861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.268914958016708e-06,
+      "loss": 3.0864,
+      "step": 1862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.267868027165857e-06,
+      "loss": 2.9041,
+      "step": 1863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.266820406450327e-06,
+      "loss": 3.0184,
+      "step": 1864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.265772096039464e-06,
+      "loss": 3.131,
+      "step": 1865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.264723096102714e-06,
+      "loss": 2.8822,
+      "step": 1866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.263673406809642e-06,
+      "loss": 3.0052,
+      "step": 1867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.262623028329918e-06,
+      "loss": 3.1538,
+      "step": 1868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.26157196083333e-06,
+      "loss": 2.9337,
+      "step": 1869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.260520204489771e-06,
+      "loss": 2.9682,
+      "step": 1870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25946775946925e-06,
+      "loss": 2.9417,
+      "step": 1871
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.258414625941887e-06,
+      "loss": 3.1212,
+      "step": 1872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.25736080407791e-06,
+      "loss": 2.9806,
+      "step": 1873
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.256306294047662e-06,
+      "loss": 2.8922,
+      "step": 1874
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.255251096021593e-06,
+      "loss": 3.1244,
+      "step": 1875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.254195210170272e-06,
+      "loss": 3.0982,
+      "step": 1876
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.253138636664368e-06,
+      "loss": 3.066,
+      "step": 1877
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.252081375674673e-06,
+      "loss": 2.9308,
+      "step": 1878
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.251023427372077e-06,
+      "loss": 3.0695,
+      "step": 1879
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.249964791927595e-06,
+      "loss": 2.9635,
+      "step": 1880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.248905469512344e-06,
+      "loss": 3.078,
+      "step": 1881
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.247845460297554e-06,
+      "loss": 3.0874,
+      "step": 1882
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.246784764454567e-06,
+      "loss": 3.0875,
+      "step": 1883
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.245723382154835e-06,
+      "loss": 2.9965,
+      "step": 1884
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.244661313569923e-06,
+      "loss": 3.1221,
+      "step": 1885
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.243598558871502e-06,
+      "loss": 2.8833,
+      "step": 1886
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.242535118231363e-06,
+      "loss": 2.908,
+      "step": 1887
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.241470991821397e-06,
+      "loss": 2.9698,
+      "step": 1888
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.240406179813613e-06,
+      "loss": 3.1115,
+      "step": 1889
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.239340682380126e-06,
+      "loss": 3.0492,
+      "step": 1890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.238274499693172e-06,
+      "loss": 3.0562,
+      "step": 1891
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.237207631925082e-06,
+      "loss": 2.9225,
+      "step": 1892
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.23614007924831e-06,
+      "loss": 2.9776,
+      "step": 1893
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.235071841835417e-06,
+      "loss": 2.9918,
+      "step": 1894
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.234002919859075e-06,
+      "loss": 3.1517,
+      "step": 1895
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.232933313492065e-06,
+      "loss": 2.9765,
+      "step": 1896
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.231863022907279e-06,
+      "loss": 2.9245,
+      "step": 1897
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.230792048277722e-06,
+      "loss": 3.0243,
+      "step": 1898
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.229720389776507e-06,
+      "loss": 2.9771,
+      "step": 1899
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.22864804757686e-06,
+      "loss": 3.0025,
+      "step": 1900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.227575021852115e-06,
+      "loss": 3.0598,
+      "step": 1901
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.226501312775715e-06,
+      "loss": 3.1478,
+      "step": 1902
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.225426920521222e-06,
+      "loss": 2.9632,
+      "step": 1903
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.224351845262296e-06,
+      "loss": 2.9458,
+      "step": 1904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.223276087172717e-06,
+      "loss": 2.9884,
+      "step": 1905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.222199646426374e-06,
+      "loss": 3.0056,
+      "step": 1906
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.22112252319726e-06,
+      "loss": 3.0394,
+      "step": 1907
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.220044717659485e-06,
+      "loss": 2.9045,
+      "step": 1908
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.218966229987268e-06,
+      "loss": 2.97,
+      "step": 1909
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.217887060354936e-06,
+      "loss": 3.1079,
+      "step": 1910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.216807208936928e-06,
+      "loss": 3.0819,
+      "step": 1911
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.215726675907794e-06,
+      "loss": 3.0834,
+      "step": 1912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.214645461442194e-06,
+      "loss": 2.8841,
+      "step": 1913
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.21356356571489e-06,
+      "loss": 2.927,
+      "step": 1914
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.212480988900772e-06,
+      "loss": 3.0626,
+      "step": 1915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.211397731174822e-06,
+      "loss": 3.0154,
+      "step": 1916
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.21031379271214e-06,
+      "loss": 2.9871,
+      "step": 1917
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.209229173687939e-06,
+      "loss": 2.8832,
+      "step": 1918
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.208143874277533e-06,
+      "loss": 2.913,
+      "step": 1919
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.207057894656358e-06,
+      "loss": 2.8946,
+      "step": 1920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.205971234999949e-06,
+      "loss": 3.0017,
+      "step": 1921
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.204883895483956e-06,
+      "loss": 3.0429,
+      "step": 1922
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.203795876284139e-06,
+      "loss": 3.069,
+      "step": 1923
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.202707177576366e-06,
+      "loss": 2.9387,
+      "step": 1924
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.201617799536617e-06,
+      "loss": 2.9718,
+      "step": 1925
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.20052774234098e-06,
+      "loss": 3.0049,
+      "step": 1926
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.199437006165653e-06,
+      "loss": 3.0302,
+      "step": 1927
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.198345591186947e-06,
+      "loss": 3.1429,
+      "step": 1928
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.197253497581277e-06,
+      "loss": 2.8129,
+      "step": 1929
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.196160725525172e-06,
+      "loss": 3.0257,
+      "step": 1930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.19506727519527e-06,
+      "loss": 2.9186,
+      "step": 1931
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.193973146768314e-06,
+      "loss": 2.8863,
+      "step": 1932
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.192878340421165e-06,
+      "loss": 3.0522,
+      "step": 1933
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.191782856330789e-06,
+      "loss": 3.0037,
+      "step": 1934
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.19068669467426e-06,
+      "loss": 3.0042,
+      "step": 1935
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.189589855628764e-06,
+      "loss": 3.0195,
+      "step": 1936
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.188492339371594e-06,
+      "loss": 3.1046,
+      "step": 1937
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.187394146080157e-06,
+      "loss": 3.065,
+      "step": 1938
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.186295275931965e-06,
+      "loss": 3.0598,
+      "step": 1939
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.18519572910464e-06,
+      "loss": 2.9086,
+      "step": 1940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.184095505775918e-06,
+      "loss": 3.1689,
+      "step": 1941
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.182994606123636e-06,
+      "loss": 2.9578,
+      "step": 1942
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.18189303032575e-06,
+      "loss": 2.9206,
+      "step": 1943
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.180790778560316e-06,
+      "loss": 3.1901,
+      "step": 1944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.179687851005507e-06,
+      "loss": 3.0997,
+      "step": 1945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.178584247839601e-06,
+      "loss": 3.0049,
+      "step": 1946
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.177479969240985e-06,
+      "loss": 3.1029,
+      "step": 1947
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.176375015388157e-06,
+      "loss": 3.0924,
+      "step": 1948
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.175269386459723e-06,
+      "loss": 3.0278,
+      "step": 1949
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.174163082634398e-06,
+      "loss": 2.9763,
+      "step": 1950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.173056104091008e-06,
+      "loss": 3.1585,
+      "step": 1951
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.171948451008487e-06,
+      "loss": 2.9279,
+      "step": 1952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.170840123565876e-06,
+      "loss": 3.1406,
+      "step": 1953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.169731121942329e-06,
+      "loss": 2.9436,
+      "step": 1954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.168621446317104e-06,
+      "loss": 3.0896,
+      "step": 1955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.167511096869573e-06,
+      "loss": 3.0592,
+      "step": 1956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.166400073779212e-06,
+      "loss": 3.0573,
+      "step": 1957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.165288377225612e-06,
+      "loss": 2.9649,
+      "step": 1958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.164176007388466e-06,
+      "loss": 3.0891,
+      "step": 1959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.16306296444758e-06,
+      "loss": 2.9206,
+      "step": 1960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.161949248582869e-06,
+      "loss": 2.9085,
+      "step": 1961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.160834859974357e-06,
+      "loss": 2.9083,
+      "step": 1962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.159719798802171e-06,
+      "loss": 2.901,
+      "step": 1963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.158604065246555e-06,
+      "loss": 2.9073,
+      "step": 1964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.157487659487856e-06,
+      "loss": 2.9967,
+      "step": 1965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.156370581706535e-06,
+      "loss": 3.0133,
+      "step": 1966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.155252832083152e-06,
+      "loss": 2.9069,
+      "step": 1967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.154134410798386e-06,
+      "loss": 3.0638,
+      "step": 1968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.153015318033021e-06,
+      "loss": 3.0487,
+      "step": 1969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.151895553967946e-06,
+      "loss": 2.9259,
+      "step": 1970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.150775118784164e-06,
+      "loss": 3.1579,
+      "step": 1971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.14965401266278e-06,
+      "loss": 3.057,
+      "step": 1972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.148532235785016e-06,
+      "loss": 3.0588,
+      "step": 1973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.147409788332193e-06,
+      "loss": 2.9454,
+      "step": 1974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.14628667048575e-06,
+      "loss": 3.025,
+      "step": 1975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.145162882427226e-06,
+      "loss": 3.1107,
+      "step": 1976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.144038424338273e-06,
+      "loss": 3.061,
+      "step": 1977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.142913296400651e-06,
+      "loss": 2.9877,
+      "step": 1978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.141787498796226e-06,
+      "loss": 3.0146,
+      "step": 1979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.140661031706975e-06,
+      "loss": 3.0046,
+      "step": 1980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.139533895314979e-06,
+      "loss": 2.9954,
+      "step": 1981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.138406089802432e-06,
+      "loss": 3.0394,
+      "step": 1982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.137277615351635e-06,
+      "loss": 3.1675,
+      "step": 1983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.136148472144996e-06,
+      "loss": 3.1164,
+      "step": 1984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.13501866036503e-06,
+      "loss": 3.086,
+      "step": 1985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.133888180194363e-06,
+      "loss": 3.0725,
+      "step": 1986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.132757031815727e-06,
+      "loss": 2.8819,
+      "step": 1987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.131625215411962e-06,
+      "loss": 2.9677,
+      "step": 1988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.130492731166016e-06,
+      "loss": 2.9244,
+      "step": 1989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.129359579260947e-06,
+      "loss": 3.0656,
+      "step": 1990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.12822575987992e-06,
+      "loss": 2.9798,
+      "step": 1991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.127091273206206e-06,
+      "loss": 3.0265,
+      "step": 1992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.125956119423185e-06,
+      "loss": 2.9461,
+      "step": 1993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.124820298714345e-06,
+      "loss": 3.026,
+      "step": 1994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.123683811263283e-06,
+      "loss": 3.0352,
+      "step": 1995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.122546657253701e-06,
+      "loss": 3.0703,
+      "step": 1996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.121408836869411e-06,
+      "loss": 2.9737,
+      "step": 1997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.120270350294332e-06,
+      "loss": 2.8946,
+      "step": 1998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.11913119771249e-06,
+      "loss": 3.055,
+      "step": 1999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.11799137930802e-06,
+      "loss": 3.0778,
+      "step": 2000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.116850895265166e-06,
+      "loss": 3.0144,
+      "step": 2001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.115709745768272e-06,
+      "loss": 3.0043,
+      "step": 2002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.114567931001802e-06,
+      "loss": 3.085,
+      "step": 2003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.113425451150317e-06,
+      "loss": 2.9302,
+      "step": 2004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.112282306398488e-06,
+      "loss": 2.9327,
+      "step": 2005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.111138496931097e-06,
+      "loss": 3.1455,
+      "step": 2006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.10999402293303e-06,
+      "loss": 3.0408,
+      "step": 2007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.108848884589281e-06,
+      "loss": 2.9265,
+      "step": 2008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.107703082084956e-06,
+      "loss": 2.9659,
+      "step": 2009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.106556615605258e-06,
+      "loss": 3.0192,
+      "step": 2010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.105409485335507e-06,
+      "loss": 3.0619,
+      "step": 2011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.104261691461127e-06,
+      "loss": 3.012,
+      "step": 2012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.103113234167649e-06,
+      "loss": 3.0857,
+      "step": 2013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.10196411364071e-06,
+      "loss": 2.862,
+      "step": 2014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.100814330066057e-06,
+      "loss": 2.9122,
+      "step": 2015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.099663883629542e-06,
+      "loss": 2.9587,
+      "step": 2016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.098512774517126e-06,
+      "loss": 2.9485,
+      "step": 2017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.097361002914874e-06,
+      "loss": 3.0121,
+      "step": 2018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.096208569008963e-06,
+      "loss": 2.9634,
+      "step": 2019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.095055472985672e-06,
+      "loss": 2.9961,
+      "step": 2020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.09390171503139e-06,
+      "loss": 3.0793,
+      "step": 2021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.092747295332612e-06,
+      "loss": 3.1557,
+      "step": 2022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.091592214075938e-06,
+      "loss": 3.0324,
+      "step": 2023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.090436471448081e-06,
+      "loss": 3.1392,
+      "step": 2024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.089280067635855e-06,
+      "loss": 2.9349,
+      "step": 2025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.088123002826183e-06,
+      "loss": 3.0584,
+      "step": 2026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.086965277206095e-06,
+      "loss": 3.1018,
+      "step": 2027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.085806890962727e-06,
+      "loss": 2.928,
+      "step": 2028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.084647844283324e-06,
+      "loss": 2.9199,
+      "step": 2029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.083488137355231e-06,
+      "loss": 3.1375,
+      "step": 2030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.082327770365911e-06,
+      "loss": 3.0454,
+      "step": 2031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.081166743502926e-06,
+      "loss": 2.8991,
+      "step": 2032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.080005056953945e-06,
+      "loss": 3.0637,
+      "step": 2033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.078842710906742e-06,
+      "loss": 2.8845,
+      "step": 2034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.077679705549204e-06,
+      "loss": 2.9301,
+      "step": 2035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.076516041069323e-06,
+      "loss": 3.0255,
+      "step": 2036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.075351717655189e-06,
+      "loss": 3.0869,
+      "step": 2037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.074186735495012e-06,
+      "loss": 3.0254,
+      "step": 2038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.073021094777097e-06,
+      "loss": 2.8502,
+      "step": 2039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.07185479568986e-06,
+      "loss": 2.9584,
+      "step": 2040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.070687838421826e-06,
+      "loss": 3.0772,
+      "step": 2041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.069520223161623e-06,
+      "loss": 2.9592,
+      "step": 2042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.068351950097984e-06,
+      "loss": 2.9788,
+      "step": 2043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.067183019419752e-06,
+      "loss": 2.9863,
+      "step": 2044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.066013431315875e-06,
+      "loss": 3.0384,
+      "step": 2045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.064843185975405e-06,
+      "loss": 3.0651,
+      "step": 2046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.063672283587504e-06,
+      "loss": 2.9557,
+      "step": 2047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.062500724341438e-06,
+      "loss": 3.1018,
+      "step": 2048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.06132850842658e-06,
+      "loss": 2.9168,
+      "step": 2049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.06015563603241e-06,
+      "loss": 3.0976,
+      "step": 2050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.058982107348512e-06,
+      "loss": 3.0663,
+      "step": 2051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.057807922564575e-06,
+      "loss": 2.932,
+      "step": 2052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.056633081870397e-06,
+      "loss": 2.867,
+      "step": 2053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.055457585455882e-06,
+      "loss": 3.0394,
+      "step": 2054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.05428143351104e-06,
+      "loss": 3.0076,
+      "step": 2055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.053104626225982e-06,
+      "loss": 2.9676,
+      "step": 2056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.051927163790931e-06,
+      "loss": 2.832,
+      "step": 2057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.050749046396216e-06,
+      "loss": 3.0404,
+      "step": 2058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.049570274232267e-06,
+      "loss": 3.0621,
+      "step": 2059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.048390847489624e-06,
+      "loss": 2.9531,
+      "step": 2060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.047210766358932e-06,
+      "loss": 2.8368,
+      "step": 2061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.046030031030937e-06,
+      "loss": 3.1481,
+      "step": 2062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.044848641696499e-06,
+      "loss": 3.0262,
+      "step": 2063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.043666598546579e-06,
+      "loss": 3.1255,
+      "step": 2064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.042483901772244e-06,
+      "loss": 3.0703,
+      "step": 2065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.041300551564664e-06,
+      "loss": 3.0045,
+      "step": 2066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.04011654811512e-06,
+      "loss": 3.0757,
+      "step": 2067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.038931891614999e-06,
+      "loss": 3.0533,
+      "step": 2068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.037746582255787e-06,
+      "loss": 3.03,
+      "step": 2069
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.036560620229077e-06,
+      "loss": 2.948,
+      "step": 2070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.035374005726577e-06,
+      "loss": 3.1442,
+      "step": 2071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.034186738940085e-06,
+      "loss": 2.8998,
+      "step": 2072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.032998820061518e-06,
+      "loss": 2.9463,
+      "step": 2073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.031810249282892e-06,
+      "loss": 3.0388,
+      "step": 2074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.03062102679633e-06,
+      "loss": 3.1005,
+      "step": 2075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.029431152794054e-06,
+      "loss": 3.0269,
+      "step": 2076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.028240627468406e-06,
+      "loss": 3.0069,
+      "step": 2077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.027049451011818e-06,
+      "loss": 3.1501,
+      "step": 2078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.025857623616837e-06,
+      "loss": 3.1065,
+      "step": 2079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.024665145476109e-06,
+      "loss": 3.0937,
+      "step": 2080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.023472016782388e-06,
+      "loss": 3.0486,
+      "step": 2081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.022278237728536e-06,
+      "loss": 3.1616,
+      "step": 2082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.021083808507516e-06,
+      "loss": 3.1258,
+      "step": 2083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.019888729312398e-06,
+      "loss": 2.9847,
+      "step": 2084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.018693000336353e-06,
+      "loss": 3.1191,
+      "step": 2085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.017496621772665e-06,
+      "loss": 3.0061,
+      "step": 2086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.016299593814715e-06,
+      "loss": 3.0578,
+      "step": 2087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.015101916655995e-06,
+      "loss": 3.0433,
+      "step": 2088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.013903590490097e-06,
+      "loss": 2.8903,
+      "step": 2089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.012704615510722e-06,
+      "loss": 3.1705,
+      "step": 2090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.011504991911673e-06,
+      "loss": 3.0541,
+      "step": 2091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.01030471988686e-06,
+      "loss": 3.0376,
+      "step": 2092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.009103799630296e-06,
+      "loss": 2.8931,
+      "step": 2093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.0079022313361e-06,
+      "loss": 3.0866,
+      "step": 2094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.006700015198495e-06,
+      "loss": 3.053,
+      "step": 2095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.00549715141181e-06,
+      "loss": 2.9092,
+      "step": 2096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.004293640170477e-06,
+      "loss": 3.0918,
+      "step": 2097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.003089481669031e-06,
+      "loss": 2.9531,
+      "step": 2098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.00188467610212e-06,
+      "loss": 3.0133,
+      "step": 2099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.000679223664484e-06,
+      "loss": 3.0163,
+      "step": 2100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.99947312455098e-06,
+      "loss": 3.044,
+      "step": 2101
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.998266378956557e-06,
+      "loss": 2.8747,
+      "step": 2102
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.997058987076281e-06,
+      "loss": 3.0387,
+      "step": 2103
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.995850949105316e-06,
+      "loss": 3.0825,
+      "step": 2104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.994642265238928e-06,
+      "loss": 3.0975,
+      "step": 2105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.993432935672494e-06,
+      "loss": 3.1151,
+      "step": 2106
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.992222960601489e-06,
+      "loss": 2.8761,
+      "step": 2107
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.991012340221497e-06,
+      "loss": 2.8665,
+      "step": 2108
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.989801074728202e-06,
+      "loss": 2.9714,
+      "step": 2109
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.988589164317398e-06,
+      "loss": 3.0457,
+      "step": 2110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.987376609184978e-06,
+      "loss": 3.1051,
+      "step": 2111
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.986163409526944e-06,
+      "loss": 2.9133,
+      "step": 2112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.984949565539396e-06,
+      "loss": 2.9359,
+      "step": 2113
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.983735077418543e-06,
+      "loss": 2.9295,
+      "step": 2114
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.982519945360697e-06,
+      "loss": 2.9121,
+      "step": 2115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.981304169562273e-06,
+      "loss": 3.1025,
+      "step": 2116
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.980087750219792e-06,
+      "loss": 3.0133,
+      "step": 2117
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.978870687529876e-06,
+      "loss": 2.8413,
+      "step": 2118
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.977652981689255e-06,
+      "loss": 2.9038,
+      "step": 2119
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.97643463289476e-06,
+      "loss": 3.0689,
+      "step": 2120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.975215641343327e-06,
+      "loss": 2.9806,
+      "step": 2121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.973996007231996e-06,
+      "loss": 2.9934,
+      "step": 2122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.972775730757907e-06,
+      "loss": 3.0063,
+      "step": 2123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.971554812118313e-06,
+      "loss": 2.9918,
+      "step": 2124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.97033325151056e-06,
+      "loss": 3.1107,
+      "step": 2125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.969111049132106e-06,
+      "loss": 3.0558,
+      "step": 2126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.967888205180509e-06,
+      "loss": 3.1309,
+      "step": 2127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.966664719853432e-06,
+      "loss": 2.9183,
+      "step": 2128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.96544059334864e-06,
+      "loss": 2.9474,
+      "step": 2129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.964215825864002e-06,
+      "loss": 3.1104,
+      "step": 2130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.962990417597492e-06,
+      "loss": 3.1129,
+      "step": 2131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.961764368747187e-06,
+      "loss": 3.0761,
+      "step": 2132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.96053767951127e-06,
+      "loss": 2.8639,
+      "step": 2133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.959310350088019e-06,
+      "loss": 2.9064,
+      "step": 2134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.958082380675828e-06,
+      "loss": 2.9328,
+      "step": 2135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.956853771473182e-06,
+      "loss": 2.9496,
+      "step": 2136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.955624522678678e-06,
+      "loss": 3.1856,
+      "step": 2137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.954394634491016e-06,
+      "loss": 3.0058,
+      "step": 2138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.95316410710899e-06,
+      "loss": 2.9198,
+      "step": 2139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.951932940731514e-06,
+      "loss": 3.131,
+      "step": 2140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.950701135557588e-06,
+      "loss": 2.8448,
+      "step": 2141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.949468691786325e-06,
+      "loss": 3.0459,
+      "step": 2142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.94823560961694e-06,
+      "loss": 2.9543,
+      "step": 2143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.94700188924875e-06,
+      "loss": 2.9251,
+      "step": 2144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.945767530881172e-06,
+      "loss": 3.0333,
+      "step": 2145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.944532534713737e-06,
+      "loss": 2.9699,
+      "step": 2146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.943296900946064e-06,
+      "loss": 3.0163,
+      "step": 2147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.942060629777888e-06,
+      "loss": 3.0869,
+      "step": 2148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.940823721409039e-06,
+      "loss": 2.8551,
+      "step": 2149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.939586176039454e-06,
+      "loss": 3.0237,
+      "step": 2150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.93834799386917e-06,
+      "loss": 2.8933,
+      "step": 2151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.937109175098332e-06,
+      "loss": 2.9783,
+      "step": 2152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.935869719927179e-06,
+      "loss": 2.9135,
+      "step": 2153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.934629628556064e-06,
+      "loss": 3.0245,
+      "step": 2154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.933388901185436e-06,
+      "loss": 3.0844,
+      "step": 2155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.932147538015847e-06,
+      "loss": 2.9693,
+      "step": 2156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.930905539247952e-06,
+      "loss": 2.9241,
+      "step": 2157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.92966290508251e-06,
+      "loss": 2.9993,
+      "step": 2158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.928419635720383e-06,
+      "loss": 3.0954,
+      "step": 2159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.927175731362536e-06,
+      "loss": 2.9725,
+      "step": 2160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.925931192210033e-06,
+      "loss": 2.9546,
+      "step": 2161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.924686018464045e-06,
+      "loss": 3.0423,
+      "step": 2162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.923440210325844e-06,
+      "loss": 3.1195,
+      "step": 2163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.922193767996803e-06,
+      "loss": 3.0239,
+      "step": 2164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.920946691678401e-06,
+      "loss": 2.9734,
+      "step": 2165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.919698981572213e-06,
+      "loss": 2.8494,
+      "step": 2166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.918450637879926e-06,
+      "loss": 2.9113,
+      "step": 2167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.91720166080332e-06,
+      "loss": 2.9708,
+      "step": 2168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.915952050544285e-06,
+      "loss": 2.9,
+      "step": 2169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.914701807304809e-06,
+      "loss": 3.0308,
+      "step": 2170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.91345093128698e-06,
+      "loss": 3.0104,
+      "step": 2171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.912199422692995e-06,
+      "loss": 2.9248,
+      "step": 2172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.910947281725149e-06,
+      "loss": 3.0017,
+      "step": 2173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.909694508585838e-06,
+      "loss": 3.0386,
+      "step": 2174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.908441103477565e-06,
+      "loss": 2.9899,
+      "step": 2175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.90718706660293e-06,
+      "loss": 3.0822,
+      "step": 2176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.90593239816464e-06,
+      "loss": 2.9562,
+      "step": 2177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.904677098365499e-06,
+      "loss": 3.1011,
+      "step": 2178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.903421167408418e-06,
+      "loss": 2.8739,
+      "step": 2179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.902164605496406e-06,
+      "loss": 3.0071,
+      "step": 2180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.900907412832573e-06,
+      "loss": 3.1565,
+      "step": 2181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.899649589620139e-06,
+      "loss": 2.988,
+      "step": 2182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.898391136062417e-06,
+      "loss": 2.9102,
+      "step": 2183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.897132052362827e-06,
+      "loss": 2.8932,
+      "step": 2184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.895872338724887e-06,
+      "loss": 2.9605,
+      "step": 2185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.89461199535222e-06,
+      "loss": 3.1183,
+      "step": 2186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.893351022448551e-06,
+      "loss": 3.1527,
+      "step": 2187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.892089420217706e-06,
+      "loss": 3.0045,
+      "step": 2188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.89082718886361e-06,
+      "loss": 2.9143,
+      "step": 2189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.889564328590293e-06,
+      "loss": 2.976,
+      "step": 2190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.888300839601884e-06,
+      "loss": 2.862,
+      "step": 2191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.88703672210262e-06,
+      "loss": 3.1854,
+      "step": 2192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.88577197629683e-06,
+      "loss": 2.9877,
+      "step": 2193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.884506602388952e-06,
+      "loss": 3.1344,
+      "step": 2194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.883240600583522e-06,
+      "loss": 3.0691,
+      "step": 2195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.881973971085178e-06,
+      "loss": 2.9696,
+      "step": 2196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.88070671409866e-06,
+      "loss": 2.8997,
+      "step": 2197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.879438829828813e-06,
+      "loss": 2.9963,
+      "step": 2198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.878170318480575e-06,
+      "loss": 3.0343,
+      "step": 2199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.876901180258991e-06,
+      "loss": 2.891,
+      "step": 2200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.875631415369208e-06,
+      "loss": 3.0444,
+      "step": 2201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.874361024016472e-06,
+      "loss": 3.0146,
+      "step": 2202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.87309000640613e-06,
+      "loss": 3.2445,
+      "step": 2203
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.871818362743633e-06,
+      "loss": 3.0018,
+      "step": 2204
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.870546093234531e-06,
+      "loss": 2.9361,
+      "step": 2205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.869273198084473e-06,
+      "loss": 2.9879,
+      "step": 2206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.867999677499218e-06,
+      "loss": 2.8805,
+      "step": 2207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.866725531684612e-06,
+      "loss": 2.9161,
+      "step": 2208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.865450760846615e-06,
+      "loss": 3.1182,
+      "step": 2209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.864175365191281e-06,
+      "loss": 2.9176,
+      "step": 2210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.862899344924767e-06,
+      "loss": 2.9686,
+      "step": 2211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.861622700253332e-06,
+      "loss": 2.9138,
+      "step": 2212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.860345431383333e-06,
+      "loss": 2.906,
+      "step": 2213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.859067538521231e-06,
+      "loss": 3.0006,
+      "step": 2214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.857789021873586e-06,
+      "loss": 3.1346,
+      "step": 2215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.85650988164706e-06,
+      "loss": 2.9866,
+      "step": 2216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.855230118048415e-06,
+      "loss": 2.8583,
+      "step": 2217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.853949731284515e-06,
+      "loss": 2.8453,
+      "step": 2218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.85266872156232e-06,
+      "loss": 3.0733,
+      "step": 2219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.851387089088897e-06,
+      "loss": 2.9437,
+      "step": 2220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.850104834071411e-06,
+      "loss": 2.9604,
+      "step": 2221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.848821956717127e-06,
+      "loss": 3.09,
+      "step": 2222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.847538457233411e-06,
+      "loss": 3.0296,
+      "step": 2223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.846254335827733e-06,
+      "loss": 3.0894,
+      "step": 2224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.844969592707657e-06,
+      "loss": 2.9589,
+      "step": 2225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.84368422808085e-06,
+      "loss": 2.9722,
+      "step": 2226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.842398242155083e-06,
+      "loss": 2.9769,
+      "step": 2227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.841111635138224e-06,
+      "loss": 2.9911,
+      "step": 2228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.83982440723824e-06,
+      "loss": 3.0463,
+      "step": 2229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.838536558663201e-06,
+      "loss": 2.9972,
+      "step": 2230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.837248089621282e-06,
+      "loss": 2.9517,
+      "step": 2231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.835959000320747e-06,
+      "loss": 2.991,
+      "step": 2232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.834669290969968e-06,
+      "loss": 2.8185,
+      "step": 2233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.833378961777416e-06,
+      "loss": 3.0572,
+      "step": 2234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.832088012951663e-06,
+      "loss": 2.9003,
+      "step": 2235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.83079644470138e-06,
+      "loss": 3.0504,
+      "step": 2236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.829504257235332e-06,
+      "loss": 3.0897,
+      "step": 2237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.8282114507624e-06,
+      "loss": 3.0511,
+      "step": 2238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.826918025491547e-06,
+      "loss": 3.0551,
+      "step": 2239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.825623981631848e-06,
+      "loss": 3.0284,
+      "step": 2240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.824329319392476e-06,
+      "loss": 2.9644,
+      "step": 2241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.823034038982696e-06,
+      "loss": 2.8513,
+      "step": 2242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.821738140611884e-06,
+      "loss": 2.9868,
+      "step": 2243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.820441624489511e-06,
+      "loss": 3.0072,
+      "step": 2244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.819144490825144e-06,
+      "loss": 3.0093,
+      "step": 2245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.817846739828459e-06,
+      "loss": 3.0772,
+      "step": 2246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.816548371709224e-06,
+      "loss": 2.9072,
+      "step": 2247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.815249386677307e-06,
+      "loss": 3.0232,
+      "step": 2248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.813949784942682e-06,
+      "loss": 3.099,
+      "step": 2249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.812649566715415e-06,
+      "loss": 3.075,
+      "step": 2250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.811348732205676e-06,
+      "loss": 3.0487,
+      "step": 2251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.810047281623734e-06,
+      "loss": 3.0048,
+      "step": 2252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.808745215179961e-06,
+      "loss": 3.0894,
+      "step": 2253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.80744253308482e-06,
+      "loss": 3.0876,
+      "step": 2254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.80613923554888e-06,
+      "loss": 3.1243,
+      "step": 2255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.80483532278281e-06,
+      "loss": 3.1416,
+      "step": 2256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.803530794997376e-06,
+      "loss": 2.925,
+      "step": 2257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.80222565240344e-06,
+      "loss": 2.8464,
+      "step": 2258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.80091989521197e-06,
+      "loss": 3.0707,
+      "step": 2259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.799613523634033e-06,
+      "loss": 2.9006,
+      "step": 2260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.79830653788079e-06,
+      "loss": 3.0065,
+      "step": 2261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.796998938163503e-06,
+      "loss": 2.9803,
+      "step": 2262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.795690724693538e-06,
+      "loss": 3.0992,
+      "step": 2263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.794381897682354e-06,
+      "loss": 2.9574,
+      "step": 2264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.793072457341512e-06,
+      "loss": 3.1148,
+      "step": 2265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.791762403882674e-06,
+      "loss": 2.9873,
+      "step": 2266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.790451737517595e-06,
+      "loss": 3.0174,
+      "step": 2267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.78914045845814e-06,
+      "loss": 3.0109,
+      "step": 2268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.787828566916257e-06,
+      "loss": 2.9842,
+      "step": 2269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.786516063104009e-06,
+      "loss": 2.9906,
+      "step": 2270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.785202947233547e-06,
+      "loss": 3.0407,
+      "step": 2271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.783889219517129e-06,
+      "loss": 2.8413,
+      "step": 2272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.782574880167106e-06,
+      "loss": 2.9952,
+      "step": 2273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.781259929395929e-06,
+      "loss": 2.9429,
+      "step": 2274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.779944367416149e-06,
+      "loss": 3.1132,
+      "step": 2275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.778628194440415e-06,
+      "loss": 3.0462,
+      "step": 2276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.77731141068148e-06,
+      "loss": 3.0772,
+      "step": 2277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.775994016352183e-06,
+      "loss": 3.0859,
+      "step": 2278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.774676011665475e-06,
+      "loss": 3.0672,
+      "step": 2279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.7733573968344e-06,
+      "loss": 3.067,
+      "step": 2280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.772038172072098e-06,
+      "loss": 2.9516,
+      "step": 2281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.770718337591813e-06,
+      "loss": 2.9338,
+      "step": 2282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.769397893606884e-06,
+      "loss": 3.0914,
+      "step": 2283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.76807684033075e-06,
+      "loss": 3.0375,
+      "step": 2284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.766755177976947e-06,
+      "loss": 3.0129,
+      "step": 2285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.765432906759112e-06,
+      "loss": 2.9595,
+      "step": 2286
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.764110026890979e-06,
+      "loss": 3.1031,
+      "step": 2287
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.762786538586381e-06,
+      "loss": 2.9719,
+      "step": 2288
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.761462442059245e-06,
+      "loss": 3.0892,
+      "step": 2289
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.760137737523604e-06,
+      "loss": 3.0404,
+      "step": 2290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.758812425193583e-06,
+      "loss": 3.0414,
+      "step": 2291
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.757486505283407e-06,
+      "loss": 2.9014,
+      "step": 2292
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.756159978007403e-06,
+      "loss": 3.1211,
+      "step": 2293
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.754832843579989e-06,
+      "loss": 2.9726,
+      "step": 2294
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.753505102215686e-06,
+      "loss": 3.0274,
+      "step": 2295
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.752176754129115e-06,
+      "loss": 2.9026,
+      "step": 2296
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.75084779953499e-06,
+      "loss": 3.0073,
+      "step": 2297
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.749518238648123e-06,
+      "loss": 3.0418,
+      "step": 2298
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.74818807168343e-06,
+      "loss": 2.9095,
+      "step": 2299
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.746857298855919e-06,
+      "loss": 2.9511,
+      "step": 2300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.745525920380701e-06,
+      "loss": 3.0463,
+      "step": 2301
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.744193936472978e-06,
+      "loss": 3.0407,
+      "step": 2302
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.742861347348056e-06,
+      "loss": 3.09,
+      "step": 2303
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.741528153221337e-06,
+      "loss": 2.8709,
+      "step": 2304
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.74019435430832e-06,
+      "loss": 2.9846,
+      "step": 2305
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.738859950824604e-06,
+      "loss": 3.1819,
+      "step": 2306
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.73752494298588e-06,
+      "loss": 3.1364,
+      "step": 2307
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.736189331007942e-06,
+      "loss": 2.9236,
+      "step": 2308
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.734853115106685e-06,
+      "loss": 3.0179,
+      "step": 2309
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.733516295498092e-06,
+      "loss": 2.8938,
+      "step": 2310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.732178872398249e-06,
+      "loss": 2.9704,
+      "step": 2311
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.73084084602334e-06,
+      "loss": 3.0619,
+      "step": 2312
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.729502216589647e-06,
+      "loss": 3.0846,
+      "step": 2313
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.728162984313545e-06,
+      "loss": 2.8442,
+      "step": 2314
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.726823149411513e-06,
+      "loss": 2.9994,
+      "step": 2315
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.725482712100122e-06,
+      "loss": 3.0964,
+      "step": 2316
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.724141672596043e-06,
+      "loss": 2.8325,
+      "step": 2317
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.72280003111604e-06,
+      "loss": 3.0296,
+      "step": 2318
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.721457787876986e-06,
+      "loss": 2.8081,
+      "step": 2319
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.720114943095836e-06,
+      "loss": 3.113,
+      "step": 2320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.71877149698965e-06,
+      "loss": 2.9647,
+      "step": 2321
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.71742744977559e-06,
+      "loss": 3.0884,
+      "step": 2322
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.716082801670904e-06,
+      "loss": 2.8289,
+      "step": 2323
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.714737552892946e-06,
+      "loss": 2.9928,
+      "step": 2324
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.713391703659166e-06,
+      "loss": 2.9661,
+      "step": 2325
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.712045254187103e-06,
+      "loss": 2.9742,
+      "step": 2326
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.710698204694406e-06,
+      "loss": 3.1063,
+      "step": 2327
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.709350555398807e-06,
+      "loss": 3.041,
+      "step": 2328
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.708002306518148e-06,
+      "loss": 3.0547,
+      "step": 2329
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.706653458270361e-06,
+      "loss": 3.0107,
+      "step": 2330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.705304010873472e-06,
+      "loss": 2.9486,
+      "step": 2331
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.703953964545614e-06,
+      "loss": 2.962,
+      "step": 2332
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.702603319505005e-06,
+      "loss": 2.9263,
+      "step": 2333
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.701252075969968e-06,
+      "loss": 3.0397,
+      "step": 2334
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.699900234158917e-06,
+      "loss": 3.0374,
+      "step": 2335
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.698547794290372e-06,
+      "loss": 2.9994,
+      "step": 2336
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.697194756582938e-06,
+      "loss": 3.0587,
+      "step": 2337
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.695841121255322e-06,
+      "loss": 3.0053,
+      "step": 2338
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.694486888526333e-06,
+      "loss": 2.8525,
+      "step": 2339
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.693132058614865e-06,
+      "loss": 3.0281,
+      "step": 2340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.691776631739918e-06,
+      "loss": 2.9812,
+      "step": 2341
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.690420608120585e-06,
+      "loss": 3.1067,
+      "step": 2342
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.689063987976053e-06,
+      "loss": 2.9931,
+      "step": 2343
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.687706771525614e-06,
+      "loss": 3.0514,
+      "step": 2344
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.686348958988646e-06,
+      "loss": 2.9799,
+      "step": 2345
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.68499055058463e-06,
+      "loss": 2.968,
+      "step": 2346
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.683631546533136e-06,
+      "loss": 3.0974,
+      "step": 2347
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.682271947053842e-06,
+      "loss": 3.0189,
+      "step": 2348
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.680911752366514e-06,
+      "loss": 2.9782,
+      "step": 2349
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.679550962691014e-06,
+      "loss": 2.9568,
+      "step": 2350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.678189578247305e-06,
+      "loss": 3.0427,
+      "step": 2351
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.676827599255441e-06,
+      "loss": 2.8786,
+      "step": 2352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.675465025935574e-06,
+      "loss": 3.0274,
+      "step": 2353
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.674101858507955e-06,
+      "loss": 3.0124,
+      "step": 2354
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.672738097192926e-06,
+      "loss": 3.0033,
+      "step": 2355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.671373742210929e-06,
+      "loss": 2.9208,
+      "step": 2356
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.670008793782498e-06,
+      "loss": 3.1593,
+      "step": 2357
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.668643252128269e-06,
+      "loss": 2.9877,
+      "step": 2358
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.667277117468966e-06,
+      "loss": 2.9921,
+      "step": 2359
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.665910390025418e-06,
+      "loss": 2.9916,
+      "step": 2360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.664543070018541e-06,
+      "loss": 2.9971,
+      "step": 2361
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.66317515766935e-06,
+      "loss": 3.0912,
+      "step": 2362
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.661806653198961e-06,
+      "loss": 2.9266,
+      "step": 2363
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.660437556828577e-06,
+      "loss": 3.0786,
+      "step": 2364
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.659067868779505e-06,
+      "loss": 3.0159,
+      "step": 2365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.657697589273139e-06,
+      "loss": 3.0869,
+      "step": 2366
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.656326718530975e-06,
+      "loss": 3.0282,
+      "step": 2367
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.654955256774602e-06,
+      "loss": 2.9323,
+      "step": 2368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.653583204225706e-06,
+      "loss": 2.9028,
+      "step": 2369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.652210561106068e-06,
+      "loss": 3.0646,
+      "step": 2370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.650837327637562e-06,
+      "loss": 3.023,
+      "step": 2371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.649463504042162e-06,
+      "loss": 3.0416,
+      "step": 2372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.648089090541935e-06,
+      "loss": 3.126,
+      "step": 2373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.646714087359041e-06,
+      "loss": 3.0795,
+      "step": 2374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.645338494715738e-06,
+      "loss": 2.918,
+      "step": 2375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.64396231283438e-06,
+      "loss": 2.9267,
+      "step": 2376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.642585541937416e-06,
+      "loss": 2.9802,
+      "step": 2377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.641208182247388e-06,
+      "loss": 3.0692,
+      "step": 2378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.639830233986933e-06,
+      "loss": 3.0851,
+      "step": 2379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.63845169737879e-06,
+      "loss": 2.9031,
+      "step": 2380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.637072572645781e-06,
+      "loss": 3.0012,
+      "step": 2381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.635692860010833e-06,
+      "loss": 3.0617,
+      "step": 2382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.634312559696965e-06,
+      "loss": 3.0734,
+      "step": 2383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.632931671927291e-06,
+      "loss": 3.0765,
+      "step": 2384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.63155019692502e-06,
+      "loss": 2.9272,
+      "step": 2385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.630168134913453e-06,
+      "loss": 3.1117,
+      "step": 2386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.628785486115993e-06,
+      "loss": 3.1892,
+      "step": 2387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.62740225075613e-06,
+      "loss": 2.8927,
+      "step": 2388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.626018429057454e-06,
+      "loss": 3.017,
+      "step": 2389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.624634021243647e-06,
+      "loss": 3.0221,
+      "step": 2390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.623249027538488e-06,
+      "loss": 2.9537,
+      "step": 2391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.621863448165847e-06,
+      "loss": 3.0426,
+      "step": 2392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.620477283349696e-06,
+      "loss": 3.1606,
+      "step": 2393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.619090533314091e-06,
+      "loss": 2.9563,
+      "step": 2394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.617703198283192e-06,
+      "loss": 3.0193,
+      "step": 2395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.616315278481248e-06,
+      "loss": 3.0212,
+      "step": 2396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.614926774132607e-06,
+      "loss": 2.9209,
+      "step": 2397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.613537685461707e-06,
+      "loss": 2.9921,
+      "step": 2398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.612148012693082e-06,
+      "loss": 3.0458,
+      "step": 2399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.610757756051363e-06,
+      "loss": 2.8784,
+      "step": 2400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.609366915761271e-06,
+      "loss": 3.079,
+      "step": 2401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.607975492047624e-06,
+      "loss": 3.0222,
+      "step": 2402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.606583485135336e-06,
+      "loss": 2.9659,
+      "step": 2403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.605190895249411e-06,
+      "loss": 3.022,
+      "step": 2404
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.603797722614948e-06,
+      "loss": 3.122,
+      "step": 2405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.602403967457145e-06,
+      "loss": 3.151,
+      "step": 2406
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.601009630001287e-06,
+      "loss": 2.9626,
+      "step": 2407
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.59961471047276e-06,
+      "loss": 3.1058,
+      "step": 2408
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.598219209097041e-06,
+      "loss": 3.1114,
+      "step": 2409
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.596823126099699e-06,
+      "loss": 3.0562,
+      "step": 2410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.595426461706398e-06,
+      "loss": 2.9657,
+      "step": 2411
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.594029216142898e-06,
+      "loss": 2.8955,
+      "step": 2412
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.592631389635055e-06,
+      "loss": 2.981,
+      "step": 2413
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.591232982408813e-06,
+      "loss": 3.0687,
+      "step": 2414
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.589833994690211e-06,
+      "loss": 2.9637,
+      "step": 2415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.588434426705387e-06,
+      "loss": 3.058,
+      "step": 2416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.587034278680569e-06,
+      "loss": 3.0879,
+      "step": 2417
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.585633550842075e-06,
+      "loss": 2.9122,
+      "step": 2418
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.584232243416324e-06,
+      "loss": 3.0146,
+      "step": 2419
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.582830356629826e-06,
+      "loss": 2.9768,
+      "step": 2420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.581427890709184e-06,
+      "loss": 2.9823,
+      "step": 2421
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.580024845881092e-06,
+      "loss": 3.0565,
+      "step": 2422
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.578621222372343e-06,
+      "loss": 2.9045,
+      "step": 2423
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.577217020409821e-06,
+      "loss": 3.123,
+      "step": 2424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.575812240220502e-06,
+      "loss": 2.9225,
+      "step": 2425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.574406882031456e-06,
+      "loss": 2.9114,
+      "step": 2426
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.57300094606985e-06,
+      "loss": 2.987,
+      "step": 2427
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.571594432562938e-06,
+      "loss": 3.0825,
+      "step": 2428
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.570187341738074e-06,
+      "loss": 2.9838,
+      "step": 2429
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.568779673822701e-06,
+      "loss": 2.9981,
+      "step": 2430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.567371429044357e-06,
+      "loss": 2.9898,
+      "step": 2431
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.565962607630674e-06,
+      "loss": 2.9985,
+      "step": 2432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.564553209809372e-06,
+      "loss": 2.9952,
+      "step": 2433
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.563143235808274e-06,
+      "loss": 2.9876,
+      "step": 2434
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.561732685855287e-06,
+      "loss": 3.0061,
+      "step": 2435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.560321560178413e-06,
+      "loss": 3.0215,
+      "step": 2436
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.558909859005751e-06,
+      "loss": 3.177,
+      "step": 2437
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.557497582565489e-06,
+      "loss": 3.0923,
+      "step": 2438
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.556084731085911e-06,
+      "loss": 3.0433,
+      "step": 2439
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.55467130479539e-06,
+      "loss": 2.9578,
+      "step": 2440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.553257303922398e-06,
+      "loss": 2.8564,
+      "step": 2441
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.551842728695495e-06,
+      "loss": 3.0087,
+      "step": 2442
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.550427579343333e-06,
+      "loss": 3.0979,
+      "step": 2443
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.54901185609466e-06,
+      "loss": 3.1697,
+      "step": 2444
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.547595559178318e-06,
+      "loss": 3.0503,
+      "step": 2445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.546178688823237e-06,
+      "loss": 3.0694,
+      "step": 2446
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.544761245258442e-06,
+      "loss": 3.0706,
+      "step": 2447
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.54334322871305e-06,
+      "loss": 3.0372,
+      "step": 2448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.541924639416272e-06,
+      "loss": 2.8429,
+      "step": 2449
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.540505477597414e-06,
+      "loss": 3.0367,
+      "step": 2450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.539085743485868e-06,
+      "loss": 3.057,
+      "step": 2451
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.537665437311122e-06,
+      "loss": 3.0465,
+      "step": 2452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.536244559302758e-06,
+      "loss": 3.012,
+      "step": 2453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.53482310969045e-06,
+      "loss": 3.003,
+      "step": 2454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.533401088703958e-06,
+      "loss": 3.1287,
+      "step": 2455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.531978496573145e-06,
+      "loss": 3.0793,
+      "step": 2456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.530555333527959e-06,
+      "loss": 2.9988,
+      "step": 2457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.529131599798444e-06,
+      "loss": 2.9488,
+      "step": 2458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.52770729561473e-06,
+      "loss": 2.9398,
+      "step": 2459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.526282421207048e-06,
+      "loss": 3.1638,
+      "step": 2460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.524856976805715e-06,
+      "loss": 3.1272,
+      "step": 2461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.523430962641141e-06,
+      "loss": 2.9235,
+      "step": 2462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.522004378943833e-06,
+      "loss": 2.9105,
+      "step": 2463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.520577225944383e-06,
+      "loss": 2.9238,
+      "step": 2464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.519149503873478e-06,
+      "loss": 3.1413,
+      "step": 2465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.517721212961898e-06,
+      "loss": 3.0805,
+      "step": 2466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.516292353440516e-06,
+      "loss": 3.0,
+      "step": 2467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.514862925540292e-06,
+      "loss": 3.0584,
+      "step": 2468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.513432929492282e-06,
+      "loss": 3.1135,
+      "step": 2469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.512002365527632e-06,
+      "loss": 3.045,
+      "step": 2470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.510571233877582e-06,
+      "loss": 3.0165,
+      "step": 2471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.509139534773461e-06,
+      "loss": 2.9384,
+      "step": 2472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.507707268446692e-06,
+      "loss": 2.9752,
+      "step": 2473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.50627443512879e-06,
+      "loss": 3.0786,
+      "step": 2474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.504841035051356e-06,
+      "loss": 3.1079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.50340706844609e-06,
+      "loss": 2.9332,
+      "step": 2476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.501972535544781e-06,
+      "loss": 3.0652,
+      "step": 2477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.500537436579308e-06,
+      "loss": 3.112,
+      "step": 2478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.499101771781643e-06,
+      "loss": 3.019,
+      "step": 2479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.49766554138385e-06,
+      "loss": 3.0529,
+      "step": 2480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.49622874561808e-06,
+      "loss": 2.9572,
+      "step": 2481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.494791384716584e-06,
+      "loss": 3.0337,
+      "step": 2482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.493353458911694e-06,
+      "loss": 2.8817,
+      "step": 2483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.491914968435841e-06,
+      "loss": 3.0094,
+      "step": 2484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.490475913521547e-06,
+      "loss": 3.0096,
+      "step": 2485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.489036294401418e-06,
+      "loss": 2.9534,
+      "step": 2486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.487596111308161e-06,
+      "loss": 2.9455,
+      "step": 2487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.486155364474566e-06,
+      "loss": 3.1917,
+      "step": 2488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.484714054133518e-06,
+      "loss": 2.9332,
+      "step": 2489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.483272180517997e-06,
+      "loss": 3.0416,
+      "step": 2490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.481829743861061e-06,
+      "loss": 2.8449,
+      "step": 2491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.480386744395877e-06,
+      "loss": 3.1047,
+      "step": 2492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.478943182355687e-06,
+      "loss": 2.9387,
+      "step": 2493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.477499057973835e-06,
+      "loss": 2.9707,
+      "step": 2494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.476054371483749e-06,
+      "loss": 3.0071,
+      "step": 2495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.474609123118949e-06,
+      "loss": 3.0001,
+      "step": 2496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.473163313113048e-06,
+      "loss": 3.0074,
+      "step": 2497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.471716941699753e-06,
+      "loss": 3.0026,
+      "step": 2498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.470270009112854e-06,
+      "loss": 2.9142,
+      "step": 2499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.468822515586232e-06,
+      "loss": 3.0381,
+      "step": 2500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.46737446135387e-06,
+      "loss": 3.0364,
+      "step": 2501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.465925846649827e-06,
+      "loss": 3.0343,
+      "step": 2502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.464476671708263e-06,
+      "loss": 2.9347,
+      "step": 2503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.463026936763424e-06,
+      "loss": 3.0899,
+      "step": 2504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.461576642049647e-06,
+      "loss": 2.9907,
+      "step": 2505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.460125787801359e-06,
+      "loss": 3.0069,
+      "step": 2506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.45867437425308e-06,
+      "loss": 2.9765,
+      "step": 2507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.457222401639418e-06,
+      "loss": 2.9836,
+      "step": 2508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.455769870195071e-06,
+      "loss": 2.9073,
+      "step": 2509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.454316780154829e-06,
+      "loss": 2.9569,
+      "step": 2510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.452863131753575e-06,
+      "loss": 3.1154,
+      "step": 2511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.451408925226274e-06,
+      "loss": 3.0385,
+      "step": 2512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.449954160807987e-06,
+      "loss": 3.0386,
+      "step": 2513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.448498838733866e-06,
+      "loss": 2.9457,
+      "step": 2514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.447042959239152e-06,
+      "loss": 3.0912,
+      "step": 2515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.445586522559173e-06,
+      "loss": 2.9574,
+      "step": 2516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.444129528929352e-06,
+      "loss": 2.9023,
+      "step": 2517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.4426719785852e-06,
+      "loss": 2.9143,
+      "step": 2518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.441213871762313e-06,
+      "loss": 2.8136,
+      "step": 2519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.439755208696388e-06,
+      "loss": 2.9444,
+      "step": 2520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.438295989623202e-06,
+      "loss": 3.1135,
+      "step": 2521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.436836214778625e-06,
+      "loss": 3.0213,
+      "step": 2522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.435375884398619e-06,
+      "loss": 3.1655,
+      "step": 2523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.433914998719232e-06,
+      "loss": 2.9866,
+      "step": 2524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.432453557976605e-06,
+      "loss": 3.2152,
+      "step": 2525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.43099156240697e-06,
+      "loss": 2.9225,
+      "step": 2526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.429529012246642e-06,
+      "loss": 3.0231,
+      "step": 2527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.428065907732032e-06,
+      "loss": 3.1511,
+      "step": 2528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.426602249099636e-06,
+      "loss": 2.802,
+      "step": 2529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.425138036586046e-06,
+      "loss": 2.7769,
+      "step": 2530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.423673270427936e-06,
+      "loss": 2.9941,
+      "step": 2531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.422207950862076e-06,
+      "loss": 3.0052,
+      "step": 2532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.42074207812532e-06,
+      "loss": 2.9489,
+      "step": 2533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.419275652454615e-06,
+      "loss": 3.1003,
+      "step": 2534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.417808674086996e-06,
+      "loss": 2.8846,
+      "step": 2535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.416341143259587e-06,
+      "loss": 2.9658,
+      "step": 2536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.414873060209605e-06,
+      "loss": 2.8131,
+      "step": 2537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.41340442517435e-06,
+      "loss": 2.9342,
+      "step": 2538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.411935238391215e-06,
+      "loss": 2.9907,
+      "step": 2539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.41046550009768e-06,
+      "loss": 3.0785,
+      "step": 2540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.40899521053132e-06,
+      "loss": 2.999,
+      "step": 2541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.40752436992979e-06,
+      "loss": 3.2058,
+      "step": 2542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.406052978530846e-06,
+      "loss": 2.9233,
+      "step": 2543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.404581036572317e-06,
+      "loss": 3.1028,
+      "step": 2544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.403108544292136e-06,
+      "loss": 3.0027,
+      "step": 2545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.401635501928316e-06,
+      "loss": 3.1499,
+      "step": 2546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.400161909718962e-06,
+      "loss": 2.8865,
+      "step": 2547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.398687767902271e-06,
+      "loss": 3.14,
+      "step": 2548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.397213076716522e-06,
+      "loss": 3.0622,
+      "step": 2549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.395737836400086e-06,
+      "loss": 3.0506,
+      "step": 2550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.394262047191427e-06,
+      "loss": 3.1126,
+      "step": 2551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.39278570932909e-06,
+      "loss": 3.1442,
+      "step": 2552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.391308823051716e-06,
+      "loss": 3.0467,
+      "step": 2553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.389831388598025e-06,
+      "loss": 3.0775,
+      "step": 2554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.388353406206838e-06,
+      "loss": 2.9206,
+      "step": 2555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.386874876117057e-06,
+      "loss": 2.9882,
+      "step": 2556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.385395798567671e-06,
+      "loss": 2.9537,
+      "step": 2557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.383916173797762e-06,
+      "loss": 3.0593,
+      "step": 2558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.3824360020465e-06,
+      "loss": 3.1018,
+      "step": 2559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.380955283553141e-06,
+      "loss": 3.1075,
+      "step": 2560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.379474018557032e-06,
+      "loss": 3.0151,
+      "step": 2561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.377992207297605e-06,
+      "loss": 3.0005,
+      "step": 2562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.376509850014383e-06,
+      "loss": 2.8593,
+      "step": 2563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.375026946946976e-06,
+      "loss": 2.9553,
+      "step": 2564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.373543498335082e-06,
+      "loss": 2.9535,
+      "step": 2565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.37205950441849e-06,
+      "loss": 3.07,
+      "step": 2566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.370574965437075e-06,
+      "loss": 3.0681,
+      "step": 2567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.369089881630798e-06,
+      "loss": 2.9878,
+      "step": 2568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.367604253239712e-06,
+      "loss": 2.9937,
+      "step": 2569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.366118080503958e-06,
+      "loss": 3.0198,
+      "step": 2570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.364631363663758e-06,
+      "loss": 3.1452,
+      "step": 2571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.363144102959431e-06,
+      "loss": 2.9582,
+      "step": 2572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.361656298631377e-06,
+      "loss": 2.8492,
+      "step": 2573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.360167950920091e-06,
+      "loss": 3.0744,
+      "step": 2574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.35867906006615e-06,
+      "loss": 2.8796,
+      "step": 2575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.357189626310222e-06,
+      "loss": 3.2063,
+      "step": 2576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.355699649893057e-06,
+      "loss": 3.0973,
+      "step": 2577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.354209131055502e-06,
+      "loss": 2.9163,
+      "step": 2578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.352718070038481e-06,
+      "loss": 3.0801,
+      "step": 2579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.351226467083019e-06,
+      "loss": 2.9958,
+      "step": 2580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.349734322430215e-06,
+      "loss": 3.1317,
+      "step": 2581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.348241636321263e-06,
+      "loss": 2.8647,
+      "step": 2582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.346748408997445e-06,
+      "loss": 2.9346,
+      "step": 2583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.345254640700126e-06,
+      "loss": 3.0414,
+      "step": 2584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.343760331670764e-06,
+      "loss": 3.1068,
+      "step": 2585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.342265482150895e-06,
+      "loss": 3.0242,
+      "step": 2586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.340770092382156e-06,
+      "loss": 2.9197,
+      "step": 2587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.33927416260626e-06,
+      "loss": 3.1364,
+      "step": 2588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.337777693065014e-06,
+      "loss": 3.113,
+      "step": 2589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.336280684000308e-06,
+      "loss": 2.9427,
+      "step": 2590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.334783135654121e-06,
+      "loss": 3.0138,
+      "step": 2591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.33328504826852e-06,
+      "loss": 2.9925,
+      "step": 2592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.331786422085654e-06,
+      "loss": 2.9822,
+      "step": 2593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.33028725734777e-06,
+      "loss": 2.8973,
+      "step": 2594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.328787554297188e-06,
+      "loss": 3.117,
+      "step": 2595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.32728731317633e-06,
+      "loss": 2.9833,
+      "step": 2596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.325786534227689e-06,
+      "loss": 3.0409,
+      "step": 2597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.324285217693859e-06,
+      "loss": 3.053,
+      "step": 2598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.322783363817513e-06,
+      "loss": 3.0266,
+      "step": 2599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.321280972841414e-06,
+      "loss": 3.1958,
+      "step": 2600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.319778045008411e-06,
+      "loss": 3.0176,
+      "step": 2601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.318274580561436e-06,
+      "loss": 3.0614,
+      "step": 2602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.316770579743516e-06,
+      "loss": 2.9902,
+      "step": 2603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.315266042797758e-06,
+      "loss": 2.8561,
+      "step": 2604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.313760969967357e-06,
+      "loss": 2.9911,
+      "step": 2605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.312255361495597e-06,
+      "loss": 2.9724,
+      "step": 2606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.310749217625843e-06,
+      "loss": 2.9783,
+      "step": 2607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.309242538601554e-06,
+      "loss": 2.9983,
+      "step": 2608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.307735324666271e-06,
+      "loss": 3.1094,
+      "step": 2609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.306227576063623e-06,
+      "loss": 3.1457,
+      "step": 2610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.304719293037325e-06,
+      "loss": 2.9798,
+      "step": 2611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.303210475831174e-06,
+      "loss": 3.0451,
+      "step": 2612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.301701124689064e-06,
+      "loss": 3.1218,
+      "step": 2613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.300191239854963e-06,
+      "loss": 3.1624,
+      "step": 2614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.298680821572935e-06,
+      "loss": 2.9873,
+      "step": 2615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.297169870087125e-06,
+      "loss": 3.1089,
+      "step": 2616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.295658385641766e-06,
+      "loss": 2.8814,
+      "step": 2617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.294146368481174e-06,
+      "loss": 3.0767,
+      "step": 2618
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.292633818849758e-06,
+      "loss": 3.117,
+      "step": 2619
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.291120736992004e-06,
+      "loss": 3.1709,
+      "step": 2620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.289607123152494e-06,
+      "loss": 2.9259,
+      "step": 2621
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.288092977575887e-06,
+      "loss": 2.9714,
+      "step": 2622
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.286578300506933e-06,
+      "loss": 3.0142,
+      "step": 2623
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.285063092190468e-06,
+      "loss": 3.2504,
+      "step": 2624
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.28354735287141e-06,
+      "loss": 3.0605,
+      "step": 2625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.282031082794768e-06,
+      "loss": 2.8528,
+      "step": 2626
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.280514282205635e-06,
+      "loss": 2.8443,
+      "step": 2627
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.278996951349184e-06,
+      "loss": 2.9897,
+      "step": 2628
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.27747909047068e-06,
+      "loss": 3.0136,
+      "step": 2629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.275960699815477e-06,
+      "loss": 3.0975,
+      "step": 2630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.274441779629005e-06,
+      "loss": 2.8406,
+      "step": 2631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.272922330156789e-06,
+      "loss": 2.9465,
+      "step": 2632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.271402351644431e-06,
+      "loss": 2.9557,
+      "step": 2633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.269881844337624e-06,
+      "loss": 2.8952,
+      "step": 2634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.268360808482146e-06,
+      "loss": 3.0473,
+      "step": 2635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.266839244323859e-06,
+      "loss": 2.9043,
+      "step": 2636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.26531715210871e-06,
+      "loss": 2.9196,
+      "step": 2637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.263794532082734e-06,
+      "loss": 2.9818,
+      "step": 2638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.262271384492047e-06,
+      "loss": 3.101,
+      "step": 2639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.260747709582857e-06,
+      "loss": 3.0778,
+      "step": 2640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.25922350760145e-06,
+      "loss": 2.8777,
+      "step": 2641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.257698778794201e-06,
+      "loss": 2.9368,
+      "step": 2642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.256173523407569e-06,
+      "loss": 3.143,
+      "step": 2643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.254647741688099e-06,
+      "loss": 2.8823,
+      "step": 2644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.253121433882424e-06,
+      "loss": 3.0141,
+      "step": 2645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.251594600237255e-06,
+      "loss": 2.9801,
+      "step": 2646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.250067240999389e-06,
+      "loss": 3.1672,
+      "step": 2647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.248539356415715e-06,
+      "loss": 2.9594,
+      "step": 2648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.247010946733202e-06,
+      "loss": 3.0584,
+      "step": 2649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.245482012198904e-06,
+      "loss": 2.8914,
+      "step": 2650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.243952553059961e-06,
+      "loss": 2.9258,
+      "step": 2651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.242422569563596e-06,
+      "loss": 2.7234,
+      "step": 2652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.240892061957118e-06,
+      "loss": 3.0966,
+      "step": 2653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.239361030487922e-06,
+      "loss": 2.9969,
+      "step": 2654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.237829475403485e-06,
+      "loss": 3.0519,
+      "step": 2655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.23629739695137e-06,
+      "loss": 3.0149,
+      "step": 2656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.234764795379222e-06,
+      "loss": 2.9885,
+      "step": 2657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.233231670934776e-06,
+      "loss": 2.9161,
+      "step": 2658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.23169802386585e-06,
+      "loss": 3.0285,
+      "step": 2659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.230163854420343e-06,
+      "loss": 3.154,
+      "step": 2660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.228629162846242e-06,
+      "loss": 2.9455,
+      "step": 2661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.22709394939161e-06,
+      "loss": 2.8635,
+      "step": 2662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.225558214304612e-06,
+      "loss": 3.0043,
+      "step": 2663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.224021957833482e-06,
+      "loss": 3.0704,
+      "step": 2664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.222485180226538e-06,
+      "loss": 2.9119,
+      "step": 2665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.220947881732194e-06,
+      "loss": 3.0951,
+      "step": 2666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.219410062598937e-06,
+      "loss": 2.8176,
+      "step": 2667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.217871723075344e-06,
+      "loss": 3.0463,
+      "step": 2668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.216332863410076e-06,
+      "loss": 2.9014,
+      "step": 2669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.214793483851874e-06,
+      "loss": 2.869,
+      "step": 2670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.213253584649567e-06,
+      "loss": 2.9756,
+      "step": 2671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.211713166052065e-06,
+      "loss": 3.0735,
+      "step": 2672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.210172228308366e-06,
+      "loss": 2.9989,
+      "step": 2673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.208630771667549e-06,
+      "loss": 2.9899,
+      "step": 2674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.207088796378776e-06,
+      "loss": 3.151,
+      "step": 2675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.205546302691297e-06,
+      "loss": 3.1282,
+      "step": 2676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.204003290854437e-06,
+      "loss": 3.1047,
+      "step": 2677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.202459761117619e-06,
+      "loss": 2.9616,
+      "step": 2678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.200915713730335e-06,
+      "loss": 3.1434,
+      "step": 2679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.19937114894217e-06,
+      "loss": 3.0774,
+      "step": 2680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.19782606700279e-06,
+      "loss": 2.9042,
+      "step": 2681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.196280468161944e-06,
+      "loss": 3.1057,
+      "step": 2682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.194734352669463e-06,
+      "loss": 2.9845,
+      "step": 2683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.193187720775266e-06,
+      "loss": 2.8946,
+      "step": 2684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.191640572729353e-06,
+      "loss": 2.9636,
+      "step": 2685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.190092908781806e-06,
+      "loss": 3.128,
+      "step": 2686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.188544729182792e-06,
+      "loss": 3.0857,
+      "step": 2687
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.186996034182562e-06,
+      "loss": 3.0677,
+      "step": 2688
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.185446824031446e-06,
+      "loss": 3.0589,
+      "step": 2689
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.183897098979867e-06,
+      "loss": 2.9521,
+      "step": 2690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.18234685927832e-06,
+      "loss": 3.0666,
+      "step": 2691
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.18079610517739e-06,
+      "loss": 3.099,
+      "step": 2692
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.17924483692774e-06,
+      "loss": 3.0835,
+      "step": 2693
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.177693054780125e-06,
+      "loss": 2.9784,
+      "step": 2694
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.176140758985374e-06,
+      "loss": 2.9792,
+      "step": 2695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.1745879497944e-06,
+      "loss": 3.0958,
+      "step": 2696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.173034627458207e-06,
+      "loss": 2.9527,
+      "step": 2697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.171480792227874e-06,
+      "loss": 3.0538,
+      "step": 2698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.169926444354564e-06,
+      "loss": 3.0491,
+      "step": 2699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.168371584089524e-06,
+      "loss": 2.9911,
+      "step": 2700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.166816211684087e-06,
+      "loss": 3.0635,
+      "step": 2701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.165260327389664e-06,
+      "loss": 2.9321,
+      "step": 2702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.16370393145775e-06,
+      "loss": 3.0465,
+      "step": 2703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.162147024139924e-06,
+      "loss": 2.9293,
+      "step": 2704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.160589605687846e-06,
+      "loss": 2.9185,
+      "step": 2705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.15903167635326e-06,
+      "loss": 2.8688,
+      "step": 2706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.157473236387992e-06,
+      "loss": 2.9142,
+      "step": 2707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.155914286043949e-06,
+      "loss": 2.9811,
+      "step": 2708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.154354825573127e-06,
+      "loss": 2.9685,
+      "step": 2709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.152794855227594e-06,
+      "loss": 3.1147,
+      "step": 2710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.15123437525951e-06,
+      "loss": 2.9904,
+      "step": 2711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.149673385921108e-06,
+      "loss": 2.8293,
+      "step": 2712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.148111887464716e-06,
+      "loss": 3.018,
+      "step": 2713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.146549880142731e-06,
+      "loss": 3.06,
+      "step": 2714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.144987364207642e-06,
+      "loss": 2.9121,
+      "step": 2715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.143424339912014e-06,
+      "loss": 3.1107,
+      "step": 2716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.141860807508499e-06,
+      "loss": 3.0982,
+      "step": 2717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.140296767249826e-06,
+      "loss": 2.9688,
+      "step": 2718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.138732219388812e-06,
+      "loss": 2.853,
+      "step": 2719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.137167164178348e-06,
+      "loss": 3.1853,
+      "step": 2720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.135601601871417e-06,
+      "loss": 2.8322,
+      "step": 2721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.134035532721077e-06,
+      "loss": 3.115,
+      "step": 2722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.132468956980471e-06,
+      "loss": 2.9781,
+      "step": 2723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.13090187490282e-06,
+      "loss": 3.1936,
+      "step": 2724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.12933428674143e-06,
+      "loss": 2.9904,
+      "step": 2725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.127766192749688e-06,
+      "loss": 2.9441,
+      "step": 2726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.126197593181068e-06,
+      "loss": 3.098,
+      "step": 2727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.124628488289115e-06,
+      "loss": 2.9881,
+      "step": 2728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.123058878327464e-06,
+      "loss": 3.087,
+      "step": 2729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.121488763549828e-06,
+      "loss": 2.9842,
+      "step": 2730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.119918144210005e-06,
+      "loss": 2.9485,
+      "step": 2731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.11834702056187e-06,
+      "loss": 2.9649,
+      "step": 2732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.116775392859381e-06,
+      "loss": 3.0031,
+      "step": 2733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.115203261356583e-06,
+      "loss": 3.003,
+      "step": 2734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.113630626307593e-06,
+      "loss": 2.995,
+      "step": 2735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.112057487966616e-06,
+      "loss": 3.0206,
+      "step": 2736
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.110483846587935e-06,
+      "loss": 3.0794,
+      "step": 2737
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.10890970242592e-06,
+      "loss": 2.9882,
+      "step": 2738
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.107335055735012e-06,
+      "loss": 2.8951,
+      "step": 2739
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.105759906769743e-06,
+      "loss": 2.963,
+      "step": 2740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.104184255784723e-06,
+      "loss": 2.9967,
+      "step": 2741
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.10260810303464e-06,
+      "loss": 3.0267,
+      "step": 2742
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.10103144877427e-06,
+      "loss": 3.0502,
+      "step": 2743
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.099454293258461e-06,
+      "loss": 3.0675,
+      "step": 2744
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.097876636742148e-06,
+      "loss": 2.913,
+      "step": 2745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.096298479480349e-06,
+      "loss": 2.9629,
+      "step": 2746
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.094719821728154e-06,
+      "loss": 2.9813,
+      "step": 2747
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.093140663740747e-06,
+      "loss": 3.1496,
+      "step": 2748
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.091561005773379e-06,
+      "loss": 2.9228,
+      "step": 2749
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.089980848081388e-06,
+      "loss": 3.0728,
+      "step": 2750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.0884001909202e-06,
+      "loss": 2.8245,
+      "step": 2751
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.086819034545311e-06,
+      "loss": 3.1721,
+      "step": 2752
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.085237379212298e-06,
+      "loss": 3.0292,
+      "step": 2753
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.083655225176828e-06,
+      "loss": 3.1328,
+      "step": 2754
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.082072572694639e-06,
+      "loss": 2.9085,
+      "step": 2755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.080489422021555e-06,
+      "loss": 2.9558,
+      "step": 2756
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.078905773413479e-06,
+      "loss": 2.7987,
+      "step": 2757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.077321627126393e-06,
+      "loss": 2.925,
+      "step": 2758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.07573698341636e-06,
+      "loss": 3.0272,
+      "step": 2759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.074151842539528e-06,
+      "loss": 3.2511,
+      "step": 2760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.072566204752119e-06,
+      "loss": 2.987,
+      "step": 2761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.070980070310438e-06,
+      "loss": 2.9073,
+      "step": 2762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.06939343947087e-06,
+      "loss": 3.1118,
+      "step": 2763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.067806312489882e-06,
+      "loss": 3.0267,
+      "step": 2764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.066218689624017e-06,
+      "loss": 3.0357,
+      "step": 2765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.064630571129903e-06,
+      "loss": 2.9156,
+      "step": 2766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.063041957264246e-06,
+      "loss": 3.0993,
+      "step": 2767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.061452848283831e-06,
+      "loss": 2.9666,
+      "step": 2768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.059863244445522e-06,
+      "loss": 2.9631,
+      "step": 2769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.058273146006269e-06,
+      "loss": 2.8976,
+      "step": 2770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.056682553223096e-06,
+      "loss": 2.9681,
+      "step": 2771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.055091466353109e-06,
+      "loss": 2.8722,
+      "step": 2772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.053499885653493e-06,
+      "loss": 3.0002,
+      "step": 2773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.051907811381514e-06,
+      "loss": 3.0358,
+      "step": 2774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.050315243794518e-06,
+      "loss": 2.8022,
+      "step": 2775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.048722183149928e-06,
+      "loss": 3.0098,
+      "step": 2776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.047128629705251e-06,
+      "loss": 3.1837,
+      "step": 2777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.04553458371807e-06,
+      "loss": 3.1186,
+      "step": 2778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.043940045446052e-06,
+      "loss": 3.0949,
+      "step": 2779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.042345015146935e-06,
+      "loss": 2.8553,
+      "step": 2780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.040749493078548e-06,
+      "loss": 3.1319,
+      "step": 2781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.03915347949879e-06,
+      "loss": 2.982,
+      "step": 2782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.037556974665644e-06,
+      "loss": 2.9004,
+      "step": 2783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.03595997883717e-06,
+      "loss": 3.0472,
+      "step": 2784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.03436249227151e-06,
+      "loss": 3.1268,
+      "step": 2785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.032764515226888e-06,
+      "loss": 2.9213,
+      "step": 2786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.031166047961596e-06,
+      "loss": 3.012,
+      "step": 2787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.029567090734018e-06,
+      "loss": 2.9791,
+      "step": 2788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.02796764380261e-06,
+      "loss": 3.1377,
+      "step": 2789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.026367707425912e-06,
+      "loss": 2.9931,
+      "step": 2790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.024767281862535e-06,
+      "loss": 3.1587,
+      "step": 2791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.023166367371178e-06,
+      "loss": 3.1446,
+      "step": 2792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.021564964210614e-06,
+      "loss": 2.7836,
+      "step": 2793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.019963072639696e-06,
+      "loss": 3.071,
+      "step": 2794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.018360692917356e-06,
+      "loss": 2.9259,
+      "step": 2795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.016757825302605e-06,
+      "loss": 2.8811,
+      "step": 2796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.015154470054536e-06,
+      "loss": 3.1415,
+      "step": 2797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.013550627432314e-06,
+      "loss": 2.8679,
+      "step": 2798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.011946297695187e-06,
+      "loss": 3.106,
+      "step": 2799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.010341481102484e-06,
+      "loss": 2.8202,
+      "step": 2800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.008736177913608e-06,
+      "loss": 3.0496,
+      "step": 2801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.007130388388042e-06,
+      "loss": 2.9023,
+      "step": 2802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.00552411278535e-06,
+      "loss": 3.0746,
+      "step": 2803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.00391735136517e-06,
+      "loss": 2.9251,
+      "step": 2804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.002310104387224e-06,
+      "loss": 2.8309,
+      "step": 2805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.000702372111311e-06,
+      "loss": 2.9323,
+      "step": 2806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.999094154797304e-06,
+      "loss": 2.8148,
+      "step": 2807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.997485452705158e-06,
+      "loss": 2.8763,
+      "step": 2808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.99587626609491e-06,
+      "loss": 2.7754,
+      "step": 2809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.994266595226665e-06,
+      "loss": 2.8273,
+      "step": 2810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.992656440360618e-06,
+      "loss": 2.9641,
+      "step": 2811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.991045801757036e-06,
+      "loss": 3.1035,
+      "step": 2812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.989434679676264e-06,
+      "loss": 3.0812,
+      "step": 2813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.987823074378725e-06,
+      "loss": 3.0792,
+      "step": 2814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.986210986124923e-06,
+      "loss": 2.9881,
+      "step": 2815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.984598415175438e-06,
+      "loss": 3.0011,
+      "step": 2816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.98298536179093e-06,
+      "loss": 3.1199,
+      "step": 2817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.981371826232132e-06,
+      "loss": 2.9445,
+      "step": 2818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.97975780875986e-06,
+      "loss": 2.8507,
+      "step": 2819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.978143309635008e-06,
+      "loss": 3.1144,
+      "step": 2820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.976528329118545e-06,
+      "loss": 2.9869,
+      "step": 2821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.974912867471519e-06,
+      "loss": 2.9752,
+      "step": 2822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.973296924955051e-06,
+      "loss": 3.0029,
+      "step": 2823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.971680501830352e-06,
+      "loss": 2.9587,
+      "step": 2824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.9700635983587e-06,
+      "loss": 2.9974,
+      "step": 2825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.96844621480145e-06,
+      "loss": 2.984,
+      "step": 2826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.966828351420046e-06,
+      "loss": 3.039,
+      "step": 2827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.965210008475995e-06,
+      "loss": 2.8939,
+      "step": 2828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.963591186230893e-06,
+      "loss": 3.0163,
+      "step": 2829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.961971884946405e-06,
+      "loss": 3.0101,
+      "step": 2830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.96035210488428e-06,
+      "loss": 2.9863,
+      "step": 2831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.95873184630634e-06,
+      "loss": 2.7845,
+      "step": 2832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.957111109474488e-06,
+      "loss": 2.9913,
+      "step": 2833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.9554898946507e-06,
+      "loss": 3.2348,
+      "step": 2834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.953868202097034e-06,
+      "loss": 3.0568,
+      "step": 2835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.952246032075624e-06,
+      "loss": 2.9461,
+      "step": 2836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.950623384848675e-06,
+      "loss": 3.12,
+      "step": 2837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.949000260678477e-06,
+      "loss": 3.1453,
+      "step": 2838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.947376659827396e-06,
+      "loss": 2.9979,
+      "step": 2839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.945752582557872e-06,
+      "loss": 3.0724,
+      "step": 2840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.944128029132422e-06,
+      "loss": 3.0487,
+      "step": 2841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.942502999813641e-06,
+      "loss": 3.0425,
+      "step": 2842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.940877494864203e-06,
+      "loss": 2.8905,
+      "step": 2843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.939251514546857e-06,
+      "loss": 2.9874,
+      "step": 2844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.93762505912443e-06,
+      "loss": 2.894,
+      "step": 2845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.935998128859821e-06,
+      "loss": 3.0146,
+      "step": 2846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.934370724016012e-06,
+      "loss": 2.9836,
+      "step": 2847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.932742844856058e-06,
+      "loss": 2.8366,
+      "step": 2848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.931114491643096e-06,
+      "loss": 3.0507,
+      "step": 2849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.929485664640328e-06,
+      "loss": 2.9318,
+      "step": 2850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.927856364111046e-06,
+      "loss": 2.9722,
+      "step": 2851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.926226590318611e-06,
+      "loss": 3.1486,
+      "step": 2852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.924596343526463e-06,
+      "loss": 2.9089,
+      "step": 2853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.922965623998116e-06,
+      "loss": 2.9717,
+      "step": 2854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.921334431997162e-06,
+      "loss": 3.0805,
+      "step": 2855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.91970276778727e-06,
+      "loss": 2.9955,
+      "step": 2856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.918070631632185e-06,
+      "loss": 2.9836,
+      "step": 2857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.916438023795727e-06,
+      "loss": 3.0143,
+      "step": 2858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.914804944541795e-06,
+      "loss": 3.024,
+      "step": 2859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.91317139413436e-06,
+      "loss": 3.0386,
+      "step": 2860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.911537372837473e-06,
+      "loss": 2.9137,
+      "step": 2861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.909902880915259e-06,
+      "loss": 3.0327,
+      "step": 2862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.90826791863192e-06,
+      "loss": 2.9191,
+      "step": 2863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.906632486251736e-06,
+      "loss": 2.9584,
+      "step": 2864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.904996584039057e-06,
+      "loss": 2.9769,
+      "step": 2865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.903360212258315e-06,
+      "loss": 2.9847,
+      "step": 2866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.901723371174016e-06,
+      "loss": 3.0409,
+      "step": 2867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.900086061050737e-06,
+      "loss": 3.0758,
+      "step": 2868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.898448282153143e-06,
+      "loss": 3.0142,
+      "step": 2869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.896810034745962e-06,
+      "loss": 3.0216,
+      "step": 2870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.895171319094002e-06,
+      "loss": 3.0538,
+      "step": 2871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.893532135462151e-06,
+      "loss": 3.0342,
+      "step": 2872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.891892484115367e-06,
+      "loss": 3.0007,
+      "step": 2873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.890252365318686e-06,
+      "loss": 2.9885,
+      "step": 2874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.888611779337218e-06,
+      "loss": 3.114,
+      "step": 2875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.886970726436154e-06,
+      "loss": 2.9393,
+      "step": 2876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.885329206880752e-06,
+      "loss": 2.9859,
+      "step": 2877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.883687220936352e-06,
+      "loss": 3.1364,
+      "step": 2878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.882044768868364e-06,
+      "loss": 2.9196,
+      "step": 2879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.880401850942281e-06,
+      "loss": 2.9302,
+      "step": 2880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.878758467423663e-06,
+      "loss": 2.9031,
+      "step": 2881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.877114618578152e-06,
+      "loss": 2.8985,
+      "step": 2882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.87547030467146e-06,
+      "loss": 3.008,
+      "step": 2883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.873825525969377e-06,
+      "loss": 3.0415,
+      "step": 2884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.872180282737769e-06,
+      "loss": 3.0363,
+      "step": 2885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.870534575242575e-06,
+      "loss": 2.967,
+      "step": 2886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.868888403749805e-06,
+      "loss": 3.0782,
+      "step": 2887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.867241768525555e-06,
+      "loss": 3.1218,
+      "step": 2888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.865594669835987e-06,
+      "loss": 2.8582,
+      "step": 2889
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.863947107947342e-06,
+      "loss": 3.0005,
+      "step": 2890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.862299083125933e-06,
+      "loss": 3.1387,
+      "step": 2891
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.860650595638147e-06,
+      "loss": 3.1621,
+      "step": 2892
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.859001645750453e-06,
+      "loss": 3.105,
+      "step": 2893
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.857352233729386e-06,
+      "loss": 3.1931,
+      "step": 2894
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.855702359841562e-06,
+      "loss": 2.9738,
+      "step": 2895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.854052024353666e-06,
+      "loss": 2.9773,
+      "step": 2896
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.852401227532463e-06,
+      "loss": 2.8392,
+      "step": 2897
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.850749969644792e-06,
+      "loss": 3.0778,
+      "step": 2898
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.849098250957558e-06,
+      "loss": 2.9759,
+      "step": 2899
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.847446071737756e-06,
+      "loss": 3.0711,
+      "step": 2900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.845793432252442e-06,
+      "loss": 3.0596,
+      "step": 2901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.844140332768751e-06,
+      "loss": 3.0482,
+      "step": 2902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.842486773553894e-06,
+      "loss": 2.9828,
+      "step": 2903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.840832754875153e-06,
+      "loss": 3.0373,
+      "step": 2904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.83917827699989e-06,
+      "loss": 3.0883,
+      "step": 2905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.837523340195534e-06,
+      "loss": 2.9638,
+      "step": 2906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.835867944729592e-06,
+      "loss": 2.9219,
+      "step": 2907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.834212090869644e-06,
+      "loss": 2.9286,
+      "step": 2908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.832555778883347e-06,
+      "loss": 3.0203,
+      "step": 2909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.830899009038426e-06,
+      "loss": 3.0588,
+      "step": 2910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.829241781602688e-06,
+      "loss": 2.8962,
+      "step": 2911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.827584096844007e-06,
+      "loss": 2.9389,
+      "step": 2912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.825925955030333e-06,
+      "loss": 3.0489,
+      "step": 2913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.824267356429696e-06,
+      "loss": 2.8897,
+      "step": 2914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.822608301310187e-06,
+      "loss": 2.985,
+      "step": 2915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.820948789939984e-06,
+      "loss": 3.1367,
+      "step": 2916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.81928882258733e-06,
+      "loss": 2.9419,
+      "step": 2917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.817628399520545e-06,
+      "loss": 3.0026,
+      "step": 2918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.815967521008024e-06,
+      "loss": 2.9592,
+      "step": 2919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.81430618731823e-06,
+      "loss": 2.9508,
+      "step": 2920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.812644398719709e-06,
+      "loss": 2.7699,
+      "step": 2921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.81098215548107e-06,
+      "loss": 2.9176,
+      "step": 2922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.809319457871004e-06,
+      "loss": 3.1258,
+      "step": 2923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.80765630615827e-06,
+      "loss": 2.8757,
+      "step": 2924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.805992700611702e-06,
+      "loss": 3.0072,
+      "step": 2925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.80432864150021e-06,
+      "loss": 2.9052,
+      "step": 2926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.802664129092772e-06,
+      "loss": 3.1088,
+      "step": 2927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.800999163658445e-06,
+      "loss": 3.1279,
+      "step": 2928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.799333745466357e-06,
+      "loss": 3.0113,
+      "step": 2929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.797667874785705e-06,
+      "loss": 3.0364,
+      "step": 2930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.796001551885768e-06,
+      "loss": 2.9319,
+      "step": 2931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.794334777035888e-06,
+      "loss": 3.0561,
+      "step": 2932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.792667550505487e-06,
+      "loss": 2.9074,
+      "step": 2933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.79099987256406e-06,
+      "loss": 3.0125,
+      "step": 2934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.78933174348117e-06,
+      "loss": 3.096,
+      "step": 2935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.787663163526455e-06,
+      "loss": 3.1214,
+      "step": 2936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.785994132969632e-06,
+      "loss": 3.011,
+      "step": 2937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.784324652080481e-06,
+      "loss": 2.9529,
+      "step": 2938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.782654721128864e-06,
+      "loss": 2.9538,
+      "step": 2939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.780984340384705e-06,
+      "loss": 2.9207,
+      "step": 2940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.77931351011801e-06,
+      "loss": 3.1725,
+      "step": 2941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.777642230598856e-06,
+      "loss": 3.1998,
+      "step": 2942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.77597050209739e-06,
+      "loss": 2.9865,
+      "step": 2943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.774298324883833e-06,
+      "loss": 2.8619,
+      "step": 2944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.772625699228475e-06,
+      "loss": 2.8675,
+      "step": 2945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.770952625401687e-06,
+      "loss": 2.9955,
+      "step": 2946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.769279103673903e-06,
+      "loss": 3.022,
+      "step": 2947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.767605134315636e-06,
+      "loss": 3.059,
+      "step": 2948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.765930717597468e-06,
+      "loss": 2.8429,
+      "step": 2949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.764255853790056e-06,
+      "loss": 3.0012,
+      "step": 2950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.762580543164124e-06,
+      "loss": 2.9986,
+      "step": 2951
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.760904785990476e-06,
+      "loss": 3.0159,
+      "step": 2952
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.75922858253998e-06,
+      "loss": 2.9798,
+      "step": 2953
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.757551933083581e-06,
+      "loss": 2.9985,
+      "step": 2954
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.755874837892298e-06,
+      "loss": 3.1392,
+      "step": 2955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.754197297237217e-06,
+      "loss": 3.1129,
+      "step": 2956
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.752519311389497e-06,
+      "loss": 2.9691,
+      "step": 2957
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.750840880620372e-06,
+      "loss": 3.0888,
+      "step": 2958
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.749162005201148e-06,
+      "loss": 2.8933,
+      "step": 2959
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.747482685403197e-06,
+      "loss": 3.1127,
+      "step": 2960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.745802921497967e-06,
+      "loss": 2.9957,
+      "step": 2961
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.74412271375698e-06,
+      "loss": 3.0891,
+      "step": 2962
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.742442062451828e-06,
+      "loss": 2.94,
+      "step": 2963
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.74076096785417e-06,
+      "loss": 3.1778,
+      "step": 2964
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.739079430235745e-06,
+      "loss": 2.8988,
+      "step": 2965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.737397449868356e-06,
+      "loss": 2.9232,
+      "step": 2966
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.735715027023882e-06,
+      "loss": 2.9972,
+      "step": 2967
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.734032161974273e-06,
+      "loss": 2.9025,
+      "step": 2968
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.732348854991552e-06,
+      "loss": 3.0729,
+      "step": 2969
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.730665106347808e-06,
+      "loss": 2.9499,
+      "step": 2970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.728980916315203e-06,
+      "loss": 3.0764,
+      "step": 2971
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.727296285165977e-06,
+      "loss": 2.989,
+      "step": 2972
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.725611213172432e-06,
+      "loss": 2.9492,
+      "step": 2973
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.72392570060695e-06,
+      "loss": 2.8672,
+      "step": 2974
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.722239747741975e-06,
+      "loss": 2.9537,
+      "step": 2975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.720553354850032e-06,
+      "loss": 2.9481,
+      "step": 2976
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.718866522203707e-06,
+      "loss": 3.0604,
+      "step": 2977
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.717179250075665e-06,
+      "loss": 2.9825,
+      "step": 2978
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.71549153873864e-06,
+      "loss": 3.0178,
+      "step": 2979
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.713803388465432e-06,
+      "loss": 3.1175,
+      "step": 2980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.712114799528922e-06,
+      "loss": 3.0666,
+      "step": 2981
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.710425772202052e-06,
+      "loss": 3.054,
+      "step": 2982
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.708736306757842e-06,
+      "loss": 2.9275,
+      "step": 2983
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.707046403469377e-06,
+      "loss": 2.891,
+      "step": 2984
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.705356062609816e-06,
+      "loss": 3.008,
+      "step": 2985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.70366528445239e-06,
+      "loss": 2.987,
+      "step": 2986
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.701974069270397e-06,
+      "loss": 3.0063,
+      "step": 2987
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.700282417337208e-06,
+      "loss": 3.0288,
+      "step": 2988
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.698590328926265e-06,
+      "loss": 2.8987,
+      "step": 2989
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.69689780431108e-06,
+      "loss": 2.9503,
+      "step": 2990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.695204843765234e-06,
+      "loss": 3.2518,
+      "step": 2991
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.693511447562379e-06,
+      "loss": 3.0369,
+      "step": 2992
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.691817615976242e-06,
+      "loss": 3.0544,
+      "step": 2993
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.690123349280611e-06,
+      "loss": 2.8186,
+      "step": 2994
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.688428647749355e-06,
+      "loss": 2.865,
+      "step": 2995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.686733511656405e-06,
+      "loss": 2.9588,
+      "step": 2996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.685037941275767e-06,
+      "loss": 2.9593,
+      "step": 2997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.683341936881515e-06,
+      "loss": 2.8968,
+      "step": 2998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.681645498747791e-06,
+      "loss": 2.8578,
+      "step": 2999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.679948627148814e-06,
+      "loss": 2.9213,
+      "step": 3000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.678251322358868e-06,
+      "loss": 3.0232,
+      "step": 3001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.676553584652308e-06,
+      "loss": 2.9846,
+      "step": 3002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.674855414303557e-06,
+      "loss": 3.0231,
+      "step": 3003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.67315681158711e-06,
+      "loss": 2.9776,
+      "step": 3004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.671457776777533e-06,
+      "loss": 3.0416,
+      "step": 3005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.669758310149463e-06,
+      "loss": 3.0428,
+      "step": 3006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.668058411977598e-06,
+      "loss": 2.978,
+      "step": 3007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.666358082536718e-06,
+      "loss": 2.9446,
+      "step": 3008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.664657322101665e-06,
+      "loss": 2.9835,
+      "step": 3009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.66295613094735e-06,
+      "loss": 2.8811,
+      "step": 3010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.66125450934876e-06,
+      "loss": 2.9892,
+      "step": 3011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.659552457580943e-06,
+      "loss": 2.9212,
+      "step": 3012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.657849975919028e-06,
+      "loss": 3.0239,
+      "step": 3013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.656147064638202e-06,
+      "loss": 3.0891,
+      "step": 3014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.654443724013724e-06,
+      "loss": 2.9076,
+      "step": 3015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.652739954320931e-06,
+      "loss": 2.8103,
+      "step": 3016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.651035755835218e-06,
+      "loss": 3.1793,
+      "step": 3017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.649331128832056e-06,
+      "loss": 3.0,
+      "step": 3018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.647626073586984e-06,
+      "loss": 2.9857,
+      "step": 3019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.645920590375609e-06,
+      "loss": 2.9816,
+      "step": 3020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.644214679473607e-06,
+      "loss": 3.015,
+      "step": 3021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.642508341156726e-06,
+      "loss": 3.0005,
+      "step": 3022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.640801575700778e-06,
+      "loss": 3.1435,
+      "step": 3023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.639094383381652e-06,
+      "loss": 2.8202,
+      "step": 3024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.637386764475296e-06,
+      "loss": 2.9661,
+      "step": 3025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.635678719257736e-06,
+      "loss": 2.9875,
+      "step": 3026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.63397024800506e-06,
+      "loss": 3.0012,
+      "step": 3027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.632261350993431e-06,
+      "loss": 2.8814,
+      "step": 3028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.630552028499077e-06,
+      "loss": 3.0347,
+      "step": 3029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.628842280798292e-06,
+      "loss": 3.0233,
+      "step": 3030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.627132108167445e-06,
+      "loss": 3.052,
+      "step": 3031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.625421510882974e-06,
+      "loss": 2.9752,
+      "step": 3032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.623710489221376e-06,
+      "loss": 3.0053,
+      "step": 3033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.6219990434592274e-06,
+      "loss": 2.9846,
+      "step": 3034
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.620287173873168e-06,
+      "loss": 2.9264,
+      "step": 3035
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.618574880739906e-06,
+      "loss": 3.0149,
+      "step": 3036
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.616862164336221e-06,
+      "loss": 3.0435,
+      "step": 3037
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.615149024938958e-06,
+      "loss": 3.0039,
+      "step": 3038
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.61343546282503e-06,
+      "loss": 3.0038,
+      "step": 3039
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.611721478271423e-06,
+      "loss": 3.0549,
+      "step": 3040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.610007071555188e-06,
+      "loss": 2.9611,
+      "step": 3041
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.608292242953439e-06,
+      "loss": 2.9346,
+      "step": 3042
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.6065769927433685e-06,
+      "loss": 2.9025,
+      "step": 3043
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.604861321202231e-06,
+      "loss": 2.8673,
+      "step": 3044
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.603145228607352e-06,
+      "loss": 2.9948,
+      "step": 3045
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.601428715236119e-06,
+      "loss": 2.9758,
+      "step": 3046
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.599711781365995e-06,
+      "loss": 2.9743,
+      "step": 3047
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.597994427274507e-06,
+      "loss": 2.9741,
+      "step": 3048
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.59627665323925e-06,
+      "loss": 2.9852,
+      "step": 3049
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.594558459537888e-06,
+      "loss": 2.9716,
+      "step": 3050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.592839846448154e-06,
+      "loss": 2.9805,
+      "step": 3051
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.591120814247847e-06,
+      "loss": 3.0157,
+      "step": 3052
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.589401363214833e-06,
+      "loss": 2.9893,
+      "step": 3053
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.587681493627044e-06,
+      "loss": 3.0256,
+      "step": 3054
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.585961205762487e-06,
+      "loss": 2.9892,
+      "step": 3055
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5842404998992295e-06,
+      "loss": 2.9449,
+      "step": 3056
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5825193763154095e-06,
+      "loss": 2.8741,
+      "step": 3057
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.580797835289233e-06,
+      "loss": 2.9342,
+      "step": 3058
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5790758770989695e-06,
+      "loss": 2.9443,
+      "step": 3059
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.577353502022962e-06,
+      "loss": 3.155,
+      "step": 3060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.575630710339617e-06,
+      "loss": 3.0528,
+      "step": 3061
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.573907502327409e-06,
+      "loss": 3.1563,
+      "step": 3062
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.572183878264879e-06,
+      "loss": 3.0654,
+      "step": 3063
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5704598384306375e-06,
+      "loss": 3.0798,
+      "step": 3064
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.56873538310336e-06,
+      "loss": 2.899,
+      "step": 3065
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5670105125617935e-06,
+      "loss": 3.0274,
+      "step": 3066
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.565285227084742e-06,
+      "loss": 3.063,
+      "step": 3067
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.56355952695109e-06,
+      "loss": 2.9524,
+      "step": 3068
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5618334124397785e-06,
+      "loss": 3.0626,
+      "step": 3069
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.56010688382982e-06,
+      "loss": 2.9798,
+      "step": 3070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.558379941400294e-06,
+      "loss": 3.0371,
+      "step": 3071
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.556652585430344e-06,
+      "loss": 2.8878,
+      "step": 3072
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.554924816199185e-06,
+      "loss": 2.9592,
+      "step": 3073
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.553196633986095e-06,
+      "loss": 3.0153,
+      "step": 3074
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.551468039070418e-06,
+      "loss": 3.1065,
+      "step": 3075
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.549739031731572e-06,
+      "loss": 2.8932,
+      "step": 3076
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.548009612249032e-06,
+      "loss": 2.9946,
+      "step": 3077
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.546279780902346e-06,
+      "loss": 3.0209,
+      "step": 3078
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.544549537971124e-06,
+      "loss": 3.0828,
+      "step": 3079
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.542818883735049e-06,
+      "loss": 2.9662,
+      "step": 3080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.541087818473863e-06,
+      "loss": 2.9926,
+      "step": 3081
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.539356342467381e-06,
+      "loss": 3.1356,
+      "step": 3082
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.537624455995478e-06,
+      "loss": 3.0176,
+      "step": 3083
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.535892159338102e-06,
+      "loss": 2.9476,
+      "step": 3084
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.534159452775263e-06,
+      "loss": 3.1203,
+      "step": 3085
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.532426336587038e-06,
+      "loss": 2.8986,
+      "step": 3086
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.530692811053568e-06,
+      "loss": 3.0719,
+      "step": 3087
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.528958876455067e-06,
+      "loss": 3.0186,
+      "step": 3088
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.527224533071811e-06,
+      "loss": 3.1004,
+      "step": 3089
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.525489781184138e-06,
+      "loss": 2.9922,
+      "step": 3090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.523754621072459e-06,
+      "loss": 3.0342,
+      "step": 3091
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.522019053017246e-06,
+      "loss": 3.1053,
+      "step": 3092
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.52028307729904e-06,
+      "loss": 2.792,
+      "step": 3093
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.518546694198446e-06,
+      "loss": 3.0727,
+      "step": 3094
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.516809903996136e-06,
+      "loss": 2.8643,
+      "step": 3095
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.515072706972849e-06,
+      "loss": 3.0548,
+      "step": 3096
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.513335103409386e-06,
+      "loss": 3.1224,
+      "step": 3097
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.511597093586616e-06,
+      "loss": 2.9517,
+      "step": 3098
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.509858677785475e-06,
+      "loss": 3.0553,
+      "step": 3099
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5081198562869606e-06,
+      "loss": 3.0907,
+      "step": 3100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.506380629372141e-06,
+      "loss": 2.9015,
+      "step": 3101
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.504640997322148e-06,
+      "loss": 3.1366,
+      "step": 3102
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.502900960418176e-06,
+      "loss": 2.7287,
+      "step": 3103
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.5011605189414886e-06,
+      "loss": 2.9894,
+      "step": 3104
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.4994196731734125e-06,
+      "loss": 3.0017,
+      "step": 3105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.49767842339534e-06,
+      "loss": 2.9951,
+      "step": 3106
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.4959367698887345e-06,
+      "loss": 3.0535,
+      "step": 3107
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.494194712935112e-06,
+      "loss": 2.9355,
+      "step": 3108
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.492452252816066e-06,
+      "loss": 3.008,
+      "step": 3109
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.49070938981325e-06,
+      "loss": 2.9901,
+      "step": 3110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.488966124208382e-06,
+      "loss": 3.104,
+      "step": 3111
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.4872224562832455e-06,
+      "loss": 3.07,
+      "step": 3112
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.485478386319691e-06,
+      "loss": 3.1126,
+      "step": 3113
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.4837339145996324e-06,
+      "loss": 2.7945,
+      "step": 3114
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.481989041405049e-06,
+      "loss": 3.0566,
+      "step": 3115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.4802437670179825e-06,
+      "loss": 3.0014,
+      "step": 3116
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.478498091720545e-06,
+      "loss": 2.9497,
+      "step": 3117
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4767520157949085e-06,
+      "loss": 2.8483,
+      "step": 3118
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.47500553952331e-06,
+      "loss": 2.907,
+      "step": 3119
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.473258663188056e-06,
+      "loss": 2.9517,
+      "step": 3120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.471511387071512e-06,
+      "loss": 3.0466,
+      "step": 3121
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.469763711456108e-06,
+      "loss": 3.0729,
+      "step": 3122
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.468015636624347e-06,
+      "loss": 3.0132,
+      "step": 3123
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.466267162858786e-06,
+      "loss": 2.9397,
+      "step": 3124
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.464518290442051e-06,
+      "loss": 3.0228,
+      "step": 3125
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4627690196568346e-06,
+      "loss": 3.0773,
+      "step": 3126
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4610193507858885e-06,
+      "loss": 2.9655,
+      "step": 3127
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.459269284112036e-06,
+      "loss": 3.0279,
+      "step": 3128
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4575188199181565e-06,
+      "loss": 2.9372,
+      "step": 3129
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.455767958487201e-06,
+      "loss": 3.0943,
+      "step": 3130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.454016700102179e-06,
+      "loss": 3.0187,
+      "step": 3131
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4522650450461675e-06,
+      "loss": 3.0479,
+      "step": 3132
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.450512993602307e-06,
+      "loss": 2.9971,
+      "step": 3133
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4487605460538005e-06,
+      "loss": 3.0396,
+      "step": 3134
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.447007702683918e-06,
+      "loss": 2.9795,
+      "step": 3135
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.445254463775992e-06,
+      "loss": 3.0784,
+      "step": 3136
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.443500829613415e-06,
+      "loss": 3.0331,
+      "step": 3137
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.44174680047965e-06,
+      "loss": 2.9566,
+      "step": 3138
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.439992376658223e-06,
+      "loss": 3.1184,
+      "step": 3139
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.438237558432717e-06,
+      "loss": 2.9482,
+      "step": 3140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.436482346086788e-06,
+      "loss": 2.8735,
+      "step": 3141
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4347267399041465e-06,
+      "loss": 3.0699,
+      "step": 3142
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.432970740168576e-06,
+      "loss": 2.8215,
+      "step": 3143
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.431214347163917e-06,
+      "loss": 2.9978,
+      "step": 3144
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.429457561174074e-06,
+      "loss": 2.96,
+      "step": 3145
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.427700382483019e-06,
+      "loss": 2.9118,
+      "step": 3146
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.425942811374782e-06,
+      "loss": 2.9512,
+      "step": 3147
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.424184848133464e-06,
+      "loss": 2.9971,
+      "step": 3148
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.422426493043221e-06,
+      "loss": 2.949,
+      "step": 3149
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4206677463882795e-06,
+      "loss": 3.0846,
+      "step": 3150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.4189086084529205e-06,
+      "loss": 2.9856,
+      "step": 3151
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.417149079521498e-06,
+      "loss": 3.0456,
+      "step": 3152
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.415389159878427e-06,
+      "loss": 2.8337,
+      "step": 3153
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.413628849808179e-06,
+      "loss": 3.1046,
+      "step": 3154
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.411868149595295e-06,
+      "loss": 2.9767,
+      "step": 3155
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.410107059524378e-06,
+      "loss": 2.9063,
+      "step": 3156
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.408345579880091e-06,
+      "loss": 2.9456,
+      "step": 3157
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.406583710947163e-06,
+      "loss": 3.0575,
+      "step": 3158
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.404821453010388e-06,
+      "loss": 2.9729,
+      "step": 3159
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.403058806354617e-06,
+      "loss": 3.1625,
+      "step": 3160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.401295771264767e-06,
+      "loss": 3.0306,
+      "step": 3161
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.399532348025819e-06,
+      "loss": 2.9952,
+      "step": 3162
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.397768536922817e-06,
+      "loss": 3.0477,
+      "step": 3163
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.396004338240862e-06,
+      "loss": 3.0147,
+      "step": 3164
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.394239752265124e-06,
+      "loss": 2.9542,
+      "step": 3165
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.392474779280833e-06,
+      "loss": 3.0509,
+      "step": 3166
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3907094195732835e-06,
+      "loss": 3.0556,
+      "step": 3167
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.388943673427829e-06,
+      "loss": 2.8811,
+      "step": 3168
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3871775411298865e-06,
+      "loss": 3.0172,
+      "step": 3169
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.385411022964938e-06,
+      "loss": 2.8854,
+      "step": 3170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.383644119218525e-06,
+      "loss": 2.954,
+      "step": 3171
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.381876830176255e-06,
+      "loss": 3.0585,
+      "step": 3172
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3801091561237935e-06,
+      "loss": 3.0299,
+      "step": 3173
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.378341097346871e-06,
+      "loss": 2.9628,
+      "step": 3174
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.376572654131277e-06,
+      "loss": 3.0058,
+      "step": 3175
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.374803826762866e-06,
+      "loss": 3.1988,
+      "step": 3176
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.373034615527555e-06,
+      "loss": 3.1184,
+      "step": 3177
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.371265020711324e-06,
+      "loss": 2.9963,
+      "step": 3178
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.369495042600209e-06,
+      "loss": 2.9636,
+      "step": 3179
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.367724681480314e-06,
+      "loss": 3.0327,
+      "step": 3180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.365953937637802e-06,
+      "loss": 3.0085,
+      "step": 3181
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3641828113589e-06,
+      "loss": 3.012,
+      "step": 3182
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.362411302929893e-06,
+      "loss": 2.8809,
+      "step": 3183
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.360639412637134e-06,
+      "loss": 2.9873,
+      "step": 3184
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.358867140767031e-06,
+      "loss": 2.9505,
+      "step": 3185
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.357094487606058e-06,
+      "loss": 3.0678,
+      "step": 3186
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.355321453440749e-06,
+      "loss": 3.0428,
+      "step": 3187
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.353548038557701e-06,
+      "loss": 3.025,
+      "step": 3188
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.351774243243569e-06,
+      "loss": 2.9901,
+      "step": 3189
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3500000677850745e-06,
+      "loss": 2.9493,
+      "step": 3190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3482255124689975e-06,
+      "loss": 3.1234,
+      "step": 3191
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.346450577582178e-06,
+      "loss": 3.0368,
+      "step": 3192
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.344675263411522e-06,
+      "loss": 2.9172,
+      "step": 3193
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.342899570243993e-06,
+      "loss": 2.9547,
+      "step": 3194
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.341123498366614e-06,
+      "loss": 3.0409,
+      "step": 3195
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.339347048066474e-06,
+      "loss": 3.0099,
+      "step": 3196
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.337570219630723e-06,
+      "loss": 3.0359,
+      "step": 3197
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.335793013346568e-06,
+      "loss": 3.0139,
+      "step": 3198
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.3340154295012796e-06,
+      "loss": 3.1338,
+      "step": 3199
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.33223746838219e-06,
+      "loss": 2.9811,
+      "step": 3200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.33045913027669e-06,
+      "loss": 3.1042,
+      "step": 3201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.328680415472235e-06,
+      "loss": 2.9218,
+      "step": 3202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.326901324256337e-06,
+      "loss": 2.9527,
+      "step": 3203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.325121856916573e-06,
+      "loss": 2.9524,
+      "step": 3204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.3233420137405775e-06,
+      "loss": 2.8397,
+      "step": 3205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.321561795016045e-06,
+      "loss": 2.9923,
+      "step": 3206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.319781201030739e-06,
+      "loss": 3.0224,
+      "step": 3207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.3180002320724705e-06,
+      "loss": 3.0681,
+      "step": 3208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.316218888429123e-06,
+      "loss": 2.9108,
+      "step": 3209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.314437170388632e-06,
+      "loss": 2.9426,
+      "step": 3210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.312655078238999e-06,
+      "loss": 3.0731,
+      "step": 3211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.310872612268286e-06,
+      "loss": 2.9117,
+      "step": 3212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.3090897727646105e-06,
+      "loss": 3.0505,
+      "step": 3213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.307306560016152e-06,
+      "loss": 3.0636,
+      "step": 3214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.305522974311157e-06,
+      "loss": 3.0812,
+      "step": 3215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.303739015937922e-06,
+      "loss": 3.1449,
+      "step": 3216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.301954685184811e-06,
+      "loss": 2.9265,
+      "step": 3217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.300169982340246e-06,
+      "loss": 3.0933,
+      "step": 3218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.298384907692709e-06,
+      "loss": 3.133,
+      "step": 3219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.296599461530741e-06,
+      "loss": 2.9374,
+      "step": 3220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.294813644142946e-06,
+      "loss": 2.9431,
+      "step": 3221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2930274558179844e-06,
+      "loss": 2.8775,
+      "step": 3222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.291240896844581e-06,
+      "loss": 2.9545,
+      "step": 3223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.289453967511515e-06,
+      "loss": 2.9437,
+      "step": 3224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.28766666810763e-06,
+      "loss": 2.9717,
+      "step": 3225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.285878998921828e-06,
+      "loss": 2.9088,
+      "step": 3226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.28409096024307e-06,
+      "loss": 3.0068,
+      "step": 3227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.282302552360377e-06,
+      "loss": 3.1125,
+      "step": 3228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.280513775562829e-06,
+      "loss": 2.9113,
+      "step": 3229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.278724630139569e-06,
+      "loss": 2.9046,
+      "step": 3230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2769351163797965e-06,
+      "loss": 2.9422,
+      "step": 3231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.27514523457277e-06,
+      "loss": 2.8168,
+      "step": 3232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.27335498500781e-06,
+      "loss": 2.9264,
+      "step": 3233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2715643679742946e-06,
+      "loss": 2.9127,
+      "step": 3234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2697733837616604e-06,
+      "loss": 3.1283,
+      "step": 3235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.267982032659409e-06,
+      "loss": 2.992,
+      "step": 3236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.266190314957093e-06,
+      "loss": 3.0927,
+      "step": 3237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.264398230944331e-06,
+      "loss": 2.8827,
+      "step": 3238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.262605780910795e-06,
+      "loss": 3.0793,
+      "step": 3239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.260812965146222e-06,
+      "loss": 2.9505,
+      "step": 3240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.259019783940405e-06,
+      "loss": 3.0013,
+      "step": 3241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.257226237583197e-06,
+      "loss": 2.9772,
+      "step": 3242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2554323263645085e-06,
+      "loss": 2.9817,
+      "step": 3243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.253638050574311e-06,
+      "loss": 2.9871,
+      "step": 3244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.251843410502632e-06,
+      "loss": 3.0147,
+      "step": 3245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.250048406439562e-06,
+      "loss": 3.0411,
+      "step": 3246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.248253038675247e-06,
+      "loss": 2.8461,
+      "step": 3247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2464573074998945e-06,
+      "loss": 3.0048,
+      "step": 3248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.244661213203767e-06,
+      "loss": 2.9538,
+      "step": 3249
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.24286475607719e-06,
+      "loss": 3.084,
+      "step": 3250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.241067936410544e-06,
+      "loss": 2.9949,
+      "step": 3251
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2392707544942716e-06,
+      "loss": 2.8066,
+      "step": 3252
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.237473210618868e-06,
+      "loss": 3.0307,
+      "step": 3253
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.235675305074895e-06,
+      "loss": 3.0587,
+      "step": 3254
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.233877038152968e-06,
+      "loss": 3.0201,
+      "step": 3255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.232078410143761e-06,
+      "loss": 3.0432,
+      "step": 3256
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.230279421338007e-06,
+      "loss": 2.9068,
+      "step": 3257
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.228480072026498e-06,
+      "loss": 2.993,
+      "step": 3258
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.22668036250008e-06,
+      "loss": 3.196,
+      "step": 3259
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.224880293049665e-06,
+      "loss": 3.0663,
+      "step": 3260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.223079863966218e-06,
+      "loss": 3.1104,
+      "step": 3261
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.221279075540764e-06,
+      "loss": 3.0105,
+      "step": 3262
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.219477928064383e-06,
+      "loss": 3.024,
+      "step": 3263
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.217676421828214e-06,
+      "loss": 2.8891,
+      "step": 3264
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.215874557123458e-06,
+      "loss": 3.032,
+      "step": 3265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.214072334241371e-06,
+      "loss": 3.025,
+      "step": 3266
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.212269753473266e-06,
+      "loss": 2.9116,
+      "step": 3267
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.210466815110516e-06,
+      "loss": 2.8208,
+      "step": 3268
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.20866351944455e-06,
+      "loss": 2.93,
+      "step": 3269
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.206859866766856e-06,
+      "loss": 2.9662,
+      "step": 3270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.205055857368978e-06,
+      "loss": 3.0008,
+      "step": 3271
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.2032514915425174e-06,
+      "loss": 3.1483,
+      "step": 3272
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.20144676957914e-06,
+      "loss": 3.0996,
+      "step": 3273
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.199641691770558e-06,
+      "loss": 3.0527,
+      "step": 3274
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.197836258408552e-06,
+      "loss": 3.0007,
+      "step": 3275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.196030469784951e-06,
+      "loss": 3.1795,
+      "step": 3276
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.194224326191646e-06,
+      "loss": 3.0681,
+      "step": 3277
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.192417827920587e-06,
+      "loss": 2.9776,
+      "step": 3278
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.190610975263776e-06,
+      "loss": 3.0336,
+      "step": 3279
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.18880376851328e-06,
+      "loss": 3.1331,
+      "step": 3280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.186996207961214e-06,
+      "loss": 3.0176,
+      "step": 3281
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.185188293899757e-06,
+      "loss": 3.0759,
+      "step": 3282
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.183380026621143e-06,
+      "loss": 2.9632,
+      "step": 3283
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.1815714064176624e-06,
+      "loss": 2.8871,
+      "step": 3284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.179762433581663e-06,
+      "loss": 3.0254,
+      "step": 3285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.177953108405551e-06,
+      "loss": 3.0563,
+      "step": 3286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.176143431181788e-06,
+      "loss": 3.0275,
+      "step": 3287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.174333402202893e-06,
+      "loss": 2.9162,
+      "step": 3288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.172523021761442e-06,
+      "loss": 3.0174,
+      "step": 3289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.170712290150067e-06,
+      "loss": 2.9797,
+      "step": 3290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.168901207661458e-06,
+      "loss": 3.0986,
+      "step": 3291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.167089774588361e-06,
+      "loss": 3.0563,
+      "step": 3292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.1652779912235796e-06,
+      "loss": 2.9404,
+      "step": 3293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.163465857859972e-06,
+      "loss": 3.1757,
+      "step": 3294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.161653374790454e-06,
+      "loss": 2.9407,
+      "step": 3295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.159840542307999e-06,
+      "loss": 2.9247,
+      "step": 3296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.158027360705637e-06,
+      "loss": 2.9339,
+      "step": 3297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.156213830276451e-06,
+      "loss": 2.9568,
+      "step": 3298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.154399951313582e-06,
+      "loss": 2.9151,
+      "step": 3299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.152585724110232e-06,
+      "loss": 2.8907,
+      "step": 3300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.150771148959655e-06,
+      "loss": 2.9933,
+      "step": 3301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.148956226155159e-06,
+      "loss": 2.9017,
+      "step": 3302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.147140955990111e-06,
+      "loss": 3.0461,
+      "step": 3303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.145325338757937e-06,
+      "loss": 3.0497,
+      "step": 3304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.143509374752113e-06,
+      "loss": 3.0391,
+      "step": 3305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.141693064266177e-06,
+      "loss": 2.9902,
+      "step": 3306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.139876407593718e-06,
+      "loss": 3.026,
+      "step": 3307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.138059405028383e-06,
+      "loss": 3.0573,
+      "step": 3308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.136242056863876e-06,
+      "loss": 3.0099,
+      "step": 3309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.134424363393957e-06,
+      "loss": 3.0571,
+      "step": 3310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.132606324912436e-06,
+      "loss": 2.9871,
+      "step": 3311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.1307879417131905e-06,
+      "loss": 2.9755,
+      "step": 3312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.128969214090142e-06,
+      "loss": 2.9533,
+      "step": 3313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.127150142337275e-06,
+      "loss": 3.0893,
+      "step": 3314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.125330726748625e-06,
+      "loss": 2.933,
+      "step": 3315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.123510967618286e-06,
+      "loss": 2.909,
+      "step": 3316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.121690865240405e-06,
+      "loss": 2.9386,
+      "step": 3317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.119870419909187e-06,
+      "loss": 2.9582,
+      "step": 3318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.118049631918895e-06,
+      "loss": 2.9894,
+      "step": 3319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.11622850156384e-06,
+      "loss": 3.0101,
+      "step": 3320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.1144070291383925e-06,
+      "loss": 3.2185,
+      "step": 3321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.1125852149369805e-06,
+      "loss": 2.9364,
+      "step": 3322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.11076305925408e-06,
+      "loss": 3.053,
+      "step": 3323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.108940562384231e-06,
+      "loss": 2.975,
+      "step": 3324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.107117724622026e-06,
+      "loss": 2.9885,
+      "step": 3325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.105294546262108e-06,
+      "loss": 2.9978,
+      "step": 3326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.10347102759918e-06,
+      "loss": 2.9139,
+      "step": 3327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.1016471689279965e-06,
+      "loss": 3.0832,
+      "step": 3328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.099822970543371e-06,
+      "loss": 3.0935,
+      "step": 3329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.097998432740168e-06,
+      "loss": 2.8354,
+      "step": 3330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.096173555813309e-06,
+      "loss": 2.8889,
+      "step": 3331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0943483400577715e-06,
+      "loss": 3.009,
+      "step": 3332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.092522785768584e-06,
+      "loss": 2.9029,
+      "step": 3333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.090696893240834e-06,
+      "loss": 2.9723,
+      "step": 3334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.088870662769659e-06,
+      "loss": 3.0246,
+      "step": 3335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.087044094650255e-06,
+      "loss": 2.9143,
+      "step": 3336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.085217189177871e-06,
+      "loss": 2.9909,
+      "step": 3337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.083389946647812e-06,
+      "loss": 2.8726,
+      "step": 3338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.081562367355434e-06,
+      "loss": 2.9661,
+      "step": 3339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.079734451596152e-06,
+      "loss": 2.9739,
+      "step": 3340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.077906199665432e-06,
+      "loss": 3.1066,
+      "step": 3341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.076077611858794e-06,
+      "loss": 2.9781,
+      "step": 3342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0742486884718155e-06,
+      "loss": 2.9893,
+      "step": 3343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.072419429800127e-06,
+      "loss": 3.0601,
+      "step": 3344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.070589836139411e-06,
+      "loss": 3.1256,
+      "step": 3345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.068759907785407e-06,
+      "loss": 2.8717,
+      "step": 3346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0669296450339075e-06,
+      "loss": 3.0401,
+      "step": 3347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.065099048180757e-06,
+      "loss": 2.9947,
+      "step": 3348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.063268117521858e-06,
+      "loss": 3.088,
+      "step": 3349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.061436853353164e-06,
+      "loss": 2.887,
+      "step": 3350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.059605255970685e-06,
+      "loss": 2.9464,
+      "step": 3351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.057773325670479e-06,
+      "loss": 2.9586,
+      "step": 3352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.055941062748667e-06,
+      "loss": 3.1436,
+      "step": 3353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0541084675014175e-06,
+      "loss": 2.9133,
+      "step": 3354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.05227554022495e-06,
+      "loss": 3.0611,
+      "step": 3355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.050442281215546e-06,
+      "loss": 2.9608,
+      "step": 3356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.048608690769535e-06,
+      "loss": 2.936,
+      "step": 3357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.046774769183301e-06,
+      "loss": 3.0218,
+      "step": 3358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0449405167532834e-06,
+      "loss": 3.1076,
+      "step": 3359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.043105933775971e-06,
+      "loss": 2.9234,
+      "step": 3360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.041271020547911e-06,
+      "loss": 2.9196,
+      "step": 3361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.039435777365699e-06,
+      "loss": 3.032,
+      "step": 3362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0376002045259885e-06,
+      "loss": 2.8313,
+      "step": 3363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.035764302325484e-06,
+      "loss": 3.0847,
+      "step": 3364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.0339280710609436e-06,
+      "loss": 3.01,
+      "step": 3365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.032091511029178e-06,
+      "loss": 2.8784,
+      "step": 3366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.03025462252705e-06,
+      "loss": 3.0589,
+      "step": 3367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.028417405851482e-06,
+      "loss": 3.0946,
+      "step": 3368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.026579861299438e-06,
+      "loss": 2.9311,
+      "step": 3369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.024741989167947e-06,
+      "loss": 2.8713,
+      "step": 3370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.022903789754083e-06,
+      "loss": 2.9608,
+      "step": 3371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.021065263354974e-06,
+      "loss": 3.059,
+      "step": 3372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.019226410267805e-06,
+      "loss": 3.0631,
+      "step": 3373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.01738723078981e-06,
+      "loss": 2.9943,
+      "step": 3374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.0155477252182755e-06,
+      "loss": 2.8213,
+      "step": 3375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.013707893850544e-06,
+      "loss": 3.0873,
+      "step": 3376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.011867736984007e-06,
+      "loss": 3.0009,
+      "step": 3377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.010027254916112e-06,
+      "loss": 3.002,
+      "step": 3378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.008186447944355e-06,
+      "loss": 3.0341,
+      "step": 3379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.006345316366289e-06,
+      "loss": 2.9866,
+      "step": 3380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.004503860479514e-06,
+      "loss": 3.034,
+      "step": 3381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.00266208058169e-06,
+      "loss": 2.8444,
+      "step": 3382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.000819976970523e-06,
+      "loss": 2.9612,
+      "step": 3383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.998977549943775e-06,
+      "loss": 2.923,
+      "step": 3384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.997134799799257e-06,
+      "loss": 2.8837,
+      "step": 3385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.995291726834833e-06,
+      "loss": 3.006,
+      "step": 3386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.99344833134842e-06,
+      "loss": 2.9696,
+      "step": 3387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.99160461363799e-06,
+      "loss": 3.0204,
+      "step": 3388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9897605740015635e-06,
+      "loss": 2.9393,
+      "step": 3389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.987916212737214e-06,
+      "loss": 3.0859,
+      "step": 3390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.986071530143066e-06,
+      "loss": 3.118,
+      "step": 3391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.984226526517295e-06,
+      "loss": 2.9809,
+      "step": 3392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.982381202158134e-06,
+      "loss": 2.9718,
+      "step": 3393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.980535557363861e-06,
+      "loss": 2.9765,
+      "step": 3394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.978689592432811e-06,
+      "loss": 2.9415,
+      "step": 3395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.976843307663367e-06,
+      "loss": 2.8816,
+      "step": 3396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.974996703353966e-06,
+      "loss": 3.0121,
+      "step": 3397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.973149779803097e-06,
+      "loss": 3.0556,
+      "step": 3398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9713025373092965e-06,
+      "loss": 2.8786,
+      "step": 3399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.969454976171158e-06,
+      "loss": 2.9992,
+      "step": 3400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9676070966873235e-06,
+      "loss": 2.9443,
+      "step": 3401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.965758899156488e-06,
+      "loss": 2.8415,
+      "step": 3402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.963910383877396e-06,
+      "loss": 2.9809,
+      "step": 3403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.962061551148845e-06,
+      "loss": 2.9919,
+      "step": 3404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.96021240126968e-06,
+      "loss": 3.1062,
+      "step": 3405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.958362934538805e-06,
+      "loss": 3.0477,
+      "step": 3406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.956513151255166e-06,
+      "loss": 2.888,
+      "step": 3407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.954663051717769e-06,
+      "loss": 2.8781,
+      "step": 3408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.952812636225665e-06,
+      "loss": 2.9388,
+      "step": 3409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9509619050779594e-06,
+      "loss": 2.7438,
+      "step": 3410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.949110858573804e-06,
+      "loss": 2.9925,
+      "step": 3411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9472594970124085e-06,
+      "loss": 2.973,
+      "step": 3412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.945407820693027e-06,
+      "loss": 3.0292,
+      "step": 3413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.943555829914967e-06,
+      "loss": 2.9481,
+      "step": 3414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.941703524977591e-06,
+      "loss": 3.0667,
+      "step": 3415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9398509061803035e-06,
+      "loss": 2.7606,
+      "step": 3416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.937997973822569e-06,
+      "loss": 2.9676,
+      "step": 3417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.936144728203895e-06,
+      "loss": 2.9997,
+      "step": 3418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9342911696238435e-06,
+      "loss": 3.0187,
+      "step": 3419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.932437298382029e-06,
+      "loss": 2.968,
+      "step": 3420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.93058311477811e-06,
+      "loss": 2.975,
+      "step": 3421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.928728619111803e-06,
+      "loss": 3.0444,
+      "step": 3422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.92687381168287e-06,
+      "loss": 3.0262,
+      "step": 3423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9250186927911265e-06,
+      "loss": 3.0904,
+      "step": 3424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.923163262736434e-06,
+      "loss": 3.0591,
+      "step": 3425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.921307521818708e-06,
+      "loss": 3.0728,
+      "step": 3426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.919451470337914e-06,
+      "loss": 3.1541,
+      "step": 3427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.917595108594068e-06,
+      "loss": 3.0857,
+      "step": 3428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.915738436887234e-06,
+      "loss": 2.7111,
+      "step": 3429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.913881455517526e-06,
+      "loss": 3.112,
+      "step": 3430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.912024164785111e-06,
+      "loss": 3.0293,
+      "step": 3431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9101665649902025e-06,
+      "loss": 2.9874,
+      "step": 3432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.908308656433067e-06,
+      "loss": 3.0293,
+      "step": 3433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.9064504394140206e-06,
+      "loss": 2.7421,
+      "step": 3434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.904591914233429e-06,
+      "loss": 3.0641,
+      "step": 3435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.902733081191701e-06,
+      "loss": 3.1429,
+      "step": 3436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.900873940589309e-06,
+      "loss": 2.927,
+      "step": 3437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.8990144927267636e-06,
+      "loss": 2.854,
+      "step": 3438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.897154737904626e-06,
+      "loss": 3.0648,
+      "step": 3439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.895294676423516e-06,
+      "loss": 3.1716,
+      "step": 3440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.893434308584093e-06,
+      "loss": 3.0307,
+      "step": 3441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.89157363468707e-06,
+      "loss": 2.9841,
+      "step": 3442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.88971265503321e-06,
+      "loss": 2.9531,
+      "step": 3443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.887851369923323e-06,
+      "loss": 2.9509,
+      "step": 3444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.885989779658271e-06,
+      "loss": 2.995,
+      "step": 3445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.8841278845389635e-06,
+      "loss": 3.0255,
+      "step": 3446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.882265684866363e-06,
+      "loss": 3.0932,
+      "step": 3447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.880403180941475e-06,
+      "loss": 2.9857,
+      "step": 3448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.878540373065359e-06,
+      "loss": 3.0277,
+      "step": 3449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.8766772615391205e-06,
+      "loss": 3.0744,
+      "step": 3450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.874813846663918e-06,
+      "loss": 3.0416,
+      "step": 3451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.872950128740955e-06,
+      "loss": 2.8205,
+      "step": 3452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.871086108071486e-06,
+      "loss": 3.0946,
+      "step": 3453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.869221784956816e-06,
+      "loss": 2.984,
+      "step": 3454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.867357159698294e-06,
+      "loss": 3.1242,
+      "step": 3455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.865492232597322e-06,
+      "loss": 3.0348,
+      "step": 3456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.863627003955351e-06,
+      "loss": 3.0718,
+      "step": 3457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.861761474073877e-06,
+      "loss": 3.0035,
+      "step": 3458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.859895643254449e-06,
+      "loss": 2.9838,
+      "step": 3459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8580295117986625e-06,
+      "loss": 2.8815,
+      "step": 3460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.856163080008162e-06,
+      "loss": 2.9993,
+      "step": 3461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.854296348184639e-06,
+      "loss": 2.9847,
+      "step": 3462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.852429316629836e-06,
+      "loss": 3.0085,
+      "step": 3463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.850561985645542e-06,
+      "loss": 3.0705,
+      "step": 3464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.848694355533595e-06,
+      "loss": 3.1584,
+      "step": 3465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.846826426595884e-06,
+      "loss": 3.0979,
+      "step": 3466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.844958199134341e-06,
+      "loss": 3.064,
+      "step": 3467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.843089673450953e-06,
+      "loss": 3.0203,
+      "step": 3468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8412208498477465e-06,
+      "loss": 3.0928,
+      "step": 3469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.839351728626802e-06,
+      "loss": 2.9875,
+      "step": 3470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8374823100902486e-06,
+      "loss": 3.0955,
+      "step": 3471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.835612594540263e-06,
+      "loss": 3.1521,
+      "step": 3472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.833742582279067e-06,
+      "loss": 2.866,
+      "step": 3473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.831872273608932e-06,
+      "loss": 3.0411,
+      "step": 3474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.830001668832178e-06,
+      "loss": 3.1536,
+      "step": 3475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.828130768251175e-06,
+      "loss": 3.0436,
+      "step": 3476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.826259572168332e-06,
+      "loss": 3.0529,
+      "step": 3477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.824388080886119e-06,
+      "loss": 2.8974,
+      "step": 3478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8225162947070425e-06,
+      "loss": 3.0985,
+      "step": 3479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8206442139336604e-06,
+      "loss": 3.0071,
+      "step": 3480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.818771838868582e-06,
+      "loss": 2.9215,
+      "step": 3481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.816899169814459e-06,
+      "loss": 2.9781,
+      "step": 3482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.8150262070739916e-06,
+      "loss": 2.9315,
+      "step": 3483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.813152950949928e-06,
+      "loss": 3.014,
+      "step": 3484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.811279401745067e-06,
+      "loss": 3.1035,
+      "step": 3485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.80940555976225e-06,
+      "loss": 3.1682,
+      "step": 3486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.807531425304369e-06,
+      "loss": 3.0055,
+      "step": 3487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.80565699867436e-06,
+      "loss": 2.9238,
+      "step": 3488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.803782280175209e-06,
+      "loss": 2.9726,
+      "step": 3489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.801907270109948e-06,
+      "loss": 2.976,
+      "step": 3490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.800031968781657e-06,
+      "loss": 2.9868,
+      "step": 3491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.798156376493463e-06,
+      "loss": 3.0678,
+      "step": 3492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.796280493548539e-06,
+      "loss": 2.809,
+      "step": 3493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.794404320250105e-06,
+      "loss": 2.8789,
+      "step": 3494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.792527856901431e-06,
+      "loss": 2.8555,
+      "step": 3495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7906511038058276e-06,
+      "loss": 3.0818,
+      "step": 3496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.788774061266659e-06,
+      "loss": 2.9536,
+      "step": 3497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.786896729587332e-06,
+      "loss": 3.0739,
+      "step": 3498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7850191090713e-06,
+      "loss": 3.0708,
+      "step": 3499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.783141200022068e-06,
+      "loss": 2.9034,
+      "step": 3500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.781263002743182e-06,
+      "loss": 3.0415,
+      "step": 3501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.779384517538235e-06,
+      "loss": 2.9062,
+      "step": 3502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.77750574471087e-06,
+      "loss": 2.9649,
+      "step": 3503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.775626684564776e-06,
+      "loss": 3.103,
+      "step": 3504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.773747337403684e-06,
+      "loss": 2.9498,
+      "step": 3505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.771867703531378e-06,
+      "loss": 2.9433,
+      "step": 3506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.769987783251682e-06,
+      "loss": 3.0702,
+      "step": 3507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.768107576868472e-06,
+      "loss": 3.0109,
+      "step": 3508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.766227084685664e-06,
+      "loss": 2.8181,
+      "step": 3509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7643463070072245e-06,
+      "loss": 3.1435,
+      "step": 3510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7624652441371674e-06,
+      "loss": 3.0714,
+      "step": 3511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.760583896379549e-06,
+      "loss": 2.8348,
+      "step": 3512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.758702264038475e-06,
+      "loss": 2.9354,
+      "step": 3513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.756820347418092e-06,
+      "loss": 3.0733,
+      "step": 3514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.754938146822597e-06,
+      "loss": 2.9813,
+      "step": 3515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.753055662556233e-06,
+      "loss": 2.9742,
+      "step": 3516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.751172894923288e-06,
+      "loss": 2.9743,
+      "step": 3517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.749289844228094e-06,
+      "loss": 3.0803,
+      "step": 3518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.74740651077503e-06,
+      "loss": 2.9796,
+      "step": 3519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.745522894868522e-06,
+      "loss": 2.8343,
+      "step": 3520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.74363899681304e-06,
+      "loss": 2.9799,
+      "step": 3521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7417548169131e-06,
+      "loss": 2.7762,
+      "step": 3522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.739870355473263e-06,
+      "loss": 3.1576,
+      "step": 3523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.737985612798138e-06,
+      "loss": 3.0666,
+      "step": 3524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.736100589192377e-06,
+      "loss": 3.1436,
+      "step": 3525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.734215284960678e-06,
+      "loss": 2.8529,
+      "step": 3526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.732329700407785e-06,
+      "loss": 3.0145,
+      "step": 3527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.730443835838484e-06,
+      "loss": 2.994,
+      "step": 3528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.728557691557611e-06,
+      "loss": 2.9213,
+      "step": 3529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.726671267870046e-06,
+      "loss": 3.0185,
+      "step": 3530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.724784565080714e-06,
+      "loss": 3.2182,
+      "step": 3531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.722897583494583e-06,
+      "loss": 3.0054,
+      "step": 3532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.7210103234166665e-06,
+      "loss": 3.0101,
+      "step": 3533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.719122785152026e-06,
+      "loss": 2.906,
+      "step": 3534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.717234969005764e-06,
+      "loss": 3.0652,
+      "step": 3535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.715346875283033e-06,
+      "loss": 2.9606,
+      "step": 3536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.713458504289026e-06,
+      "loss": 3.0801,
+      "step": 3537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.71156985632898e-06,
+      "loss": 2.8661,
+      "step": 3538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.709680931708182e-06,
+      "loss": 2.9187,
+      "step": 3539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.70779173073196e-06,
+      "loss": 2.9625,
+      "step": 3540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.705902253705685e-06,
+      "loss": 3.1353,
+      "step": 3541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.704012500934778e-06,
+      "loss": 3.0863,
+      "step": 3542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.702122472724699e-06,
+      "loss": 3.0347,
+      "step": 3543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.700232169380958e-06,
+      "loss": 3.0136,
+      "step": 3544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.698341591209104e-06,
+      "loss": 3.0469,
+      "step": 3545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.696450738514735e-06,
+      "loss": 2.9452,
+      "step": 3546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.694559611603488e-06,
+      "loss": 3.0433,
+      "step": 3547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.692668210781049e-06,
+      "loss": 3.1116,
+      "step": 3548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.69077653635315e-06,
+      "loss": 3.183,
+      "step": 3549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.688884588625559e-06,
+      "loss": 3.0103,
+      "step": 3550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.686992367904098e-06,
+      "loss": 2.915,
+      "step": 3551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.685099874494626e-06,
+      "loss": 2.9984,
+      "step": 3552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.683207108703048e-06,
+      "loss": 2.8662,
+      "step": 3553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6813140708353146e-06,
+      "loss": 2.917,
+      "step": 3554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.679420761197419e-06,
+      "loss": 2.9873,
+      "step": 3555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6775271800954e-06,
+      "loss": 2.9634,
+      "step": 3556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.675633327835338e-06,
+      "loss": 2.9346,
+      "step": 3557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6737392047233565e-06,
+      "loss": 2.8475,
+      "step": 3558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.671844811065627e-06,
+      "loss": 2.9604,
+      "step": 3559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.669950147168359e-06,
+      "loss": 2.9954,
+      "step": 3560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.668055213337812e-06,
+      "loss": 3.0087,
+      "step": 3561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.666160009880286e-06,
+      "loss": 2.8792,
+      "step": 3562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.664264537102121e-06,
+      "loss": 2.9689,
+      "step": 3563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.662368795309709e-06,
+      "loss": 2.9908,
+      "step": 3564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.660472784809478e-06,
+      "loss": 3.0942,
+      "step": 3565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.658576505907899e-06,
+      "loss": 2.9001,
+      "step": 3566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.656679958911495e-06,
+      "loss": 3.0383,
+      "step": 3567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.654783144126823e-06,
+      "loss": 3.1575,
+      "step": 3568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.652886061860489e-06,
+      "loss": 2.9869,
+      "step": 3569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.65098871241914e-06,
+      "loss": 3.1065,
+      "step": 3570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.649091096109466e-06,
+      "loss": 2.9399,
+      "step": 3571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.647193213238199e-06,
+      "loss": 2.9912,
+      "step": 3572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.645295064112118e-06,
+      "loss": 3.1085,
+      "step": 3573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6433966490380434e-06,
+      "loss": 2.7541,
+      "step": 3574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.641497968322836e-06,
+      "loss": 2.9064,
+      "step": 3575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.639599022273401e-06,
+      "loss": 3.023,
+      "step": 3576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.637699811196688e-06,
+      "loss": 2.9958,
+      "step": 3577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.635800335399687e-06,
+      "loss": 3.005,
+      "step": 3578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.633900595189435e-06,
+      "loss": 2.977,
+      "step": 3579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.632000590873006e-06,
+      "loss": 3.0322,
+      "step": 3580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.630100322757522e-06,
+      "loss": 3.0568,
+      "step": 3581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6281997911501435e-06,
+      "loss": 3.0517,
+      "step": 3582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.626298996358076e-06,
+      "loss": 3.1616,
+      "step": 3583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.624397938688567e-06,
+      "loss": 3.0368,
+      "step": 3584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.622496618448906e-06,
+      "loss": 3.031,
+      "step": 3585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.620595035946424e-06,
+      "loss": 2.8569,
+      "step": 3586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.618693191488498e-06,
+      "loss": 2.99,
+      "step": 3587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.6167910853825434e-06,
+      "loss": 3.0242,
+      "step": 3588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.614888717936022e-06,
+      "loss": 3.0427,
+      "step": 3589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.612986089456432e-06,
+      "loss": 2.943,
+      "step": 3590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.61108320025132e-06,
+      "loss": 2.9229,
+      "step": 3591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.60918005062827e-06,
+      "loss": 3.1438,
+      "step": 3592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.60727664089491e-06,
+      "loss": 2.7838,
+      "step": 3593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.605372971358911e-06,
+      "loss": 2.9696,
+      "step": 3594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.603469042327984e-06,
+      "loss": 3.0836,
+      "step": 3595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.601564854109885e-06,
+      "loss": 3.0524,
+      "step": 3596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.599660407012408e-06,
+      "loss": 2.9461,
+      "step": 3597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.597755701343391e-06,
+      "loss": 2.9051,
+      "step": 3598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.595850737410714e-06,
+      "loss": 3.0445,
+      "step": 3599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.593945515522299e-06,
+      "loss": 2.9024,
+      "step": 3600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.592040035986107e-06,
+      "loss": 3.1629,
+      "step": 3601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.590134299110144e-06,
+      "loss": 2.9402,
+      "step": 3602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.588228305202456e-06,
+      "loss": 3.0797,
+      "step": 3603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.58632205457113e-06,
+      "loss": 2.924,
+      "step": 3604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.584415547524295e-06,
+      "loss": 2.9204,
+      "step": 3605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.582508784370125e-06,
+      "loss": 2.9418,
+      "step": 3606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.580601765416827e-06,
+      "loss": 3.0217,
+      "step": 3607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.578694490972659e-06,
+      "loss": 2.9692,
+      "step": 3608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.576786961345914e-06,
+      "loss": 2.9774,
+      "step": 3609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.5748791768449274e-06,
+      "loss": 2.8781,
+      "step": 3610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.572971137778076e-06,
+      "loss": 3.206,
+      "step": 3611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.571062844453778e-06,
+      "loss": 3.0691,
+      "step": 3612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.569154297180495e-06,
+      "loss": 3.0664,
+      "step": 3613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.5672454962667276e-06,
+      "loss": 2.8326,
+      "step": 3614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.565336442021015e-06,
+      "loss": 2.8964,
+      "step": 3615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.563427134751941e-06,
+      "loss": 2.8621,
+      "step": 3616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.561517574768127e-06,
+      "loss": 3.0495,
+      "step": 3617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.559607762378238e-06,
+      "loss": 3.0848,
+      "step": 3618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5576976978909824e-06,
+      "loss": 2.974,
+      "step": 3619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.555787381615102e-06,
+      "loss": 2.9417,
+      "step": 3620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.553876813859384e-06,
+      "loss": 2.9727,
+      "step": 3621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.551965994932656e-06,
+      "loss": 2.8767,
+      "step": 3622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5500549251437864e-06,
+      "loss": 2.9738,
+      "step": 3623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.548143604801681e-06,
+      "loss": 3.0761,
+      "step": 3624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.54623203421529e-06,
+      "loss": 3.0666,
+      "step": 3625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5443202136936045e-06,
+      "loss": 3.0964,
+      "step": 3626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.542408143545651e-06,
+      "loss": 3.1146,
+      "step": 3627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5404958240805016e-06,
+      "loss": 3.0135,
+      "step": 3628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.538583255607266e-06,
+      "loss": 2.8998,
+      "step": 3629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.536670438435092e-06,
+      "loss": 3.0756,
+      "step": 3630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.534757372873173e-06,
+      "loss": 3.1025,
+      "step": 3631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5328440592307405e-06,
+      "loss": 3.0119,
+      "step": 3632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.530930497817065e-06,
+      "loss": 3.0531,
+      "step": 3633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.529016688941456e-06,
+      "loss": 3.0464,
+      "step": 3634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.527102632913266e-06,
+      "loss": 3.0182,
+      "step": 3635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.525188330041883e-06,
+      "loss": 3.0438,
+      "step": 3636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.523273780636741e-06,
+      "loss": 3.0772,
+      "step": 3637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.52135898500731e-06,
+      "loss": 3.0108,
+      "step": 3638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.5194439434631005e-06,
+      "loss": 3.0078,
+      "step": 3639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.517528656313662e-06,
+      "loss": 3.051,
+      "step": 3640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.515613123868585e-06,
+      "loss": 2.9573,
+      "step": 3641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.513697346437499e-06,
+      "loss": 2.9408,
+      "step": 3642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.511781324330071e-06,
+      "loss": 2.9388,
+      "step": 3643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.509865057856014e-06,
+      "loss": 3.1403,
+      "step": 3644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.507948547325075e-06,
+      "loss": 2.8739,
+      "step": 3645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.506031793047039e-06,
+      "loss": 3.0035,
+      "step": 3646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.504114795331734e-06,
+      "loss": 3.0186,
+      "step": 3647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.502197554489029e-06,
+      "loss": 3.1015,
+      "step": 3648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.500280070828827e-06,
+      "loss": 2.9465,
+      "step": 3649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.498362344661073e-06,
+      "loss": 3.0636,
+      "step": 3650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.496444376295753e-06,
+      "loss": 2.9983,
+      "step": 3651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.494526166042889e-06,
+      "loss": 3.023,
+      "step": 3652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.492607714212545e-06,
+      "loss": 2.9749,
+      "step": 3653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.490689021114821e-06,
+      "loss": 2.958,
+      "step": 3654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.488770087059858e-06,
+      "loss": 2.924,
+      "step": 3655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.486850912357834e-06,
+      "loss": 3.1338,
+      "step": 3656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.484931497318967e-06,
+      "loss": 3.0875,
+      "step": 3657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.483011842253518e-06,
+      "loss": 2.8549,
+      "step": 3658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.481091947471781e-06,
+      "loss": 2.9503,
+      "step": 3659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.47917181328409e-06,
+      "loss": 2.9811,
+      "step": 3660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.477251440000817e-06,
+      "loss": 3.0109,
+      "step": 3661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.475330827932378e-06,
+      "loss": 2.8884,
+      "step": 3662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.473409977389219e-06,
+      "loss": 2.9393,
+      "step": 3663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.471488888681834e-06,
+      "loss": 2.9125,
+      "step": 3664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.469567562120747e-06,
+      "loss": 2.9223,
+      "step": 3665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.4676459980165255e-06,
+      "loss": 2.9287,
+      "step": 3666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.465724196679774e-06,
+      "loss": 2.925,
+      "step": 3667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.4638021584211366e-06,
+      "loss": 3.0048,
+      "step": 3668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.461879883551292e-06,
+      "loss": 2.8062,
+      "step": 3669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.459957372380961e-06,
+      "loss": 2.9655,
+      "step": 3670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.4580346252209015e-06,
+      "loss": 2.8718,
+      "step": 3671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.456111642381908e-06,
+      "loss": 3.0602,
+      "step": 3672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.454188424174818e-06,
+      "loss": 3.1103,
+      "step": 3673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.452264970910499e-06,
+      "loss": 3.0626,
+      "step": 3674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.450341282899861e-06,
+      "loss": 2.8443,
+      "step": 3675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.448417360453855e-06,
+      "loss": 2.7236,
+      "step": 3676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.446493203883466e-06,
+      "loss": 3.0145,
+      "step": 3677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.444568813499717e-06,
+      "loss": 3.1417,
+      "step": 3678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.442644189613667e-06,
+      "loss": 3.12,
+      "step": 3679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.440719332536418e-06,
+      "loss": 2.8977,
+      "step": 3680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.438794242579107e-06,
+      "loss": 2.8155,
+      "step": 3681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.436868920052907e-06,
+      "loss": 2.9484,
+      "step": 3682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.434943365269032e-06,
+      "loss": 2.9451,
+      "step": 3683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.433017578538729e-06,
+      "loss": 2.8581,
+      "step": 3684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.431091560173286e-06,
+      "loss": 2.9889,
+      "step": 3685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.429165310484029e-06,
+      "loss": 2.935,
+      "step": 3686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.427238829782318e-06,
+      "loss": 3.0624,
+      "step": 3687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.425312118379554e-06,
+      "loss": 2.8836,
+      "step": 3688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.423385176587172e-06,
+      "loss": 3.0487,
+      "step": 3689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.421458004716647e-06,
+      "loss": 2.8911,
+      "step": 3690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.419530603079489e-06,
+      "loss": 2.8346,
+      "step": 3691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.417602971987247e-06,
+      "loss": 2.9978,
+      "step": 3692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.415675111751506e-06,
+      "loss": 3.0464,
+      "step": 3693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.413747022683887e-06,
+      "loss": 2.9638,
+      "step": 3694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.4118187050960516e-06,
+      "loss": 2.9675,
+      "step": 3695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.409890159299693e-06,
+      "loss": 2.9571,
+      "step": 3696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.407961385606548e-06,
+      "loss": 3.0513,
+      "step": 3697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.4060323843283836e-06,
+      "loss": 3.0577,
+      "step": 3698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.404103155777009e-06,
+      "loss": 2.8703,
+      "step": 3699
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.402173700264263e-06,
+      "loss": 2.933,
+      "step": 3700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.400244018102029e-06,
+      "loss": 2.9278,
+      "step": 3701
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.398314109602225e-06,
+      "loss": 3.1279,
+      "step": 3702
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.396383975076803e-06,
+      "loss": 3.032,
+      "step": 3703
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.394453614837752e-06,
+      "loss": 2.9648,
+      "step": 3704
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.392523029197099e-06,
+      "loss": 2.9867,
+      "step": 3705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.390592218466908e-06,
+      "loss": 2.9827,
+      "step": 3706
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3886611829592745e-06,
+      "loss": 2.9137,
+      "step": 3707
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.386729922986339e-06,
+      "loss": 2.9536,
+      "step": 3708
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.384798438860271e-06,
+      "loss": 2.9567,
+      "step": 3709
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.382866730893276e-06,
+      "loss": 2.9686,
+      "step": 3710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.380934799397603e-06,
+      "loss": 2.9619,
+      "step": 3711
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.37900264468553e-06,
+      "loss": 3.0007,
+      "step": 3712
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.377070267069372e-06,
+      "loss": 2.9281,
+      "step": 3713
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3751376668614835e-06,
+      "loss": 3.034,
+      "step": 3714
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.373204844374253e-06,
+      "loss": 3.0805,
+      "step": 3715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3712717999201024e-06,
+      "loss": 2.7092,
+      "step": 3716
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.369338533811496e-06,
+      "loss": 2.9693,
+      "step": 3717
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.367405046360926e-06,
+      "loss": 3.0072,
+      "step": 3718
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.365471337880925e-06,
+      "loss": 2.9062,
+      "step": 3719
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.363537408684063e-06,
+      "loss": 3.0571,
+      "step": 3720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.36160325908294e-06,
+      "loss": 2.9462,
+      "step": 3721
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.359668889390198e-06,
+      "loss": 2.8996,
+      "step": 3722
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.357734299918511e-06,
+      "loss": 2.972,
+      "step": 3723
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.355799490980586e-06,
+      "loss": 3.0745,
+      "step": 3724
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.353864462889171e-06,
+      "loss": 2.8702,
+      "step": 3725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.351929215957046e-06,
+      "loss": 3.1087,
+      "step": 3726
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.349993750497028e-06,
+      "loss": 3.0466,
+      "step": 3727
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.348058066821968e-06,
+      "loss": 3.1364,
+      "step": 3728
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.346122165244753e-06,
+      "loss": 2.9796,
+      "step": 3729
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3441860460783046e-06,
+      "loss": 3.0054,
+      "step": 3730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.342249709635582e-06,
+      "loss": 2.8822,
+      "step": 3731
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.340313156229575e-06,
+      "loss": 3.1388,
+      "step": 3732
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.338376386173311e-06,
+      "loss": 3.1025,
+      "step": 3733
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.336439399779854e-06,
+      "loss": 2.8955,
+      "step": 3734
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.334502197362303e-06,
+      "loss": 2.9992,
+      "step": 3735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.332564779233787e-06,
+      "loss": 3.0179,
+      "step": 3736
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3306271457074755e-06,
+      "loss": 3.0556,
+      "step": 3737
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.328689297096568e-06,
+      "loss": 2.9891,
+      "step": 3738
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3267512337143035e-06,
+      "loss": 2.9388,
+      "step": 3739
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.324812955873954e-06,
+      "loss": 3.0365,
+      "step": 3740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3228744638888224e-06,
+      "loss": 2.9234,
+      "step": 3741
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.320935758072253e-06,
+      "loss": 2.9962,
+      "step": 3742
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3189968387376214e-06,
+      "loss": 2.7847,
+      "step": 3743
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.317057706198333e-06,
+      "loss": 2.8689,
+      "step": 3744
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.315118360767836e-06,
+      "loss": 2.9459,
+      "step": 3745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.313178802759607e-06,
+      "loss": 2.7553,
+      "step": 3746
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.311239032487162e-06,
+      "loss": 2.9119,
+      "step": 3747
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.309299050264047e-06,
+      "loss": 2.8745,
+      "step": 3748
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.307358856403842e-06,
+      "loss": 2.9316,
+      "step": 3749
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.305418451220164e-06,
+      "loss": 2.9767,
+      "step": 3750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.3034778350266655e-06,
+      "loss": 2.9293,
+      "step": 3751
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.301537008137026e-06,
+      "loss": 3.0383,
+      "step": 3752
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.299595970864968e-06,
+      "loss": 3.0028,
+      "step": 3753
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.297654723524241e-06,
+      "loss": 2.8813,
+      "step": 3754
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.295713266428631e-06,
+      "loss": 3.0483,
+      "step": 3755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.29377159989196e-06,
+      "loss": 3.0439,
+      "step": 3756
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.2918297242280826e-06,
+      "loss": 2.8738,
+      "step": 3757
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.289887639750882e-06,
+      "loss": 2.9508,
+      "step": 3758
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.287945346774283e-06,
+      "loss": 3.0668,
+      "step": 3759
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.286002845612243e-06,
+      "loss": 2.9707,
+      "step": 3760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.284060136578747e-06,
+      "loss": 2.946,
+      "step": 3761
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.282117219987818e-06,
+      "loss": 3.144,
+      "step": 3762
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.280174096153514e-06,
+      "loss": 3.0178,
+      "step": 3763
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.2782307653899214e-06,
+      "loss": 2.9112,
+      "step": 3764
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.276287228011165e-06,
+      "loss": 3.0736,
+      "step": 3765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.274343484331403e-06,
+      "loss": 2.961,
+      "step": 3766
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.272399534664822e-06,
+      "loss": 3.0793,
+      "step": 3767
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.270455379325647e-06,
+      "loss": 2.8462,
+      "step": 3768
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.268511018628132e-06,
+      "loss": 2.9005,
+      "step": 3769
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.266566452886569e-06,
+      "loss": 2.9569,
+      "step": 3770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.264621682415278e-06,
+      "loss": 2.991,
+      "step": 3771
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.262676707528617e-06,
+      "loss": 2.9337,
+      "step": 3772
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.260731528540973e-06,
+      "loss": 2.8875,
+      "step": 3773
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.258786145766769e-06,
+      "loss": 3.1073,
+      "step": 3774
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.25684055952046e-06,
+      "loss": 2.9551,
+      "step": 3775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.254894770116531e-06,
+      "loss": 2.9146,
+      "step": 3776
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.252948777869504e-06,
+      "loss": 2.7944,
+      "step": 3777
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.251002583093931e-06,
+      "loss": 3.1168,
+      "step": 3778
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.249056186104401e-06,
+      "loss": 2.9381,
+      "step": 3779
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.247109587215529e-06,
+      "loss": 3.0046,
+      "step": 3780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.24516278674197e-06,
+      "loss": 3.2034,
+      "step": 3781
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.243215784998404e-06,
+      "loss": 3.0207,
+      "step": 3782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.241268582299548e-06,
+      "loss": 3.0764,
+      "step": 3783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.239321178960152e-06,
+      "loss": 2.8775,
+      "step": 3784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.237373575294997e-06,
+      "loss": 2.9583,
+      "step": 3785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.235425771618897e-06,
+      "loss": 3.0419,
+      "step": 3786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.233477768246697e-06,
+      "loss": 3.0191,
+      "step": 3787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.231529565493275e-06,
+      "loss": 2.9479,
+      "step": 3788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.229581163673544e-06,
+      "loss": 3.077,
+      "step": 3789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.227632563102442e-06,
+      "loss": 3.0071,
+      "step": 3790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.225683764094948e-06,
+      "loss": 2.9879,
+      "step": 3791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.223734766966068e-06,
+      "loss": 3.0038,
+      "step": 3792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.221785572030839e-06,
+      "loss": 2.8921,
+      "step": 3793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.219836179604333e-06,
+      "loss": 3.0702,
+      "step": 3794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.217886590001653e-06,
+      "loss": 2.948,
+      "step": 3795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.215936803537934e-06,
+      "loss": 3.0535,
+      "step": 3796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.21398682052834e-06,
+      "loss": 2.8836,
+      "step": 3797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.212036641288072e-06,
+      "loss": 2.8653,
+      "step": 3798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.210086266132359e-06,
+      "loss": 3.0672,
+      "step": 3799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.208135695376464e-06,
+      "loss": 3.04,
+      "step": 3800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.2061849293356764e-06,
+      "loss": 2.8123,
+      "step": 3801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.204233968325326e-06,
+      "loss": 3.0305,
+      "step": 3802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.2022828126607634e-06,
+      "loss": 3.0761,
+      "step": 3803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.20033146265738e-06,
+      "loss": 2.8989,
+      "step": 3804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.198379918630597e-06,
+      "loss": 3.0262,
+      "step": 3805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.196428180895862e-06,
+      "loss": 2.9688,
+      "step": 3806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1944762497686564e-06,
+      "loss": 2.8935,
+      "step": 3807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.192524125564493e-06,
+      "loss": 3.0239,
+      "step": 3808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.190571808598918e-06,
+      "loss": 2.9749,
+      "step": 3809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.188619299187506e-06,
+      "loss": 2.8864,
+      "step": 3810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.186666597645864e-06,
+      "loss": 2.8758,
+      "step": 3811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.18471370428963e-06,
+      "loss": 3.0478,
+      "step": 3812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.182760619434471e-06,
+      "loss": 3.0559,
+      "step": 3813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.180807343396089e-06,
+      "loss": 2.9244,
+      "step": 3814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.178853876490212e-06,
+      "loss": 2.9064,
+      "step": 3815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1769002190326025e-06,
+      "loss": 2.9139,
+      "step": 3816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1749463713390534e-06,
+      "loss": 3.0638,
+      "step": 3817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.172992333725386e-06,
+      "loss": 3.0313,
+      "step": 3818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1710381065074556e-06,
+      "loss": 2.8878,
+      "step": 3819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.169083690001145e-06,
+      "loss": 3.12,
+      "step": 3820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.167129084522371e-06,
+      "loss": 3.104,
+      "step": 3821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.165174290387076e-06,
+      "loss": 3.0324,
+      "step": 3822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.163219307911238e-06,
+      "loss": 2.8594,
+      "step": 3823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.161264137410864e-06,
+      "loss": 3.0807,
+      "step": 3824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.159308779201989e-06,
+      "loss": 2.9933,
+      "step": 3825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.157353233600682e-06,
+      "loss": 3.0675,
+      "step": 3826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.155397500923036e-06,
+      "loss": 3.0378,
+      "step": 3827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.153441581485184e-06,
+      "loss": 3.1575,
+      "step": 3828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.151485475603281e-06,
+      "loss": 2.9359,
+      "step": 3829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.149529183593516e-06,
+      "loss": 2.8285,
+      "step": 3830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.147572705772106e-06,
+      "loss": 3.0202,
+      "step": 3831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.145616042455298e-06,
+      "loss": 2.9138,
+      "step": 3832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.143659193959374e-06,
+      "loss": 3.0003,
+      "step": 3833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.141702160600639e-06,
+      "loss": 3.0482,
+      "step": 3834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.139744942695431e-06,
+      "loss": 2.9933,
+      "step": 3835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1377875405601175e-06,
+      "loss": 3.0307,
+      "step": 3836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.135829954511095e-06,
+      "loss": 3.2262,
+      "step": 3837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.133872184864795e-06,
+      "loss": 2.8896,
+      "step": 3838
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1319142319376716e-06,
+      "loss": 3.0907,
+      "step": 3839
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1299560960462105e-06,
+      "loss": 2.9338,
+      "step": 3840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.127997777506927e-06,
+      "loss": 2.9689,
+      "step": 3841
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.126039276636368e-06,
+      "loss": 2.9758,
+      "step": 3842
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1240805937511104e-06,
+      "loss": 3.1154,
+      "step": 3843
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.122121729167757e-06,
+      "loss": 2.9782,
+      "step": 3844
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.120162683202941e-06,
+      "loss": 3.0266,
+      "step": 3845
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.1182034561733275e-06,
+      "loss": 2.9485,
+      "step": 3846
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.116244048395606e-06,
+      "loss": 3.009,
+      "step": 3847
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.114284460186499e-06,
+      "loss": 3.0375,
+      "step": 3848
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.112324691862761e-06,
+      "loss": 2.9751,
+      "step": 3849
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.110364743741167e-06,
+      "loss": 2.8968,
+      "step": 3850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.108404616138529e-06,
+      "loss": 2.7854,
+      "step": 3851
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.106444309371683e-06,
+      "loss": 2.9328,
+      "step": 3852
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.104483823757499e-06,
+      "loss": 2.9162,
+      "step": 3853
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.102523159612868e-06,
+      "loss": 2.9848,
+      "step": 3854
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.100562317254719e-06,
+      "loss": 2.8764,
+      "step": 3855
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.098601297000005e-06,
+      "loss": 3.1135,
+      "step": 3856
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.096640099165706e-06,
+      "loss": 3.0361,
+      "step": 3857
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.094678724068836e-06,
+      "loss": 3.0344,
+      "step": 3858
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.092717172026432e-06,
+      "loss": 2.8831,
+      "step": 3859
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.090755443355563e-06,
+      "loss": 2.8674,
+      "step": 3860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.088793538373327e-06,
+      "loss": 2.939,
+      "step": 3861
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.086831457396848e-06,
+      "loss": 3.0765,
+      "step": 3862
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.084869200743281e-06,
+      "loss": 2.9791,
+      "step": 3863
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.082906768729807e-06,
+      "loss": 3.0585,
+      "step": 3864
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.080944161673636e-06,
+      "loss": 2.9712,
+      "step": 3865
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.078981379892008e-06,
+      "loss": 3.1926,
+      "step": 3866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.077018423702191e-06,
+      "loss": 2.9445,
+      "step": 3867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.075055293421477e-06,
+      "loss": 3.0617,
+      "step": 3868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.073091989367192e-06,
+      "loss": 2.951,
+      "step": 3869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.071128511856687e-06,
+      "loss": 2.8377,
+      "step": 3870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.069164861207341e-06,
+      "loss": 2.967,
+      "step": 3871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.067201037736562e-06,
+      "loss": 2.879,
+      "step": 3872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.065237041761786e-06,
+      "loss": 2.8447,
+      "step": 3873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.063272873600473e-06,
+      "loss": 3.1023,
+      "step": 3874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.061308533570119e-06,
+      "loss": 3.0915,
+      "step": 3875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.059344021988237e-06,
+      "loss": 3.0437,
+      "step": 3876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.05737933917238e-06,
+      "loss": 2.9962,
+      "step": 3877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.055414485440119e-06,
+      "loss": 3.039,
+      "step": 3878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.053449461109056e-06,
+      "loss": 2.9767,
+      "step": 3879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.0514842664968194e-06,
+      "loss": 2.9367,
+      "step": 3880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.049518901921069e-06,
+      "loss": 2.9512,
+      "step": 3881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.047553367699488e-06,
+      "loss": 3.084,
+      "step": 3882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.045587664149789e-06,
+      "loss": 2.99,
+      "step": 3883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.04362179158971e-06,
+      "loss": 3.0465,
+      "step": 3884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.041655750337019e-06,
+      "loss": 2.9682,
+      "step": 3885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.039689540709507e-06,
+      "loss": 2.9524,
+      "step": 3886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.037723163024998e-06,
+      "loss": 2.9436,
+      "step": 3887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.03575661760134e-06,
+      "loss": 3.0206,
+      "step": 3888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.033789904756409e-06,
+      "loss": 3.0524,
+      "step": 3889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.031823024808107e-06,
+      "loss": 2.9277,
+      "step": 3890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.02985597807436e-06,
+      "loss": 3.0035,
+      "step": 3891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.027888764873129e-06,
+      "loss": 2.8875,
+      "step": 3892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.025921385522395e-06,
+      "loss": 3.1394,
+      "step": 3893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.02395384034017e-06,
+      "loss": 2.8978,
+      "step": 3894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.02198612964449e-06,
+      "loss": 3.0721,
+      "step": 3895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.0200182537534155e-06,
+      "loss": 3.0999,
+      "step": 3896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.018050212985043e-06,
+      "loss": 2.9336,
+      "step": 3897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.016082007657485e-06,
+      "loss": 3.0295,
+      "step": 3898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.014113638088888e-06,
+      "loss": 3.1153,
+      "step": 3899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.01214510459742e-06,
+      "loss": 2.9967,
+      "step": 3900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.010176407501279e-06,
+      "loss": 2.9546,
+      "step": 3901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.008207547118689e-06,
+      "loss": 2.9331,
+      "step": 3902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.0062385237678975e-06,
+      "loss": 2.9496,
+      "step": 3903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.0042693377671835e-06,
+      "loss": 3.0435,
+      "step": 3904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.002299989434845e-06,
+      "loss": 2.9706,
+      "step": 3905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.000330479089212e-06,
+      "loss": 3.0884,
+      "step": 3906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.998360807048642e-06,
+      "loss": 2.955,
+      "step": 3907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.996390973631513e-06,
+      "loss": 3.1542,
+      "step": 3908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.994420979156232e-06,
+      "loss": 2.8527,
+      "step": 3909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.992450823941234e-06,
+      "loss": 3.0929,
+      "step": 3910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.990480508304975e-06,
+      "loss": 2.9138,
+      "step": 3911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.9885100325659394e-06,
+      "loss": 2.7344,
+      "step": 3912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.986539397042641e-06,
+      "loss": 3.0255,
+      "step": 3913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.984568602053615e-06,
+      "loss": 2.9937,
+      "step": 3914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.9825976479174245e-06,
+      "loss": 2.8823,
+      "step": 3915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.980626534952656e-06,
+      "loss": 2.9331,
+      "step": 3916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.978655263477924e-06,
+      "loss": 2.9433,
+      "step": 3917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.976683833811866e-06,
+      "loss": 2.9293,
+      "step": 3918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.97471224627315e-06,
+      "loss": 2.9519,
+      "step": 3919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.972740501180464e-06,
+      "loss": 2.8348,
+      "step": 3920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.970768598852525e-06,
+      "loss": 2.9953,
+      "step": 3921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.968796539608076e-06,
+      "loss": 2.9375,
+      "step": 3922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.966824323765881e-06,
+      "loss": 2.8848,
+      "step": 3923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.964851951644731e-06,
+      "loss": 3.1142,
+      "step": 3924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.962879423563444e-06,
+      "loss": 2.9234,
+      "step": 3925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.960906739840867e-06,
+      "loss": 2.9409,
+      "step": 3926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.958933900795862e-06,
+      "loss": 3.1819,
+      "step": 3927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.956960906747325e-06,
+      "loss": 3.0548,
+      "step": 3928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.954987758014171e-06,
+      "loss": 2.9113,
+      "step": 3929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.953014454915344e-06,
+      "loss": 3.0557,
+      "step": 3930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.951040997769812e-06,
+      "loss": 3.0446,
+      "step": 3931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.949067386896569e-06,
+      "loss": 3.1181,
+      "step": 3932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.94709362261463e-06,
+      "loss": 2.9543,
+      "step": 3933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.945119705243039e-06,
+      "loss": 2.949,
+      "step": 3934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.94314563510086e-06,
+      "loss": 2.9456,
+      "step": 3935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.941171412507188e-06,
+      "loss": 3.0047,
+      "step": 3936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.939197037781139e-06,
+      "loss": 2.9355,
+      "step": 3937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.937222511241852e-06,
+      "loss": 3.1727,
+      "step": 3938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.935247833208496e-06,
+      "loss": 2.8891,
+      "step": 3939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.933273004000256e-06,
+      "loss": 3.029,
+      "step": 3940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.9312980239363504e-06,
+      "loss": 2.9561,
+      "step": 3941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.929322893336017e-06,
+      "loss": 2.7676,
+      "step": 3942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.927347612518518e-06,
+      "loss": 3.0022,
+      "step": 3943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.92537218180314e-06,
+      "loss": 2.9991,
+      "step": 3944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.923396601509197e-06,
+      "loss": 2.8265,
+      "step": 3945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.9214208719560255e-06,
+      "loss": 3.0918,
+      "step": 3946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.919444993462982e-06,
+      "loss": 2.943,
+      "step": 3947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.917468966349452e-06,
+      "loss": 2.8093,
+      "step": 3948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.915492790934845e-06,
+      "loss": 2.9337,
+      "step": 3949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.91351646753859e-06,
+      "loss": 2.8492,
+      "step": 3950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.911539996480145e-06,
+      "loss": 2.9748,
+      "step": 3951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.909563378078989e-06,
+      "loss": 3.0502,
+      "step": 3952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.907586612654628e-06,
+      "loss": 3.0716,
+      "step": 3953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.905609700526587e-06,
+      "loss": 3.0755,
+      "step": 3954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.903632642014415e-06,
+      "loss": 2.9886,
+      "step": 3955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.901655437437691e-06,
+      "loss": 2.8837,
+      "step": 3956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.899678087116011e-06,
+      "loss": 2.9342,
+      "step": 3957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.897700591368998e-06,
+      "loss": 3.0581,
+      "step": 3958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.895722950516297e-06,
+      "loss": 2.973,
+      "step": 3959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.893745164877576e-06,
+      "loss": 3.0266,
+      "step": 3960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.891767234772528e-06,
+      "loss": 2.9316,
+      "step": 3961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.88978916052087e-06,
+      "loss": 2.815,
+      "step": 3962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.887810942442338e-06,
+      "loss": 3.0973,
+      "step": 3963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8858325808566955e-06,
+      "loss": 2.9274,
+      "step": 3964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.88385407608373e-06,
+      "loss": 3.024,
+      "step": 3965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.881875428443247e-06,
+      "loss": 2.9901,
+      "step": 3966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.879896638255081e-06,
+      "loss": 2.8732,
+      "step": 3967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.877917705839082e-06,
+      "loss": 2.9205,
+      "step": 3968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.875938631515132e-06,
+      "loss": 3.0916,
+      "step": 3969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.87395941560313e-06,
+      "loss": 3.0108,
+      "step": 3970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.871980058423e-06,
+      "loss": 3.0861,
+      "step": 3971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.870000560294687e-06,
+      "loss": 3.0429,
+      "step": 3972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.868020921538164e-06,
+      "loss": 3.149,
+      "step": 3973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.866041142473415e-06,
+      "loss": 2.9378,
+      "step": 3974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.864061223420463e-06,
+      "loss": 2.8662,
+      "step": 3975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.86208116469934e-06,
+      "loss": 3.0961,
+      "step": 3976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.860100966630108e-06,
+      "loss": 2.8256,
+      "step": 3977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.858120629532847e-06,
+      "loss": 3.1584,
+      "step": 3978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.856140153727664e-06,
+      "loss": 2.975,
+      "step": 3979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.854159539534685e-06,
+      "loss": 3.0058,
+      "step": 3980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.852178787274061e-06,
+      "loss": 2.9936,
+      "step": 3981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8501978972659605e-06,
+      "loss": 2.9062,
+      "step": 3982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8482168698305805e-06,
+      "loss": 2.9317,
+      "step": 3983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8462357052881355e-06,
+      "loss": 2.9459,
+      "step": 3984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.844254403958866e-06,
+      "loss": 2.9138,
+      "step": 3985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.842272966163031e-06,
+      "loss": 2.9228,
+      "step": 3986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.840291392220914e-06,
+      "loss": 3.0179,
+      "step": 3987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.838309682452818e-06,
+      "loss": 2.9908,
+      "step": 3988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.83632783717907e-06,
+      "loss": 3.139,
+      "step": 3989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.834345856720021e-06,
+      "loss": 3.0269,
+      "step": 3990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.832363741396039e-06,
+      "loss": 3.0636,
+      "step": 3991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.830381491527517e-06,
+      "loss": 3.0423,
+      "step": 3992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.828399107434868e-06,
+      "loss": 2.8651,
+      "step": 3993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.826416589438527e-06,
+      "loss": 3.023,
+      "step": 3994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8244339378589535e-06,
+      "loss": 3.0621,
+      "step": 3995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.822451153016624e-06,
+      "loss": 2.8763,
+      "step": 3996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.820468235232042e-06,
+      "loss": 2.9307,
+      "step": 3997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.818485184825727e-06,
+      "loss": 2.9328,
+      "step": 3998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.816502002118221e-06,
+      "loss": 2.944,
+      "step": 3999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.814518687430091e-06,
+      "loss": 2.9047,
+      "step": 4000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.812535241081922e-06,
+      "loss": 2.9442,
+      "step": 4001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.810551663394321e-06,
+      "loss": 2.9324,
+      "step": 4002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.808567954687919e-06,
+      "loss": 3.001,
+      "step": 4003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.806584115283363e-06,
+      "loss": 2.8597,
+      "step": 4004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8046001455013235e-06,
+      "loss": 3.0387,
+      "step": 4005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.802616045662495e-06,
+      "loss": 3.0127,
+      "step": 4006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.800631816087588e-06,
+      "loss": 2.9829,
+      "step": 4007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.798647457097337e-06,
+      "loss": 3.0384,
+      "step": 4008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.796662969012498e-06,
+      "loss": 3.0945,
+      "step": 4009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.794678352153845e-06,
+      "loss": 2.9365,
+      "step": 4010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7926936068421756e-06,
+      "loss": 2.8802,
+      "step": 4011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.790708733398307e-06,
+      "loss": 2.9758,
+      "step": 4012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.788723732143076e-06,
+      "loss": 3.1039,
+      "step": 4013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.786738603397343e-06,
+      "loss": 2.9249,
+      "step": 4014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.784753347481984e-06,
+      "loss": 3.0538,
+      "step": 4015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.782767964717904e-06,
+      "loss": 2.9746,
+      "step": 4016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.78078245542602e-06,
+      "loss": 3.1398,
+      "step": 4017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.778796819927271e-06,
+      "loss": 2.9803,
+      "step": 4018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.776811058542622e-06,
+      "loss": 2.9551,
+      "step": 4019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.774825171593053e-06,
+      "loss": 2.7815,
+      "step": 4020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.772839159399563e-06,
+      "loss": 3.033,
+      "step": 4021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7708530222831785e-06,
+      "loss": 2.8837,
+      "step": 4022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7688667605649384e-06,
+      "loss": 3.0125,
+      "step": 4023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.766880374565909e-06,
+      "loss": 2.9103,
+      "step": 4024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.764893864607169e-06,
+      "loss": 2.9273,
+      "step": 4025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.762907231009822e-06,
+      "loss": 2.9832,
+      "step": 4026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7609204740949885e-06,
+      "loss": 2.999,
+      "step": 4027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.758933594183815e-06,
+      "loss": 3.0148,
+      "step": 4028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.756946591597462e-06,
+      "loss": 2.9408,
+      "step": 4029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7549594666571105e-06,
+      "loss": 3.1041,
+      "step": 4030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.752972219683965e-06,
+      "loss": 3.0184,
+      "step": 4031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.7509848509992435e-06,
+      "loss": 3.0028,
+      "step": 4032
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.748997360924189e-06,
+      "loss": 2.9438,
+      "step": 4033
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.747009749780062e-06,
+      "loss": 2.8425,
+      "step": 4034
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.745022017888145e-06,
+      "loss": 2.8314,
+      "step": 4035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.743034165569737e-06,
+      "loss": 3.012,
+      "step": 4036
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.741046193146157e-06,
+      "loss": 2.983,
+      "step": 4037
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.739058100938743e-06,
+      "loss": 2.9679,
+      "step": 4038
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.737069889268855e-06,
+      "loss": 2.9773,
+      "step": 4039
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.735081558457869e-06,
+      "loss": 2.8802,
+      "step": 4040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.733093108827182e-06,
+      "loss": 3.0584,
+      "step": 4041
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.731104540698213e-06,
+      "loss": 2.8701,
+      "step": 4042
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.729115854392392e-06,
+      "loss": 3.029,
+      "step": 4043
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.727127050231177e-06,
+      "loss": 3.0764,
+      "step": 4044
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.7251381285360404e-06,
+      "loss": 3.0807,
+      "step": 4045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.723149089628475e-06,
+      "loss": 3.1807,
+      "step": 4046
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.72115993382999e-06,
+      "loss": 2.937,
+      "step": 4047
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.719170661462119e-06,
+      "loss": 2.8905,
+      "step": 4048
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.717181272846409e-06,
+      "loss": 3.1206,
+      "step": 4049
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.715191768304427e-06,
+      "loss": 2.977,
+      "step": 4050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.71320214815776e-06,
+      "loss": 2.8845,
+      "step": 4051
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.711212412728014e-06,
+      "loss": 3.13,
+      "step": 4052
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.709222562336812e-06,
+      "loss": 2.984,
+      "step": 4053
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.707232597305798e-06,
+      "loss": 3.0652,
+      "step": 4054
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.705242517956633e-06,
+      "loss": 2.8913,
+      "step": 4055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.703252324610995e-06,
+      "loss": 2.8401,
+      "step": 4056
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.701262017590581e-06,
+      "loss": 2.8425,
+      "step": 4057
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.69927159721711e-06,
+      "loss": 2.8998,
+      "step": 4058
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.697281063812313e-06,
+      "loss": 3.0477,
+      "step": 4059
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6952904176979465e-06,
+      "loss": 3.059,
+      "step": 4060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.69329965919578e-06,
+      "loss": 2.9576,
+      "step": 4061
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.691308788627602e-06,
+      "loss": 2.9947,
+      "step": 4062
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.689317806315221e-06,
+      "loss": 2.9749,
+      "step": 4063
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.687326712580463e-06,
+      "loss": 2.8673,
+      "step": 4064
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6853355077451685e-06,
+      "loss": 2.9113,
+      "step": 4065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6833441921312025e-06,
+      "loss": 3.1255,
+      "step": 4066
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.68135276606044e-06,
+      "loss": 3.068,
+      "step": 4067
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.679361229854784e-06,
+      "loss": 3.031,
+      "step": 4068
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6773695838361455e-06,
+      "loss": 2.7237,
+      "step": 4069
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.675377828326457e-06,
+      "loss": 2.9648,
+      "step": 4070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.67338596364767e-06,
+      "loss": 3.0836,
+      "step": 4071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.671393990121751e-06,
+      "loss": 2.8838,
+      "step": 4072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.669401908070688e-06,
+      "loss": 2.9402,
+      "step": 4073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6674097178164836e-06,
+      "loss": 2.9387,
+      "step": 4074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.665417419681157e-06,
+      "loss": 3.0605,
+      "step": 4075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.663425013986748e-06,
+      "loss": 2.9839,
+      "step": 4076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.661432501055311e-06,
+      "loss": 3.0763,
+      "step": 4077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.659439881208918e-06,
+      "loss": 2.947,
+      "step": 4078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6574471547696616e-06,
+      "loss": 3.0394,
+      "step": 4079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.655454322059648e-06,
+      "loss": 2.959,
+      "step": 4080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.653461383401002e-06,
+      "loss": 2.9821,
+      "step": 4081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.651468339115863e-06,
+      "loss": 3.1089,
+      "step": 4082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.649475189526392e-06,
+      "loss": 3.0459,
+      "step": 4083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.647481934954765e-06,
+      "loss": 3.094,
+      "step": 4084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.645488575723172e-06,
+      "loss": 2.9868,
+      "step": 4085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.643495112153827e-06,
+      "loss": 2.9321,
+      "step": 4086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.641501544568951e-06,
+      "loss": 3.0927,
+      "step": 4087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.639507873290792e-06,
+      "loss": 2.8713,
+      "step": 4088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.637514098641609e-06,
+      "loss": 2.8647,
+      "step": 4089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.635520220943677e-06,
+      "loss": 3.0362,
+      "step": 4090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6335262405192895e-06,
+      "loss": 3.0839,
+      "step": 4091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.631532157690758e-06,
+      "loss": 3.0514,
+      "step": 4092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.62953797278041e-06,
+      "loss": 2.9614,
+      "step": 4093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.627543686110586e-06,
+      "loss": 2.9528,
+      "step": 4094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.625549298003648e-06,
+      "loss": 2.8899,
+      "step": 4095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.62355480878197e-06,
+      "loss": 3.0343,
+      "step": 4096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.621560218767943e-06,
+      "loss": 3.0526,
+      "step": 4097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.619565528283979e-06,
+      "loss": 3.0029,
+      "step": 4098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.617570737652501e-06,
+      "loss": 2.9902,
+      "step": 4099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.615575847195951e-06,
+      "loss": 3.0465,
+      "step": 4100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.613580857236786e-06,
+      "loss": 2.8612,
+      "step": 4101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.611585768097476e-06,
+      "loss": 2.9401,
+      "step": 4102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.609590580100515e-06,
+      "loss": 3.0265,
+      "step": 4103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.607595293568404e-06,
+      "loss": 3.0725,
+      "step": 4104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.605599908823668e-06,
+      "loss": 2.8666,
+      "step": 4105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.6036044261888415e-06,
+      "loss": 2.982,
+      "step": 4106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.601608845986478e-06,
+      "loss": 2.9432,
+      "step": 4107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.599613168539147e-06,
+      "loss": 2.8787,
+      "step": 4108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.597617394169431e-06,
+      "loss": 2.8897,
+      "step": 4109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.595621523199931e-06,
+      "loss": 2.8128,
+      "step": 4110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.5936255559532625e-06,
+      "loss": 3.0997,
+      "step": 4111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.5916294927520575e-06,
+      "loss": 3.1575,
+      "step": 4112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.5896333339189625e-06,
+      "loss": 2.9086,
+      "step": 4113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.587637079776639e-06,
+      "loss": 3.0985,
+      "step": 4114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.585640730647765e-06,
+      "loss": 2.8895,
+      "step": 4115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.583644286855031e-06,
+      "loss": 3.0167,
+      "step": 4116
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.581647748721149e-06,
+      "loss": 2.9703,
+      "step": 4117
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.57965111656884e-06,
+      "loss": 2.9785,
+      "step": 4118
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.577654390720844e-06,
+      "loss": 3.0288,
+      "step": 4119
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.575657571499914e-06,
+      "loss": 2.8859,
+      "step": 4120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5736606592288185e-06,
+      "loss": 2.972,
+      "step": 4121
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.571663654230344e-06,
+      "loss": 2.9967,
+      "step": 4122
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.569666556827285e-06,
+      "loss": 2.7457,
+      "step": 4123
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.56766936734246e-06,
+      "loss": 2.9331,
+      "step": 4124
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.565672086098694e-06,
+      "loss": 2.8798,
+      "step": 4125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.563674713418833e-06,
+      "loss": 2.846,
+      "step": 4126
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.561677249625734e-06,
+      "loss": 3.1226,
+      "step": 4127
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.559679695042271e-06,
+      "loss": 2.9274,
+      "step": 4128
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.557682049991331e-06,
+      "loss": 3.1159,
+      "step": 4129
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5556843147958185e-06,
+      "loss": 3.2252,
+      "step": 4130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5536864897786455e-06,
+      "loss": 3.0457,
+      "step": 4131
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.551688575262749e-06,
+      "loss": 2.8815,
+      "step": 4132
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.549690571571072e-06,
+      "loss": 3.0534,
+      "step": 4133
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.547692479026576e-06,
+      "loss": 2.9851,
+      "step": 4134
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5456942979522345e-06,
+      "loss": 3.0468,
+      "step": 4135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.543696028671038e-06,
+      "loss": 2.9336,
+      "step": 4136
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.541697671505988e-06,
+      "loss": 3.12,
+      "step": 4137
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.539699226780103e-06,
+      "loss": 2.9142,
+      "step": 4138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.537700694816416e-06,
+      "loss": 2.9274,
+      "step": 4139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.535702075937971e-06,
+      "loss": 2.9271,
+      "step": 4140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.533703370467827e-06,
+      "loss": 3.0564,
+      "step": 4141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5317045787290576e-06,
+      "loss": 3.0009,
+      "step": 4142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5297057010447555e-06,
+      "loss": 2.9576,
+      "step": 4143
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.527706737738017e-06,
+      "loss": 2.9828,
+      "step": 4144
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.52570768913196e-06,
+      "loss": 3.1199,
+      "step": 4145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.523708555549712e-06,
+      "loss": 2.9197,
+      "step": 4146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.521709337314417e-06,
+      "loss": 3.0519,
+      "step": 4147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.519710034749232e-06,
+      "loss": 2.8689,
+      "step": 4148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.5177106481773266e-06,
+      "loss": 3.1244,
+      "step": 4149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.515711177921885e-06,
+      "loss": 2.8792,
+      "step": 4150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.513711624306106e-06,
+      "loss": 2.8565,
+      "step": 4151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.511711987653198e-06,
+      "loss": 2.8712,
+      "step": 4152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.509712268286388e-06,
+      "loss": 2.9646,
+      "step": 4153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.50771246652891e-06,
+      "loss": 3.0881,
+      "step": 4154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.505712582704019e-06,
+      "loss": 3.1329,
+      "step": 4155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.503712617134976e-06,
+      "loss": 2.9872,
+      "step": 4156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.501712570145061e-06,
+      "loss": 3.0976,
+      "step": 4157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.499712442057563e-06,
+      "loss": 3.0134,
+      "step": 4158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.497712233195788e-06,
+      "loss": 2.8695,
+      "step": 4159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.495711943883048e-06,
+      "loss": 2.9343,
+      "step": 4160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.493711574442678e-06,
+      "loss": 3.0245,
+      "step": 4161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.491711125198018e-06,
+      "loss": 2.8735,
+      "step": 4162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.489710596472425e-06,
+      "loss": 3.0645,
+      "step": 4163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.487709988589266e-06,
+      "loss": 2.8596,
+      "step": 4164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.4857093018719235e-06,
+      "loss": 2.9761,
+      "step": 4165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.48370853664379e-06,
+      "loss": 3.0473,
+      "step": 4166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.481707693228276e-06,
+      "loss": 3.1455,
+      "step": 4167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.479706771948796e-06,
+      "loss": 2.974,
+      "step": 4168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.477705773128785e-06,
+      "loss": 2.8956,
+      "step": 4169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.475704697091686e-06,
+      "loss": 3.0545,
+      "step": 4170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.473703544160957e-06,
+      "loss": 3.0313,
+      "step": 4171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.4717023146600676e-06,
+      "loss": 3.0283,
+      "step": 4172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.4697010089125e-06,
+      "loss": 2.8697,
+      "step": 4173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.467699627241745e-06,
+      "loss": 3.0124,
+      "step": 4174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.465698169971312e-06,
+      "loss": 2.9382,
+      "step": 4175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.46369663742472e-06,
+      "loss": 2.9481,
+      "step": 4176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.461695029925499e-06,
+      "loss": 3.0002,
+      "step": 4177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.45969334779719e-06,
+      "loss": 2.8756,
+      "step": 4178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.457691591363351e-06,
+      "loss": 3.0066,
+      "step": 4179
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.4556897609475455e-06,
+      "loss": 2.9821,
+      "step": 4180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.453687856873355e-06,
+      "loss": 2.8796,
+      "step": 4181
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.451685879464371e-06,
+      "loss": 3.0569,
+      "step": 4182
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.449683829044194e-06,
+      "loss": 2.9582,
+      "step": 4183
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.44768170593644e-06,
+      "loss": 2.9436,
+      "step": 4184
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.445679510464735e-06,
+      "loss": 2.9746,
+      "step": 4185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.443677242952715e-06,
+      "loss": 2.9343,
+      "step": 4186
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.44167490372403e-06,
+      "loss": 2.9117,
+      "step": 4187
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.439672493102345e-06,
+      "loss": 3.0818,
+      "step": 4188
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.437670011411329e-06,
+      "loss": 2.981,
+      "step": 4189
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.435667458974666e-06,
+      "loss": 2.9698,
+      "step": 4190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.433664836116054e-06,
+      "loss": 3.0162,
+      "step": 4191
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.431662143159199e-06,
+      "loss": 3.0485,
+      "step": 4192
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.429659380427816e-06,
+      "loss": 2.8888,
+      "step": 4193
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.427656548245641e-06,
+      "loss": 3.0328,
+      "step": 4194
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.4256536469364095e-06,
+      "loss": 2.8533,
+      "step": 4195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.423650676823877e-06,
+      "loss": 2.9268,
+      "step": 4196
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.421647638231805e-06,
+      "loss": 3.0609,
+      "step": 4197
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.419644531483967e-06,
+      "loss": 2.9963,
+      "step": 4198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.417641356904149e-06,
+      "loss": 3.021,
+      "step": 4199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.415638114816147e-06,
+      "loss": 2.9066,
+      "step": 4200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.413634805543769e-06,
+      "loss": 2.873,
+      "step": 4201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.4116314294108325e-06,
+      "loss": 3.0227,
+      "step": 4202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.409627986741167e-06,
+      "loss": 3.0355,
+      "step": 4203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.40762447785861e-06,
+      "loss": 3.0341,
+      "step": 4204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.405620903087012e-06,
+      "loss": 3.1058,
+      "step": 4205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.403617262750234e-06,
+      "loss": 2.9458,
+      "step": 4206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.401613557172152e-06,
+      "loss": 3.091,
+      "step": 4207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.399609786676642e-06,
+      "loss": 3.0782,
+      "step": 4208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.397605951587597e-06,
+      "loss": 2.9765,
+      "step": 4209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.395602052228923e-06,
+      "loss": 3.0552,
+      "step": 4210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.393598088924532e-06,
+      "loss": 2.995,
+      "step": 4211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3915940619983475e-06,
+      "loss": 3.0518,
+      "step": 4212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3895899717743035e-06,
+      "loss": 2.9203,
+      "step": 4213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.387585818576344e-06,
+      "loss": 2.9103,
+      "step": 4214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.385581602728424e-06,
+      "loss": 2.9034,
+      "step": 4215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.383577324554508e-06,
+      "loss": 3.0012,
+      "step": 4216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.381572984378572e-06,
+      "loss": 2.912,
+      "step": 4217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3795685825245945e-06,
+      "loss": 2.9332,
+      "step": 4218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.377564119316576e-06,
+      "loss": 2.9957,
+      "step": 4219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.375559595078521e-06,
+      "loss": 3.0488,
+      "step": 4220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.373555010134442e-06,
+      "loss": 2.854,
+      "step": 4221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.371550364808362e-06,
+      "loss": 2.9903,
+      "step": 4222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.369545659424317e-06,
+      "loss": 3.0545,
+      "step": 4223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.367540894306349e-06,
+      "loss": 2.9576,
+      "step": 4224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.365536069778512e-06,
+      "loss": 2.9274,
+      "step": 4225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3635311861648705e-06,
+      "loss": 3.196,
+      "step": 4226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3615262437894955e-06,
+      "loss": 2.9957,
+      "step": 4227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.359521242976469e-06,
+      "loss": 2.9034,
+      "step": 4228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.35751618404988e-06,
+      "loss": 2.9311,
+      "step": 4229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.355511067333834e-06,
+      "loss": 2.9691,
+      "step": 4230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.353505893152438e-06,
+      "loss": 2.9562,
+      "step": 4231
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.351500661829811e-06,
+      "loss": 2.906,
+      "step": 4232
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.349495373690085e-06,
+      "loss": 2.9889,
+      "step": 4233
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.347490029057393e-06,
+      "loss": 2.8547,
+      "step": 4234
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.345484628255886e-06,
+      "loss": 2.9567,
+      "step": 4235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.343479171609718e-06,
+      "loss": 3.1969,
+      "step": 4236
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.341473659443056e-06,
+      "loss": 2.9537,
+      "step": 4237
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.33946809208007e-06,
+      "loss": 2.9928,
+      "step": 4238
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.337462469844947e-06,
+      "loss": 3.1151,
+      "step": 4239
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.335456793061877e-06,
+      "loss": 2.8945,
+      "step": 4240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.333451062055063e-06,
+      "loss": 3.0787,
+      "step": 4241
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.331445277148711e-06,
+      "loss": 3.0283,
+      "step": 4242
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3294394386670404e-06,
+      "loss": 2.9255,
+      "step": 4243
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.3274335469342786e-06,
+      "loss": 2.9669,
+      "step": 4244
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.32542760227466e-06,
+      "loss": 3.0209,
+      "step": 4245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.32342160501243e-06,
+      "loss": 2.9052,
+      "step": 4246
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.321415555471839e-06,
+      "loss": 2.9558,
+      "step": 4247
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.319409453977149e-06,
+      "loss": 2.8837,
+      "step": 4248
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.317403300852632e-06,
+      "loss": 3.1619,
+      "step": 4249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.315397096422561e-06,
+      "loss": 3.0612,
+      "step": 4250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.313390841011223e-06,
+      "loss": 2.9657,
+      "step": 4251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.311384534942914e-06,
+      "loss": 3.1005,
+      "step": 4252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.309378178541935e-06,
+      "loss": 3.0103,
+      "step": 4253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.307371772132598e-06,
+      "loss": 2.9187,
+      "step": 4254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.30536531603922e-06,
+      "loss": 3.0912,
+      "step": 4255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.303358810586129e-06,
+      "loss": 2.9758,
+      "step": 4256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.301352256097657e-06,
+      "loss": 2.9363,
+      "step": 4257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.299345652898149e-06,
+      "loss": 2.9081,
+      "step": 4258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.297339001311955e-06,
+      "loss": 2.9824,
+      "step": 4259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.295332301663433e-06,
+      "loss": 2.9142,
+      "step": 4260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.293325554276949e-06,
+      "loss": 2.8849,
+      "step": 4261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.291318759476877e-06,
+      "loss": 2.8082,
+      "step": 4262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.289311917587596e-06,
+      "loss": 2.9474,
+      "step": 4263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.2873050289334974e-06,
+      "loss": 2.942,
+      "step": 4264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.2852980938389774e-06,
+      "loss": 2.9353,
+      "step": 4265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.283291112628441e-06,
+      "loss": 2.942,
+      "step": 4266
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.281284085626298e-06,
+      "loss": 3.0373,
+      "step": 4267
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.279277013156967e-06,
+      "loss": 2.9617,
+      "step": 4268
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.277269895544875e-06,
+      "loss": 3.0362,
+      "step": 4269
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.275262733114456e-06,
+      "loss": 3.1755,
+      "step": 4270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.273255526190151e-06,
+      "loss": 3.1269,
+      "step": 4271
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.271248275096408e-06,
+      "loss": 2.8191,
+      "step": 4272
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.26924098015768e-06,
+      "loss": 2.9982,
+      "step": 4273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.267233641698432e-06,
+      "loss": 2.9394,
+      "step": 4274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.26522626004313e-06,
+      "loss": 2.9965,
+      "step": 4275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.2632188355162536e-06,
+      "loss": 3.0669,
+      "step": 4276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.261211368442285e-06,
+      "loss": 2.9625,
+      "step": 4277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.259203859145712e-06,
+      "loss": 2.9318,
+      "step": 4278
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.257196307951035e-06,
+      "loss": 2.967,
+      "step": 4279
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.255188715182755e-06,
+      "loss": 2.8596,
+      "step": 4280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.2531810811653834e-06,
+      "loss": 2.9696,
+      "step": 4281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.251173406223436e-06,
+      "loss": 3.0579,
+      "step": 4282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.249165690681438e-06,
+      "loss": 2.9395,
+      "step": 4283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.24715793486392e-06,
+      "loss": 3.1079,
+      "step": 4284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.245150139095417e-06,
+      "loss": 2.9547,
+      "step": 4285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.243142303700474e-06,
+      "loss": 3.0087,
+      "step": 4286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.241134429003638e-06,
+      "loss": 3.046,
+      "step": 4287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.239126515329467e-06,
+      "loss": 2.938,
+      "step": 4288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2371185630025215e-06,
+      "loss": 3.0857,
+      "step": 4289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2351105723473715e-06,
+      "loss": 3.0839,
+      "step": 4290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.233102543688593e-06,
+      "loss": 2.9422,
+      "step": 4291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.231094477350762e-06,
+      "loss": 3.0163,
+      "step": 4292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.229086373658471e-06,
+      "loss": 2.8826,
+      "step": 4293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.22707823293631e-06,
+      "loss": 2.7796,
+      "step": 4294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.225070055508876e-06,
+      "loss": 2.9787,
+      "step": 4295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.223061841700778e-06,
+      "loss": 2.9713,
+      "step": 4296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.221053591836624e-06,
+      "loss": 2.8309,
+      "step": 4297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.219045306241033e-06,
+      "loss": 2.8549,
+      "step": 4298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.217036985238625e-06,
+      "loss": 2.9296,
+      "step": 4299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.21502862915403e-06,
+      "loss": 2.7515,
+      "step": 4300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.213020238311879e-06,
+      "loss": 2.8517,
+      "step": 4301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.211011813036811e-06,
+      "loss": 3.1593,
+      "step": 4302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.209003353653474e-06,
+      "loss": 3.0834,
+      "step": 4303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2069948604865195e-06,
+      "loss": 3.0438,
+      "step": 4304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2049863338606e-06,
+      "loss": 2.9492,
+      "step": 4305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2029777741003785e-06,
+      "loss": 2.8325,
+      "step": 4306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.2009691815305185e-06,
+      "loss": 2.9013,
+      "step": 4307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.198960556475697e-06,
+      "loss": 2.8681,
+      "step": 4308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.196951899260587e-06,
+      "loss": 3.0249,
+      "step": 4309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.194943210209874e-06,
+      "loss": 2.9072,
+      "step": 4310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1929344896482436e-06,
+      "loss": 2.8979,
+      "step": 4311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.190925737900389e-06,
+      "loss": 3.0051,
+      "step": 4312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.188916955291008e-06,
+      "loss": 2.8925,
+      "step": 4313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.186908142144803e-06,
+      "loss": 2.8616,
+      "step": 4314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.184899298786482e-06,
+      "loss": 3.1142,
+      "step": 4315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.182890425540757e-06,
+      "loss": 3.0061,
+      "step": 4316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.180881522732346e-06,
+      "loss": 2.8751,
+      "step": 4317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.178872590685971e-06,
+      "loss": 3.0279,
+      "step": 4318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.176863629726358e-06,
+      "loss": 3.0147,
+      "step": 4319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1748546401782394e-06,
+      "loss": 3.0492,
+      "step": 4320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.172845622366351e-06,
+      "loss": 2.9946,
+      "step": 4321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.170836576615432e-06,
+      "loss": 2.9158,
+      "step": 4322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.168827503250231e-06,
+      "loss": 3.0144,
+      "step": 4323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1668184025954945e-06,
+      "loss": 3.0948,
+      "step": 4324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.164809274975979e-06,
+      "loss": 3.0367,
+      "step": 4325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.162800120716442e-06,
+      "loss": 3.0577,
+      "step": 4326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.160790940141645e-06,
+      "loss": 2.8953,
+      "step": 4327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.158781733576355e-06,
+      "loss": 2.9604,
+      "step": 4328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.156772501345346e-06,
+      "loss": 2.9822,
+      "step": 4329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.154763243773392e-06,
+      "loss": 3.0037,
+      "step": 4330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.152753961185272e-06,
+      "loss": 2.8864,
+      "step": 4331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.150744653905767e-06,
+      "loss": 3.0245,
+      "step": 4332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.148735322259669e-06,
+      "loss": 3.0561,
+      "step": 4333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1467259665717665e-06,
+      "loss": 2.9841,
+      "step": 4334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.144716587166857e-06,
+      "loss": 2.9089,
+      "step": 4335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.142707184369737e-06,
+      "loss": 2.9276,
+      "step": 4336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.140697758505212e-06,
+      "loss": 2.9863,
+      "step": 4337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1386883098980865e-06,
+      "loss": 2.9211,
+      "step": 4338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.136678838873172e-06,
+      "loss": 2.9637,
+      "step": 4339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.134669345755283e-06,
+      "loss": 2.8716,
+      "step": 4340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.132659830869235e-06,
+      "loss": 3.0206,
+      "step": 4341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.130650294539851e-06,
+      "loss": 2.9782,
+      "step": 4342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.128640737091956e-06,
+      "loss": 2.9081,
+      "step": 4343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.126631158850374e-06,
+      "loss": 2.971,
+      "step": 4344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.124621560139942e-06,
+      "loss": 3.0312,
+      "step": 4345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.122611941285489e-06,
+      "loss": 3.0109,
+      "step": 4346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.120602302611857e-06,
+      "loss": 2.9023,
+      "step": 4347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.118592644443885e-06,
+      "loss": 2.9806,
+      "step": 4348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.116582967106418e-06,
+      "loss": 2.9446,
+      "step": 4349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.1145732709243035e-06,
+      "loss": 2.8903,
+      "step": 4350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.11256355622239e-06,
+      "loss": 2.9307,
+      "step": 4351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.110553823325532e-06,
+      "loss": 2.9401,
+      "step": 4352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.108544072558587e-06,
+      "loss": 2.9352,
+      "step": 4353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.106534304246412e-06,
+      "loss": 2.9926,
+      "step": 4354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.104524518713871e-06,
+      "loss": 2.9647,
+      "step": 4355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.102514716285828e-06,
+      "loss": 3.1297,
+      "step": 4356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.100504897287151e-06,
+      "loss": 3.0762,
+      "step": 4357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.098495062042709e-06,
+      "loss": 2.8609,
+      "step": 4358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.096485210877374e-06,
+      "loss": 2.9441,
+      "step": 4359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.094475344116025e-06,
+      "loss": 2.857,
+      "step": 4360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.0924654620835355e-06,
+      "loss": 3.0103,
+      "step": 4361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.09045556510479e-06,
+      "loss": 3.0105,
+      "step": 4362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.08844565350467e-06,
+      "loss": 3.03,
+      "step": 4363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.08643572760806e-06,
+      "loss": 3.0095,
+      "step": 4364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0844257877398465e-06,
+      "loss": 2.9993,
+      "step": 4365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.08241583422492e-06,
+      "loss": 3.0485,
+      "step": 4366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.080405867388173e-06,
+      "loss": 2.9771,
+      "step": 4367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0783958875545e-06,
+      "loss": 3.0037,
+      "step": 4368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.076385895048798e-06,
+      "loss": 3.1455,
+      "step": 4369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0743758901959616e-06,
+      "loss": 2.9214,
+      "step": 4370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0723658733208935e-06,
+      "loss": 3.0049,
+      "step": 4371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.070355844748496e-06,
+      "loss": 2.9907,
+      "step": 4372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0683458048036724e-06,
+      "loss": 3.09,
+      "step": 4373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.06633575381133e-06,
+      "loss": 3.0023,
+      "step": 4374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.064325692096375e-06,
+      "loss": 3.064,
+      "step": 4375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.062315619983717e-06,
+      "loss": 2.9049,
+      "step": 4376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.060305537798268e-06,
+      "loss": 2.9072,
+      "step": 4377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.05829544586494e-06,
+      "loss": 2.8635,
+      "step": 4378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.056285344508647e-06,
+      "loss": 3.0199,
+      "step": 4379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0542752340543064e-06,
+      "loss": 2.9073,
+      "step": 4380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.052265114826833e-06,
+      "loss": 3.0092,
+      "step": 4381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.050254987151149e-06,
+      "loss": 3.0309,
+      "step": 4382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.048244851352171e-06,
+      "loss": 3.1331,
+      "step": 4383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.046234707754822e-06,
+      "loss": 2.9222,
+      "step": 4384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.044224556684025e-06,
+      "loss": 2.8896,
+      "step": 4385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.042214398464703e-06,
+      "loss": 2.8614,
+      "step": 4386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.040204233421782e-06,
+      "loss": 3.0735,
+      "step": 4387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.038194061880187e-06,
+      "loss": 3.069,
+      "step": 4388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.036183884164847e-06,
+      "loss": 3.0345,
+      "step": 4389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.034173700600688e-06,
+      "loss": 3.1081,
+      "step": 4390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.03216351151264e-06,
+      "loss": 2.9368,
+      "step": 4391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.030153317225632e-06,
+      "loss": 2.9668,
+      "step": 4392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.028143118064598e-06,
+      "loss": 2.8936,
+      "step": 4393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.026132914354468e-06,
+      "loss": 2.9458,
+      "step": 4394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.024122706420171e-06,
+      "loss": 2.8931,
+      "step": 4395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.022112494586646e-06,
+      "loss": 2.8889,
+      "step": 4396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.020102279178824e-06,
+      "loss": 2.9277,
+      "step": 4397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.018092060521637e-06,
+      "loss": 2.9428,
+      "step": 4398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.0160818389400225e-06,
+      "loss": 2.9899,
+      "step": 4399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.014071614758914e-06,
+      "loss": 2.9402,
+      "step": 4400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.01206138830325e-06,
+      "loss": 3.0359,
+      "step": 4401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.010051159897965e-06,
+      "loss": 3.0024,
+      "step": 4402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.008040929867995e-06,
+      "loss": 2.8671,
+      "step": 4403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.006030698538275e-06,
+      "loss": 2.9201,
+      "step": 4404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.004020466233746e-06,
+      "loss": 2.9653,
+      "step": 4405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.002010233279342e-06,
+      "loss": 2.9298,
+      "step": 4406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 5e-06,
+      "loss": 3.0232,
+      "step": 4407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.99798976672066e-06,
+      "loss": 2.9747,
+      "step": 4408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.995979533766257e-06,
+      "loss": 3.0202,
+      "step": 4409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.993969301461726e-06,
+      "loss": 2.9366,
+      "step": 4410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.991959070132007e-06,
+      "loss": 3.0293,
+      "step": 4411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.989948840102035e-06,
+      "loss": 2.8018,
+      "step": 4412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9879386116967495e-06,
+      "loss": 2.9277,
+      "step": 4413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.985928385241086e-06,
+      "loss": 2.9099,
+      "step": 4414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.983918161059979e-06,
+      "loss": 2.8731,
+      "step": 4415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.981907939478365e-06,
+      "loss": 3.144,
+      "step": 4416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.979897720821178e-06,
+      "loss": 2.9115,
+      "step": 4417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.977887505413356e-06,
+      "loss": 2.9387,
+      "step": 4418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.975877293579829e-06,
+      "loss": 3.0015,
+      "step": 4419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9738670856455355e-06,
+      "loss": 3.1076,
+      "step": 4420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.971856881935404e-06,
+      "loss": 2.9457,
+      "step": 4421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.969846682774368e-06,
+      "loss": 3.0823,
+      "step": 4422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.967836488487361e-06,
+      "loss": 2.9629,
+      "step": 4423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.965826299399313e-06,
+      "loss": 2.9777,
+      "step": 4424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.963816115835154e-06,
+      "loss": 2.9305,
+      "step": 4425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9618059381198135e-06,
+      "loss": 3.1028,
+      "step": 4426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.95979576657822e-06,
+      "loss": 3.1117,
+      "step": 4427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9577856015352985e-06,
+      "loss": 3.0307,
+      "step": 4428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.955775443315976e-06,
+      "loss": 3.2492,
+      "step": 4429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.953765292245179e-06,
+      "loss": 2.918,
+      "step": 4430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.951755148647831e-06,
+      "loss": 3.1232,
+      "step": 4431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.949745012848852e-06,
+      "loss": 2.9781,
+      "step": 4432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.947734885173167e-06,
+      "loss": 2.9359,
+      "step": 4433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.945724765945695e-06,
+      "loss": 2.9296,
+      "step": 4434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.943714655491355e-06,
+      "loss": 3.0141,
+      "step": 4435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.941704554135061e-06,
+      "loss": 2.8386,
+      "step": 4436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9396944622017335e-06,
+      "loss": 3.0328,
+      "step": 4437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.937684380016285e-06,
+      "loss": 2.9319,
+      "step": 4438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9356743079036276e-06,
+      "loss": 2.9175,
+      "step": 4439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.933664246188673e-06,
+      "loss": 2.9487,
+      "step": 4440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.931654195196329e-06,
+      "loss": 3.0052,
+      "step": 4441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.929644155251505e-06,
+      "loss": 2.9292,
+      "step": 4442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.927634126679108e-06,
+      "loss": 2.9678,
+      "step": 4443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.92562410980404e-06,
+      "loss": 2.8557,
+      "step": 4444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9236141049512045e-06,
+      "loss": 3.0055,
+      "step": 4445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.921604112445501e-06,
+      "loss": 2.9706,
+      "step": 4446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.919594132611828e-06,
+      "loss": 2.97,
+      "step": 4447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.917584165775082e-06,
+      "loss": 2.9224,
+      "step": 4448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.915574212260156e-06,
+      "loss": 2.9385,
+      "step": 4449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.913564272391943e-06,
+      "loss": 2.9476,
+      "step": 4450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.911554346495332e-06,
+      "loss": 3.0643,
+      "step": 4451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.90954443489521e-06,
+      "loss": 2.9024,
+      "step": 4452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.907534537916464e-06,
+      "loss": 2.9626,
+      "step": 4453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.905524655883976e-06,
+      "loss": 2.9691,
+      "step": 4454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.903514789122627e-06,
+      "loss": 2.931,
+      "step": 4455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.901504937957293e-06,
+      "loss": 3.1123,
+      "step": 4456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8994951027128516e-06,
+      "loss": 2.9845,
+      "step": 4457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8974852837141735e-06,
+      "loss": 3.0466,
+      "step": 4458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.895475481286131e-06,
+      "loss": 2.9873,
+      "step": 4459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.89346569575359e-06,
+      "loss": 2.9862,
+      "step": 4460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.891455927441414e-06,
+      "loss": 3.0003,
+      "step": 4461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.889446176674469e-06,
+      "loss": 3.0911,
+      "step": 4462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.887436443777611e-06,
+      "loss": 2.7397,
+      "step": 4463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.885426729075698e-06,
+      "loss": 3.022,
+      "step": 4464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.883417032893584e-06,
+      "loss": 2.8842,
+      "step": 4465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.881407355556117e-06,
+      "loss": 2.8997,
+      "step": 4466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.879397697388145e-06,
+      "loss": 2.9658,
+      "step": 4467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877388058714512e-06,
+      "loss": 2.9932,
+      "step": 4468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8753784398600605e-06,
+      "loss": 3.0,
+      "step": 4469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.873368841149627e-06,
+      "loss": 2.8318,
+      "step": 4470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.871359262908045e-06,
+      "loss": 3.0477,
+      "step": 4471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.869349705460149e-06,
+      "loss": 3.0734,
+      "step": 4472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.867340169130765e-06,
+      "loss": 3.0578,
+      "step": 4473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.865330654244719e-06,
+      "loss": 2.7439,
+      "step": 4474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8633211611268285e-06,
+      "loss": 3.0379,
+      "step": 4475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.861311690101914e-06,
+      "loss": 2.9669,
+      "step": 4476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.85930224149479e-06,
+      "loss": 3.064,
+      "step": 4477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8572928156302645e-06,
+      "loss": 3.1027,
+      "step": 4478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.855283412833146e-06,
+      "loss": 2.9812,
+      "step": 4479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.853274033428234e-06,
+      "loss": 2.9201,
+      "step": 4480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.851264677740332e-06,
+      "loss": 3.0192,
+      "step": 4481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.849255346094233e-06,
+      "loss": 2.907,
+      "step": 4482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.84724603881473e-06,
+      "loss": 2.9944,
+      "step": 4483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.84523675622661e-06,
+      "loss": 2.9492,
+      "step": 4484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.843227498654655e-06,
+      "loss": 2.8748,
+      "step": 4485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.841218266423646e-06,
+      "loss": 2.8805,
+      "step": 4486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.839209059858357e-06,
+      "loss": 3.0338,
+      "step": 4487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.83719987928356e-06,
+      "loss": 2.9701,
+      "step": 4488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.835190725024023e-06,
+      "loss": 2.8468,
+      "step": 4489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.833181597404507e-06,
+      "loss": 3.0089,
+      "step": 4490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.83117249674977e-06,
+      "loss": 3.0695,
+      "step": 4491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.829163423384568e-06,
+      "loss": 3.0716,
+      "step": 4492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.827154377633651e-06,
+      "loss": 2.9865,
+      "step": 4493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.825145359821761e-06,
+      "loss": 3.0856,
+      "step": 4494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.823136370273643e-06,
+      "loss": 2.9542,
+      "step": 4495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.821127409314031e-06,
+      "loss": 3.0109,
+      "step": 4496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.819118477267656e-06,
+      "loss": 2.9246,
+      "step": 4497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.817109574459245e-06,
+      "loss": 3.0545,
+      "step": 4498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.81510070121352e-06,
+      "loss": 3.0668,
+      "step": 4499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8130918578551985e-06,
+      "loss": 2.9765,
+      "step": 4500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.811083044708992e-06,
+      "loss": 2.9075,
+      "step": 4501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.809074262099611e-06,
+      "loss": 2.9656,
+      "step": 4502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.807065510351758e-06,
+      "loss": 2.8568,
+      "step": 4503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.805056789790128e-06,
+      "loss": 2.9029,
+      "step": 4504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.803048100739415e-06,
+      "loss": 3.1269,
+      "step": 4505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.801039443524304e-06,
+      "loss": 2.944,
+      "step": 4506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.799030818469483e-06,
+      "loss": 2.9403,
+      "step": 4507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.797022225899625e-06,
+      "loss": 2.9392,
+      "step": 4508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.795013666139402e-06,
+      "loss": 3.0113,
+      "step": 4509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7930051395134805e-06,
+      "loss": 2.9695,
+      "step": 4510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.790996646346525e-06,
+      "loss": 3.0084,
+      "step": 4511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7889881869631895e-06,
+      "loss": 3.0621,
+      "step": 4512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.786979761688123e-06,
+      "loss": 2.9844,
+      "step": 4513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7849713708459725e-06,
+      "loss": 2.9209,
+      "step": 4514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7829630147613755e-06,
+      "loss": 2.9193,
+      "step": 4515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.780954693758968e-06,
+      "loss": 2.9251,
+      "step": 4516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7789464081633765e-06,
+      "loss": 3.0626,
+      "step": 4517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7769381582992225e-06,
+      "loss": 2.7277,
+      "step": 4518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.774929944491125e-06,
+      "loss": 2.85,
+      "step": 4519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.772921767063692e-06,
+      "loss": 2.9986,
+      "step": 4520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.770913626341529e-06,
+      "loss": 2.9411,
+      "step": 4521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.768905522649238e-06,
+      "loss": 3.0365,
+      "step": 4522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.766897456311409e-06,
+      "loss": 2.9045,
+      "step": 4523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.764889427652629e-06,
+      "loss": 2.8918,
+      "step": 4524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.76288143699748e-06,
+      "loss": 3.1296,
+      "step": 4525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.760873484670535e-06,
+      "loss": 2.9053,
+      "step": 4526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.758865570996364e-06,
+      "loss": 2.9364,
+      "step": 4527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.756857696299529e-06,
+      "loss": 3.0224,
+      "step": 4528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.754849860904585e-06,
+      "loss": 3.0601,
+      "step": 4529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.75284206513608e-06,
+      "loss": 2.9981,
+      "step": 4530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7508343093185625e-06,
+      "loss": 2.8814,
+      "step": 4531
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.7488265937765644e-06,
+      "loss": 2.8393,
+      "step": 4532
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.746818918834617e-06,
+      "loss": 2.856,
+      "step": 4533
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.744811284817246e-06,
+      "loss": 2.9804,
+      "step": 4534
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.742803692048967e-06,
+      "loss": 2.9349,
+      "step": 4535
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.74079614085429e-06,
+      "loss": 2.932,
+      "step": 4536
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.738788631557718e-06,
+      "loss": 3.0163,
+      "step": 4537
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.736781164483748e-06,
+      "loss": 2.9038,
+      "step": 4538
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.734773739956872e-06,
+      "loss": 2.9346,
+      "step": 4539
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.732766358301569e-06,
+      "loss": 3.0818,
+      "step": 4540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.730759019842321e-06,
+      "loss": 3.0383,
+      "step": 4541
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.728751724903594e-06,
+      "loss": 2.9525,
+      "step": 4542
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.72674447380985e-06,
+      "loss": 3.0142,
+      "step": 4543
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.724737266885545e-06,
+      "loss": 3.1113,
+      "step": 4544
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.722730104455126e-06,
+      "loss": 2.9302,
+      "step": 4545
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.720722986843035e-06,
+      "loss": 3.0343,
+      "step": 4546
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.718715914373705e-06,
+      "loss": 2.9859,
+      "step": 4547
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.716708887371562e-06,
+      "loss": 2.9407,
+      "step": 4548
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.7147019061610225e-06,
+      "loss": 2.9787,
+      "step": 4549
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.712694971066503e-06,
+      "loss": 2.8889,
+      "step": 4550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.710688082412405e-06,
+      "loss": 2.8434,
+      "step": 4551
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.708681240523125e-06,
+      "loss": 3.0861,
+      "step": 4552
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.7066744457230524e-06,
+      "loss": 3.1501,
+      "step": 4553
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.704667698336569e-06,
+      "loss": 3.0175,
+      "step": 4554
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.702660998688047e-06,
+      "loss": 2.8655,
+      "step": 4555
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.7006543471018526e-06,
+      "loss": 2.9652,
+      "step": 4556
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6986477439023445e-06,
+      "loss": 2.9761,
+      "step": 4557
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.696641189413873e-06,
+      "loss": 3.089,
+      "step": 4558
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.69463468396078e-06,
+      "loss": 2.9166,
+      "step": 4559
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.692628227867402e-06,
+      "loss": 2.975,
+      "step": 4560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.690621821458065e-06,
+      "loss": 2.9487,
+      "step": 4561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.688615465057087e-06,
+      "loss": 3.0633,
+      "step": 4562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6866091589887785e-06,
+      "loss": 2.8723,
+      "step": 4563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.684602903577441e-06,
+      "loss": 2.9131,
+      "step": 4564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.68259669914737e-06,
+      "loss": 2.9241,
+      "step": 4565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.680590546022852e-06,
+      "loss": 2.9329,
+      "step": 4566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6785844445281635e-06,
+      "loss": 2.8993,
+      "step": 4567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.676578394987573e-06,
+      "loss": 2.8756,
+      "step": 4568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.67457239772534e-06,
+      "loss": 2.8943,
+      "step": 4569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.672566453065723e-06,
+      "loss": 2.902,
+      "step": 4570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.67056056133296e-06,
+      "loss": 3.0015,
+      "step": 4571
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6685547228512905e-06,
+      "loss": 2.8398,
+      "step": 4572
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.666548937944938e-06,
+      "loss": 3.0751,
+      "step": 4573
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.664543206938124e-06,
+      "loss": 2.9555,
+      "step": 4574
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.662537530155055e-06,
+      "loss": 2.8817,
+      "step": 4575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.660531907919931e-06,
+      "loss": 3.004,
+      "step": 4576
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.658526340556947e-06,
+      "loss": 2.946,
+      "step": 4577
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.656520828390284e-06,
+      "loss": 2.8595,
+      "step": 4578
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.654515371744114e-06,
+      "loss": 2.9401,
+      "step": 4579
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.652509970942607e-06,
+      "loss": 2.9475,
+      "step": 4580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.650504626309918e-06,
+      "loss": 2.8463,
+      "step": 4581
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6484993381701905e-06,
+      "loss": 3.0715,
+      "step": 4582
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.646494106847563e-06,
+      "loss": 2.9773,
+      "step": 4583
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.644488932666168e-06,
+      "loss": 2.8763,
+      "step": 4584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.642483815950121e-06,
+      "loss": 3.0254,
+      "step": 4585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.640478757023535e-06,
+      "loss": 2.9381,
+      "step": 4586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.638473756210508e-06,
+      "loss": 2.9895,
+      "step": 4587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.636468813835131e-06,
+      "loss": 3.0963,
+      "step": 4588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.634463930221489e-06,
+      "loss": 3.0073,
+      "step": 4589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6324591056936525e-06,
+      "loss": 2.9249,
+      "step": 4590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.630454340575684e-06,
+      "loss": 2.9088,
+      "step": 4591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.62844963519164e-06,
+      "loss": 2.9183,
+      "step": 4592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.62644498986556e-06,
+      "loss": 2.9467,
+      "step": 4593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.624440404921481e-06,
+      "loss": 3.0445,
+      "step": 4594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.622435880683425e-06,
+      "loss": 2.95,
+      "step": 4595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.620431417475406e-06,
+      "loss": 2.9169,
+      "step": 4596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.618427015621432e-06,
+      "loss": 3.0515,
+      "step": 4597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.616422675445492e-06,
+      "loss": 3.0091,
+      "step": 4598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.614418397271576e-06,
+      "loss": 2.9043,
+      "step": 4599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.612414181423656e-06,
+      "loss": 3.1269,
+      "step": 4600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.610410028225697e-06,
+      "loss": 2.9395,
+      "step": 4601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.608405938001654e-06,
+      "loss": 2.9749,
+      "step": 4602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.606401911075468e-06,
+      "loss": 2.8371,
+      "step": 4603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6043979477710784e-06,
+      "loss": 2.9862,
+      "step": 4604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.602394048412405e-06,
+      "loss": 2.8557,
+      "step": 4605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.600390213323362e-06,
+      "loss": 2.858,
+      "step": 4606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5983864428278515e-06,
+      "loss": 2.878,
+      "step": 4607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.596382737249765e-06,
+      "loss": 3.0725,
+      "step": 4608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.594379096912989e-06,
+      "loss": 3.0861,
+      "step": 4609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.59237552214139e-06,
+      "loss": 2.9102,
+      "step": 4610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.590372013258834e-06,
+      "loss": 2.9075,
+      "step": 4611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.588368570589168e-06,
+      "loss": 3.0567,
+      "step": 4612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.586365194456232e-06,
+      "loss": 2.9467,
+      "step": 4613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.584361885183854e-06,
+      "loss": 2.9802,
+      "step": 4614
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.582358643095852e-06,
+      "loss": 2.9262,
+      "step": 4615
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5803554685160345e-06,
+      "loss": 2.9432,
+      "step": 4616
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5783523617681974e-06,
+      "loss": 2.9114,
+      "step": 4617
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.576349323176123e-06,
+      "loss": 3.0525,
+      "step": 4618
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5743463530635905e-06,
+      "loss": 3.0653,
+      "step": 4619
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.57234345175436e-06,
+      "loss": 3.0399,
+      "step": 4620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5703406195721845e-06,
+      "loss": 3.0071,
+      "step": 4621
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.568337856840803e-06,
+      "loss": 3.0552,
+      "step": 4622
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.566335163883948e-06,
+      "loss": 2.9846,
+      "step": 4623
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5643325410253355e-06,
+      "loss": 2.9168,
+      "step": 4624
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.562329988588674e-06,
+      "loss": 2.9262,
+      "step": 4625
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.560327506897657e-06,
+      "loss": 2.9068,
+      "step": 4626
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5583250962759706e-06,
+      "loss": 2.915,
+      "step": 4627
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.556322757047287e-06,
+      "loss": 2.9637,
+      "step": 4628
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.554320489535266e-06,
+      "loss": 2.9494,
+      "step": 4629
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.552318294063561e-06,
+      "loss": 3.083,
+      "step": 4630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.550316170955807e-06,
+      "loss": 2.9682,
+      "step": 4631
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.54831412053563e-06,
+      "loss": 3.0047,
+      "step": 4632
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.546312143126646e-06,
+      "loss": 2.9227,
+      "step": 4633
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.544310239052455e-06,
+      "loss": 2.8927,
+      "step": 4634
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.542308408636652e-06,
+      "loss": 3.0483,
+      "step": 4635
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.540306652202812e-06,
+      "loss": 3.0979,
+      "step": 4636
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5383049700745044e-06,
+      "loss": 2.919,
+      "step": 4637
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5363033625752804e-06,
+      "loss": 2.925,
+      "step": 4638
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.534301830028688e-06,
+      "loss": 3.0592,
+      "step": 4639
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.532300372758256e-06,
+      "loss": 3.0765,
+      "step": 4640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.530298991087502e-06,
+      "loss": 3.0425,
+      "step": 4641
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.528297685339933e-06,
+      "loss": 3.0098,
+      "step": 4642
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5262964558390434e-06,
+      "loss": 2.9499,
+      "step": 4643
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.524295302908316e-06,
+      "loss": 2.8351,
+      "step": 4644
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.522294226871218e-06,
+      "loss": 2.9228,
+      "step": 4645
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.520293228051206e-06,
+      "loss": 2.9114,
+      "step": 4646
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.518292306771725e-06,
+      "loss": 3.1493,
+      "step": 4647
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.516291463356209e-06,
+      "loss": 2.8696,
+      "step": 4648
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.514290698128077e-06,
+      "loss": 2.8323,
+      "step": 4649
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.512290011410735e-06,
+      "loss": 3.0841,
+      "step": 4650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.510289403527577e-06,
+      "loss": 2.9268,
+      "step": 4651
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.508288874801984e-06,
+      "loss": 2.9014,
+      "step": 4652
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.506288425557323e-06,
+      "loss": 3.0024,
+      "step": 4653
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5042880561169525e-06,
+      "loss": 2.9037,
+      "step": 4654
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5022877668042145e-06,
+      "loss": 2.9932,
+      "step": 4655
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.500287557942439e-06,
+      "loss": 2.985,
+      "step": 4656
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.49828742985494e-06,
+      "loss": 3.0296,
+      "step": 4657
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.496287382865025e-06,
+      "loss": 2.9155,
+      "step": 4658
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.494287417295983e-06,
+      "loss": 2.9048,
+      "step": 4659
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4922875334710915e-06,
+      "loss": 2.9911,
+      "step": 4660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.490287731713613e-06,
+      "loss": 3.1,
+      "step": 4661
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.488288012346803e-06,
+      "loss": 2.9077,
+      "step": 4662
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.486288375693896e-06,
+      "loss": 2.8455,
+      "step": 4663
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.484288822078117e-06,
+      "loss": 2.8563,
+      "step": 4664
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.482289351822676e-06,
+      "loss": 3.1115,
+      "step": 4665
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.48028996525077e-06,
+      "loss": 3.0588,
+      "step": 4666
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.478290662685585e-06,
+      "loss": 3.0511,
+      "step": 4667
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.476291444450289e-06,
+      "loss": 2.9192,
+      "step": 4668
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.474292310868042e-06,
+      "loss": 3.0134,
+      "step": 4669
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.472293262261985e-06,
+      "loss": 3.1453,
+      "step": 4670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.470294298955247e-06,
+      "loss": 2.9603,
+      "step": 4671
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.468295421270943e-06,
+      "loss": 2.9596,
+      "step": 4672
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.466296629532175e-06,
+      "loss": 3.1573,
+      "step": 4673
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.464297924062032e-06,
+      "loss": 2.8405,
+      "step": 4674
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.462299305183587e-06,
+      "loss": 2.8159,
+      "step": 4675
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.460300773219899e-06,
+      "loss": 2.986,
+      "step": 4676
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.458302328494013e-06,
+      "loss": 3.087,
+      "step": 4677
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.456303971328963e-06,
+      "loss": 3.0696,
+      "step": 4678
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.454305702047767e-06,
+      "loss": 2.8369,
+      "step": 4679
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4523075209734245e-06,
+      "loss": 2.9244,
+      "step": 4680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4503094284289285e-06,
+      "loss": 3.0016,
+      "step": 4681
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.448311424737253e-06,
+      "loss": 3.0192,
+      "step": 4682
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.446313510221355e-06,
+      "loss": 2.8323,
+      "step": 4683
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.444315685204185e-06,
+      "loss": 2.9305,
+      "step": 4684
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.44231795000867e-06,
+      "loss": 3.0253,
+      "step": 4685
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.440320304957731e-06,
+      "loss": 2.8846,
+      "step": 4686
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.438322750374266e-06,
+      "loss": 2.982,
+      "step": 4687
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.436325286581168e-06,
+      "loss": 2.8344,
+      "step": 4688
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4343279139013066e-06,
+      "loss": 3.0617,
+      "step": 4689
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.432330632657542e-06,
+      "loss": 2.8576,
+      "step": 4690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.430333443172716e-06,
+      "loss": 3.059,
+      "step": 4691
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.428336345769658e-06,
+      "loss": 3.0798,
+      "step": 4692
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.426339340771182e-06,
+      "loss": 2.8331,
+      "step": 4693
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.424342428500088e-06,
+      "loss": 2.9537,
+      "step": 4694
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.422345609279158e-06,
+      "loss": 3.0237,
+      "step": 4695
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.42034888343116e-06,
+      "loss": 2.9642,
+      "step": 4696
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.418352251278852e-06,
+      "loss": 2.8554,
+      "step": 4697
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.41635571314497e-06,
+      "loss": 2.8122,
+      "step": 4698
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.414359269352237e-06,
+      "loss": 2.9562,
+      "step": 4699
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.412362920223362e-06,
+      "loss": 2.8324,
+      "step": 4700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.410366666081039e-06,
+      "loss": 3.0802,
+      "step": 4701
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.408370507247944e-06,
+      "loss": 3.0343,
+      "step": 4702
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.406374444046738e-06,
+      "loss": 2.7868,
+      "step": 4703
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.40437847680007e-06,
+      "loss": 2.9559,
+      "step": 4704
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.402382605830571e-06,
+      "loss": 2.9333,
+      "step": 4705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.400386831460853e-06,
+      "loss": 2.9748,
+      "step": 4706
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.398391154013522e-06,
+      "loss": 2.8557,
+      "step": 4707
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.396395573811159e-06,
+      "loss": 2.9788,
+      "step": 4708
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.394400091176334e-06,
+      "loss": 2.958,
+      "step": 4709
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.392404706431598e-06,
+      "loss": 2.8754,
+      "step": 4710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.390409419899487e-06,
+      "loss": 3.1784,
+      "step": 4711
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.388414231902525e-06,
+      "loss": 2.9391,
+      "step": 4712
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.386419142763217e-06,
+      "loss": 2.9585,
+      "step": 4713
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.384424152804051e-06,
+      "loss": 2.9619,
+      "step": 4714
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3824292623475014e-06,
+      "loss": 3.0067,
+      "step": 4715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.380434471716022e-06,
+      "loss": 2.9778,
+      "step": 4716
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.378439781232058e-06,
+      "loss": 3.0409,
+      "step": 4717
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.376445191218032e-06,
+      "loss": 3.0188,
+      "step": 4718
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.374450701996354e-06,
+      "loss": 3.0342,
+      "step": 4719
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3724563138894145e-06,
+      "loss": 2.839,
+      "step": 4720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.370462027219592e-06,
+      "loss": 2.8248,
+      "step": 4721
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.368467842309243e-06,
+      "loss": 2.9533,
+      "step": 4722
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.366473759480711e-06,
+      "loss": 2.9261,
+      "step": 4723
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.364479779056325e-06,
+      "loss": 3.0074,
+      "step": 4724
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.362485901358395e-06,
+      "loss": 2.9686,
+      "step": 4725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3604921267092084e-06,
+      "loss": 3.0942,
+      "step": 4726
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.35849845543105e-06,
+      "loss": 3.0393,
+      "step": 4727
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.356504887846175e-06,
+      "loss": 2.736,
+      "step": 4728
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.354511424276829e-06,
+      "loss": 3.0399,
+      "step": 4729
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.352518065045237e-06,
+      "loss": 2.8491,
+      "step": 4730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.350524810473609e-06,
+      "loss": 3.0195,
+      "step": 4731
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.348531660884139e-06,
+      "loss": 3.0302,
+      "step": 4732
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.346538616599001e-06,
+      "loss": 2.8717,
+      "step": 4733
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.344545677940354e-06,
+      "loss": 3.1841,
+      "step": 4734
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.342552845230339e-06,
+      "loss": 2.9841,
+      "step": 4735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.340560118791082e-06,
+      "loss": 3.0334,
+      "step": 4736
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.33856749894469e-06,
+      "loss": 3.0163,
+      "step": 4737
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3365749860132525e-06,
+      "loss": 2.8073,
+      "step": 4738
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.334582580318843e-06,
+      "loss": 2.9948,
+      "step": 4739
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.332590282183518e-06,
+      "loss": 2.8244,
+      "step": 4740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.330598091929313e-06,
+      "loss": 3.0006,
+      "step": 4741
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.328606009878251e-06,
+      "loss": 3.0705,
+      "step": 4742
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3266140363523326e-06,
+      "loss": 2.8643,
+      "step": 4743
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.3246221716735455e-06,
+      "loss": 3.0305,
+      "step": 4744
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.322630416163855e-06,
+      "loss": 3.0854,
+      "step": 4745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.320638770145217e-06,
+      "loss": 2.8827,
+      "step": 4746
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.318647233939559e-06,
+      "loss": 2.8166,
+      "step": 4747
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.316655807868799e-06,
+      "loss": 2.9839,
+      "step": 4748
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.314664492254833e-06,
+      "loss": 2.9449,
+      "step": 4749
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.312673287419538e-06,
+      "loss": 3.0797,
+      "step": 4750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.31068219368478e-06,
+      "loss": 3.066,
+      "step": 4751
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.308691211372399e-06,
+      "loss": 3.1808,
+      "step": 4752
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.306700340804223e-06,
+      "loss": 3.0129,
+      "step": 4753
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.304709582302056e-06,
+      "loss": 3.0032,
+      "step": 4754
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.302718936187688e-06,
+      "loss": 2.9165,
+      "step": 4755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.300728402782892e-06,
+      "loss": 2.9922,
+      "step": 4756
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.298737982409419e-06,
+      "loss": 2.9396,
+      "step": 4757
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.296747675389007e-06,
+      "loss": 2.9261,
+      "step": 4758
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.294757482043368e-06,
+      "loss": 2.9377,
+      "step": 4759
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.292767402694203e-06,
+      "loss": 2.968,
+      "step": 4760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.2907774376631885e-06,
+      "loss": 2.9825,
+      "step": 4761
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.288787587271987e-06,
+      "loss": 3.0383,
+      "step": 4762
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.286797851842241e-06,
+      "loss": 2.9829,
+      "step": 4763
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.284808231695576e-06,
+      "loss": 2.7791,
+      "step": 4764
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.282818727153592e-06,
+      "loss": 2.9036,
+      "step": 4765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.280829338537882e-06,
+      "loss": 3.0845,
+      "step": 4766
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.27884006617001e-06,
+      "loss": 2.8253,
+      "step": 4767
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.276850910371527e-06,
+      "loss": 2.9422,
+      "step": 4768
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.27486187146396e-06,
+      "loss": 2.979,
+      "step": 4769
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.272872949768824e-06,
+      "loss": 3.1307,
+      "step": 4770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.27088414560761e-06,
+      "loss": 3.0678,
+      "step": 4771
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.26889545930179e-06,
+      "loss": 2.9605,
+      "step": 4772
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.2669068911728204e-06,
+      "loss": 2.7621,
+      "step": 4773
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.264918441542133e-06,
+      "loss": 2.8436,
+      "step": 4774
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.262930110731147e-06,
+      "loss": 3.0717,
+      "step": 4775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.260941899061257e-06,
+      "loss": 3.0591,
+      "step": 4776
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.258953806853844e-06,
+      "loss": 3.1158,
+      "step": 4777
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.2569658344302635e-06,
+      "loss": 2.9827,
+      "step": 4778
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.254977982111856e-06,
+      "loss": 3.1353,
+      "step": 4779
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.252990250219939e-06,
+      "loss": 2.9929,
+      "step": 4780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.251002639075812e-06,
+      "loss": 2.8262,
+      "step": 4781
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.249015149000759e-06,
+      "loss": 3.0315,
+      "step": 4782
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.247027780316038e-06,
+      "loss": 2.9828,
+      "step": 4783
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2450405333428895e-06,
+      "loss": 2.955,
+      "step": 4784
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2430534084025385e-06,
+      "loss": 2.8778,
+      "step": 4785
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.241066405816186e-06,
+      "loss": 3.0406,
+      "step": 4786
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.239079525905012e-06,
+      "loss": 2.9111,
+      "step": 4787
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2370927689901795e-06,
+      "loss": 2.9513,
+      "step": 4788
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.235106135392833e-06,
+      "loss": 3.0989,
+      "step": 4789
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.233119625434093e-06,
+      "loss": 3.0785,
+      "step": 4790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.231133239435062e-06,
+      "loss": 2.8582,
+      "step": 4791
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.229146977716824e-06,
+      "loss": 2.7991,
+      "step": 4792
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.227160840600438e-06,
+      "loss": 2.7306,
+      "step": 4793
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.22517482840695e-06,
+      "loss": 2.9862,
+      "step": 4794
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.223188941457379e-06,
+      "loss": 2.9297,
+      "step": 4795
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.22120318007273e-06,
+      "loss": 3.1201,
+      "step": 4796
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.219217544573983e-06,
+      "loss": 2.9967,
+      "step": 4797
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.217232035282098e-06,
+      "loss": 3.0738,
+      "step": 4798
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.215246652518017e-06,
+      "loss": 3.0272,
+      "step": 4799
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.213261396602659e-06,
+      "loss": 2.8611,
+      "step": 4800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.211276267856926e-06,
+      "loss": 3.1306,
+      "step": 4801
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.209291266601696e-06,
+      "loss": 2.8909,
+      "step": 4802
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.207306393157827e-06,
+      "loss": 3.0747,
+      "step": 4803
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.205321647846156e-06,
+      "loss": 2.8299,
+      "step": 4804
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.203337030987503e-06,
+      "loss": 2.9839,
+      "step": 4805
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2013525429026645e-06,
+      "loss": 2.9445,
+      "step": 4806
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.199368183912414e-06,
+      "loss": 3.0458,
+      "step": 4807
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.197383954337506e-06,
+      "loss": 2.8788,
+      "step": 4808
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.195399854498677e-06,
+      "loss": 3.024,
+      "step": 4809
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.19341588471664e-06,
+      "loss": 3.1137,
+      "step": 4810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.191432045312084e-06,
+      "loss": 3.0153,
+      "step": 4811
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.18944833660568e-06,
+      "loss": 2.9208,
+      "step": 4812
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.18746475891808e-06,
+      "loss": 2.8948,
+      "step": 4813
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.185481312569911e-06,
+      "loss": 3.0044,
+      "step": 4814
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.18349799788178e-06,
+      "loss": 2.815,
+      "step": 4815
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.181514815174275e-06,
+      "loss": 2.9503,
+      "step": 4816
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.17953176476796e-06,
+      "loss": 3.053,
+      "step": 4817
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.177548846983377e-06,
+      "loss": 2.8761,
+      "step": 4818
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.175566062141049e-06,
+      "loss": 2.9251,
+      "step": 4819
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.173583410561474e-06,
+      "loss": 2.8678,
+      "step": 4820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.171600892565135e-06,
+      "loss": 2.9506,
+      "step": 4821
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.169618508472486e-06,
+      "loss": 2.9385,
+      "step": 4822
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.167636258603964e-06,
+      "loss": 2.8959,
+      "step": 4823
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.16565414327998e-06,
+      "loss": 2.9661,
+      "step": 4824
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.16367216282093e-06,
+      "loss": 2.9956,
+      "step": 4825
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.161690317547183e-06,
+      "loss": 2.7489,
+      "step": 4826
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.159708607779088e-06,
+      "loss": 3.1565,
+      "step": 4827
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.15772703383697e-06,
+      "loss": 3.0333,
+      "step": 4828
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.155745596041135e-06,
+      "loss": 2.9353,
+      "step": 4829
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.153764294711866e-06,
+      "loss": 2.9673,
+      "step": 4830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.151783130169422e-06,
+      "loss": 2.9337,
+      "step": 4831
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.149802102734041e-06,
+      "loss": 2.7823,
+      "step": 4832
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.147821212725941e-06,
+      "loss": 2.9702,
+      "step": 4833
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.145840460465315e-06,
+      "loss": 2.9485,
+      "step": 4834
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.143859846272336e-06,
+      "loss": 2.9467,
+      "step": 4835
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.1418793704671535e-06,
+      "loss": 2.7302,
+      "step": 4836
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.139899033369894e-06,
+      "loss": 2.917,
+      "step": 4837
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.137918835300662e-06,
+      "loss": 2.8493,
+      "step": 4838
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.135938776579538e-06,
+      "loss": 2.9404,
+      "step": 4839
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.133958857526586e-06,
+      "loss": 3.0427,
+      "step": 4840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.13197907846184e-06,
+      "loss": 2.9455,
+      "step": 4841
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.129999439705315e-06,
+      "loss": 2.7725,
+      "step": 4842
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.128019941577001e-06,
+      "loss": 3.0173,
+      "step": 4843
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.126040584396871e-06,
+      "loss": 2.9089,
+      "step": 4844
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.124061368484869e-06,
+      "loss": 2.8607,
+      "step": 4845
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.122082294160919e-06,
+      "loss": 2.9679,
+      "step": 4846
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.120103361744922e-06,
+      "loss": 2.9698,
+      "step": 4847
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.118124571556754e-06,
+      "loss": 2.9592,
+      "step": 4848
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.116145923916272e-06,
+      "loss": 3.0111,
+      "step": 4849
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.114167419143305e-06,
+      "loss": 2.979,
+      "step": 4850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.112189057557664e-06,
+      "loss": 2.982,
+      "step": 4851
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.110210839479133e-06,
+      "loss": 2.9145,
+      "step": 4852
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.1082327652274725e-06,
+      "loss": 2.966,
+      "step": 4853
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.106254835122425e-06,
+      "loss": 2.8209,
+      "step": 4854
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.104277049483705e-06,
+      "loss": 2.9423,
+      "step": 4855
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.102299408631004e-06,
+      "loss": 3.032,
+      "step": 4856
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.100321912883991e-06,
+      "loss": 3.0569,
+      "step": 4857
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.09834456256231e-06,
+      "loss": 3.0131,
+      "step": 4858
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.096367357985587e-06,
+      "loss": 3.0421,
+      "step": 4859
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.094390299473417e-06,
+      "loss": 3.0378,
+      "step": 4860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.092413387345375e-06,
+      "loss": 3.0458,
+      "step": 4861
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.090436621921013e-06,
+      "loss": 2.9617,
+      "step": 4862
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.088460003519856e-06,
+      "loss": 3.1063,
+      "step": 4863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.086483532461412e-06,
+      "loss": 3.03,
+      "step": 4864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.084507209065157e-06,
+      "loss": 3.0074,
+      "step": 4865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.082531033650549e-06,
+      "loss": 2.9855,
+      "step": 4866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.08055500653702e-06,
+      "loss": 3.0752,
+      "step": 4867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.078579128043977e-06,
+      "loss": 3.045,
+      "step": 4868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.076603398490804e-06,
+      "loss": 2.8016,
+      "step": 4869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0746278181968604e-06,
+      "loss": 3.0565,
+      "step": 4870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.072652387481484e-06,
+      "loss": 3.0633,
+      "step": 4871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.070677106663985e-06,
+      "loss": 2.9721,
+      "step": 4872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0687019760636495e-06,
+      "loss": 3.0274,
+      "step": 4873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.066726995999744e-06,
+      "loss": 3.0827,
+      "step": 4874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.064752166791506e-06,
+      "loss": 2.9481,
+      "step": 4875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.062777488758149e-06,
+      "loss": 2.9172,
+      "step": 4876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.060802962218862e-06,
+      "loss": 3.0363,
+      "step": 4877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.058828587492813e-06,
+      "loss": 3.0294,
+      "step": 4878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.056854364899142e-06,
+      "loss": 3.0413,
+      "step": 4879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.054880294756964e-06,
+      "loss": 3.0022,
+      "step": 4880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.052906377385372e-06,
+      "loss": 2.9899,
+      "step": 4881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.050932613103432e-06,
+      "loss": 2.9893,
+      "step": 4882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.048959002230189e-06,
+      "loss": 3.1646,
+      "step": 4883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0469855450846575e-06,
+      "loss": 3.0459,
+      "step": 4884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.04501224198583e-06,
+      "loss": 3.0679,
+      "step": 4885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.043039093252676e-06,
+      "loss": 3.0176,
+      "step": 4886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.041066099204139e-06,
+      "loss": 2.9047,
+      "step": 4887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.039093260159135e-06,
+      "loss": 3.0371,
+      "step": 4888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0371205764365565e-06,
+      "loss": 2.8824,
+      "step": 4889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0351480483552705e-06,
+      "loss": 3.1061,
+      "step": 4890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0331756762341225e-06,
+      "loss": 2.8727,
+      "step": 4891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.031203460391925e-06,
+      "loss": 2.9262,
+      "step": 4892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0292314011474746e-06,
+      "loss": 2.8971,
+      "step": 4893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.027259498819536e-06,
+      "loss": 2.7896,
+      "step": 4894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.025287753726852e-06,
+      "loss": 3.0231,
+      "step": 4895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.023316166188135e-06,
+      "loss": 2.8517,
+      "step": 4896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0213447365220785e-06,
+      "loss": 3.0136,
+      "step": 4897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0193734650473455e-06,
+      "loss": 2.79,
+      "step": 4898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.017402352082578e-06,
+      "loss": 3.0722,
+      "step": 4899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0154313979463865e-06,
+      "loss": 2.7535,
+      "step": 4900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.013460602957361e-06,
+      "loss": 3.0815,
+      "step": 4901
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.011489967434061e-06,
+      "loss": 3.0281,
+      "step": 4902
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.009519491695028e-06,
+      "loss": 2.8755,
+      "step": 4903
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.007549176058768e-06,
+      "loss": 2.9446,
+      "step": 4904
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.005579020843769e-06,
+      "loss": 2.9521,
+      "step": 4905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0036090263684886e-06,
+      "loss": 2.8757,
+      "step": 4906
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.00163919295136e-06,
+      "loss": 2.8953,
+      "step": 4907
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.9996695209107895e-06,
+      "loss": 3.0172,
+      "step": 4908
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.997700010565157e-06,
+      "loss": 2.8924,
+      "step": 4909
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.99573066223282e-06,
+      "loss": 2.7798,
+      "step": 4910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.993761476232104e-06,
+      "loss": 3.0562,
+      "step": 4911
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.991792452881311e-06,
+      "loss": 2.9197,
+      "step": 4912
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.9898235924987215e-06,
+      "loss": 2.9112,
+      "step": 4913
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.987854895402581e-06,
+      "loss": 3.0073,
+      "step": 4914
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.985886361911114e-06,
+      "loss": 3.0596,
+      "step": 4915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.983917992342516e-06,
+      "loss": 2.9179,
+      "step": 4916
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.981949787014959e-06,
+      "loss": 3.1081,
+      "step": 4917
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.979981746246585e-06,
+      "loss": 2.8717,
+      "step": 4918
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.978013870355514e-06,
+      "loss": 2.9853,
+      "step": 4919
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.9760461596598336e-06,
+      "loss": 3.03,
+      "step": 4920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.974078614477607e-06,
+      "loss": 2.9673,
+      "step": 4921
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.972111235126872e-06,
+      "loss": 3.1029,
+      "step": 4922
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.970144021925641e-06,
+      "loss": 3.0092,
+      "step": 4923
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.968176975191895e-06,
+      "loss": 2.8383,
+      "step": 4924
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.966210095243592e-06,
+      "loss": 3.1032,
+      "step": 4925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.96424338239866e-06,
+      "loss": 2.9961,
+      "step": 4926
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.962276836975003e-06,
+      "loss": 2.993,
+      "step": 4927
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.960310459290494e-06,
+      "loss": 2.9538,
+      "step": 4928
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.9583442496629835e-06,
+      "loss": 2.9522,
+      "step": 4929
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.956378208410292e-06,
+      "loss": 3.0275,
+      "step": 4930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.954412335850212e-06,
+      "loss": 2.9614,
+      "step": 4931
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.952446632300512e-06,
+      "loss": 2.9913,
+      "step": 4932
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.950481098078931e-06,
+      "loss": 2.9224,
+      "step": 4933
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.948515733503181e-06,
+      "loss": 3.0718,
+      "step": 4934
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.946550538890945e-06,
+      "loss": 3.0339,
+      "step": 4935
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.944585514559883e-06,
+      "loss": 2.9306,
+      "step": 4936
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.942620660827622e-06,
+      "loss": 2.8714,
+      "step": 4937
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.940655978011764e-06,
+      "loss": 2.9216,
+      "step": 4938
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.938691466429885e-06,
+      "loss": 3.1066,
+      "step": 4939
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.936727126399529e-06,
+      "loss": 3.0169,
+      "step": 4940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.934762958238217e-06,
+      "loss": 2.9652,
+      "step": 4941
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.932798962263438e-06,
+      "loss": 3.0724,
+      "step": 4942
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.930835138792659e-06,
+      "loss": 3.1393,
+      "step": 4943
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.928871488143313e-06,
+      "loss": 2.9561,
+      "step": 4944
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.926908010632809e-06,
+      "loss": 2.887,
+      "step": 4945
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.924944706578525e-06,
+      "loss": 2.9892,
+      "step": 4946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.9229815762978105e-06,
+      "loss": 3.0204,
+      "step": 4947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.921018620107993e-06,
+      "loss": 3.0432,
+      "step": 4948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.919055838326366e-06,
+      "loss": 2.9489,
+      "step": 4949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.917093231270196e-06,
+      "loss": 2.867,
+      "step": 4950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.91513079925672e-06,
+      "loss": 3.0181,
+      "step": 4951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.913168542603152e-06,
+      "loss": 3.0575,
+      "step": 4952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.911206461626674e-06,
+      "loss": 3.0109,
+      "step": 4953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.909244556644438e-06,
+      "loss": 2.9175,
+      "step": 4954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.907282827973569e-06,
+      "loss": 3.0207,
+      "step": 4955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.905321275931166e-06,
+      "loss": 3.0234,
+      "step": 4956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.903359900834296e-06,
+      "loss": 3.0254,
+      "step": 4957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.901398702999997e-06,
+      "loss": 2.9318,
+      "step": 4958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.899437682745283e-06,
+      "loss": 2.9298,
+      "step": 4959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.897476840387133e-06,
+      "loss": 3.0093,
+      "step": 4960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8955161762425035e-06,
+      "loss": 2.8656,
+      "step": 4961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8935556906283175e-06,
+      "loss": 2.7742,
+      "step": 4962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8915953838614715e-06,
+      "loss": 3.0429,
+      "step": 4963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.889635256258834e-06,
+      "loss": 2.9076,
+      "step": 4964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.887675308137241e-06,
+      "loss": 3.0632,
+      "step": 4965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.885715539813502e-06,
+      "loss": 2.8996,
+      "step": 4966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.883755951604395e-06,
+      "loss": 3.0747,
+      "step": 4967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.881796543826675e-06,
+      "loss": 2.9412,
+      "step": 4968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8798373167970605e-06,
+      "loss": 2.9586,
+      "step": 4969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8778782708322445e-06,
+      "loss": 2.955,
+      "step": 4970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8759194062488895e-06,
+      "loss": 2.9699,
+      "step": 4971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8739607233636325e-06,
+      "loss": 2.9289,
+      "step": 4972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.872002222493074e-06,
+      "loss": 2.8545,
+      "step": 4973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.870043903953791e-06,
+      "loss": 3.0282,
+      "step": 4974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.86808576806233e-06,
+      "loss": 3.1185,
+      "step": 4975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.866127815135206e-06,
+      "loss": 2.9996,
+      "step": 4976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.864170045488906e-06,
+      "loss": 2.8906,
+      "step": 4977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.862212459439885e-06,
+      "loss": 2.9471,
+      "step": 4978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.860255057304572e-06,
+      "loss": 2.9994,
+      "step": 4979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.858297839399363e-06,
+      "loss": 2.924,
+      "step": 4980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.856340806040627e-06,
+      "loss": 3.0179,
+      "step": 4981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8543839575447015e-06,
+      "loss": 2.7958,
+      "step": 4982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.852427294227896e-06,
+      "loss": 3.049,
+      "step": 4983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.850470816406487e-06,
+      "loss": 2.9295,
+      "step": 4984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.848514524396721e-06,
+      "loss": 2.811,
+      "step": 4985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.846558418514818e-06,
+      "loss": 2.9498,
+      "step": 4986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.844602499076965e-06,
+      "loss": 3.067,
+      "step": 4987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.842646766399322e-06,
+      "loss": 3.0585,
+      "step": 4988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.840691220798014e-06,
+      "loss": 3.0701,
+      "step": 4989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.838735862589137e-06,
+      "loss": 2.8775,
+      "step": 4990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.836780692088762e-06,
+      "loss": 2.9072,
+      "step": 4991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.834825709612926e-06,
+      "loss": 3.0504,
+      "step": 4992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.832870915477631e-06,
+      "loss": 2.8292,
+      "step": 4993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.830916309998856e-06,
+      "loss": 2.9277,
+      "step": 4994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.828961893492546e-06,
+      "loss": 2.9013,
+      "step": 4995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.827007666274616e-06,
+      "loss": 3.0438,
+      "step": 4996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.825053628660949e-06,
+      "loss": 2.9005,
+      "step": 4997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.823099780967398e-06,
+      "loss": 3.0665,
+      "step": 4998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8211461235097904e-06,
+      "loss": 2.9878,
+      "step": 4999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.819192656603912e-06,
+      "loss": 3.0477,
+      "step": 5000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.81723938056553e-06,
+      "loss": 3.0882,
+      "step": 5001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.815286295710371e-06,
+      "loss": 2.95,
+      "step": 5002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8133334023541367e-06,
+      "loss": 3.072,
+      "step": 5003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8113807008124955e-06,
+      "loss": 3.0675,
+      "step": 5004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8094281914010834e-06,
+      "loss": 3.1333,
+      "step": 5005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.807475874435509e-06,
+      "loss": 2.923,
+      "step": 5006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.805523750231347e-06,
+      "loss": 3.0638,
+      "step": 5007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.803571819104141e-06,
+      "loss": 3.0542,
+      "step": 5008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.8016200813694055e-06,
+      "loss": 3.0253,
+      "step": 5009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.799668537342619e-06,
+      "loss": 2.8768,
+      "step": 5010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.797717187339237e-06,
+      "loss": 3.0395,
+      "step": 5011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.795766031674676e-06,
+      "loss": 3.006,
+      "step": 5012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.793815070664324e-06,
+      "loss": 3.0259,
+      "step": 5013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.791864304623538e-06,
+      "loss": 2.9595,
+      "step": 5014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.789913733867642e-06,
+      "loss": 3.0348,
+      "step": 5015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7879633587119298e-06,
+      "loss": 2.9434,
+      "step": 5016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7860131794716613e-06,
+      "loss": 2.9193,
+      "step": 5017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.784063196462069e-06,
+      "loss": 3.0257,
+      "step": 5018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7821134099983486e-06,
+      "loss": 3.0147,
+      "step": 5019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.780163820395668e-06,
+      "loss": 3.0435,
+      "step": 5020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7782144279691624e-06,
+      "loss": 2.963,
+      "step": 5021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.776265233033934e-06,
+      "loss": 3.0338,
+      "step": 5022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.774316235905053e-06,
+      "loss": 3.0454,
+      "step": 5023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7723674368975594e-06,
+      "loss": 3.0919,
+      "step": 5024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7704188363264584e-06,
+      "loss": 2.9433,
+      "step": 5025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.768470434506726e-06,
+      "loss": 2.936,
+      "step": 5026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.766522231753305e-06,
+      "loss": 2.9478,
+      "step": 5027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7645742283811056e-06,
+      "loss": 3.0591,
+      "step": 5028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7626264247050032e-06,
+      "loss": 2.9912,
+      "step": 5029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7606788210398487e-06,
+      "loss": 2.9533,
+      "step": 5030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.758731417700453e-06,
+      "loss": 3.0083,
+      "step": 5031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.756784215001597e-06,
+      "loss": 3.0161,
+      "step": 5032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7548372132580315e-06,
+      "loss": 3.0194,
+      "step": 5033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.752890412784471e-06,
+      "loss": 2.9343,
+      "step": 5034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7509438138956e-06,
+      "loss": 3.0179,
+      "step": 5035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.74899741690607e-06,
+      "loss": 2.9006,
+      "step": 5036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7470512221304977e-06,
+      "loss": 3.0952,
+      "step": 5037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.745105229883471e-06,
+      "loss": 2.939,
+      "step": 5038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7431594404795408e-06,
+      "loss": 3.0784,
+      "step": 5039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.741213854233231e-06,
+      "loss": 2.8774,
+      "step": 5040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.739268471459028e-06,
+      "loss": 3.028,
+      "step": 5041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7373232924713844e-06,
+      "loss": 3.1084,
+      "step": 5042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7353783175847236e-06,
+      "loss": 2.9685,
+      "step": 5043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7334335471134325e-06,
+      "loss": 3.0341,
+      "step": 5044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.731488981371869e-06,
+      "loss": 3.0396,
+      "step": 5045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7295446206743552e-06,
+      "loss": 3.0262,
+      "step": 5046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.72760046533518e-06,
+      "loss": 3.0609,
+      "step": 5047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7256565156685996e-06,
+      "loss": 2.7481,
+      "step": 5048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.723712771988835e-06,
+      "loss": 2.8201,
+      "step": 5049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.72176923461008e-06,
+      "loss": 3.0103,
+      "step": 5050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7198259038464874e-06,
+      "loss": 3.04,
+      "step": 5051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.717882780012183e-06,
+      "loss": 2.937,
+      "step": 5052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.715939863421255e-06,
+      "loss": 2.9327,
+      "step": 5053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.713997154387759e-06,
+      "loss": 2.8314,
+      "step": 5054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.712054653225718e-06,
+      "loss": 2.9168,
+      "step": 5055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7101123602491194e-06,
+      "loss": 2.9272,
+      "step": 5056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.708170275771921e-06,
+      "loss": 3.0593,
+      "step": 5057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.706228400108042e-06,
+      "loss": 2.9084,
+      "step": 5058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7042867335713693e-06,
+      "loss": 2.7459,
+      "step": 5059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.7023452764757606e-06,
+      "loss": 2.9548,
+      "step": 5060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.700404029135034e-06,
+      "loss": 3.0193,
+      "step": 5061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6984629918629757e-06,
+      "loss": 3.0737,
+      "step": 5062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.696522164973336e-06,
+      "loss": 2.9683,
+      "step": 5063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6945815487798363e-06,
+      "loss": 2.8794,
+      "step": 5064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6926411435961594e-06,
+      "loss": 2.9645,
+      "step": 5065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6907009497359556e-06,
+      "loss": 2.9817,
+      "step": 5066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.68876096751284e-06,
+      "loss": 3.0787,
+      "step": 5067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6868211972403927e-06,
+      "loss": 2.9942,
+      "step": 5068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.684881639232165e-06,
+      "loss": 2.9322,
+      "step": 5069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.682942293801668e-06,
+      "loss": 2.9654,
+      "step": 5070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.681003161262381e-06,
+      "loss": 2.9422,
+      "step": 5071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.679064241927748e-06,
+      "loss": 2.9512,
+      "step": 5072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6771255361111792e-06,
+      "loss": 2.9193,
+      "step": 5073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6751870441260482e-06,
+      "loss": 3.0208,
+      "step": 5074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6732487662856973e-06,
+      "loss": 2.8713,
+      "step": 5075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6713107029034335e-06,
+      "loss": 3.0413,
+      "step": 5076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6693728542925266e-06,
+      "loss": 2.9162,
+      "step": 5077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.667435220766213e-06,
+      "loss": 3.0891,
+      "step": 5078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6654978026376974e-06,
+      "loss": 2.9494,
+      "step": 5079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6635606002201453e-06,
+      "loss": 2.8196,
+      "step": 5080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6616236138266896e-06,
+      "loss": 3.1148,
+      "step": 5081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6596868437704257e-06,
+      "loss": 2.8839,
+      "step": 5082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6577502903644197e-06,
+      "loss": 2.9477,
+      "step": 5083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6558139539216963e-06,
+      "loss": 2.9824,
+      "step": 5084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.653877834755249e-06,
+      "loss": 2.9325,
+      "step": 5085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.651941933178034e-06,
+      "loss": 2.9624,
+      "step": 5086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6500062495029736e-06,
+      "loss": 2.871,
+      "step": 5087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6480707840429545e-06,
+      "loss": 2.9379,
+      "step": 5088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6461355371108294e-06,
+      "loss": 2.9454,
+      "step": 5089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6442005090194142e-06,
+      "loss": 2.9091,
+      "step": 5090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.642265700081491e-06,
+      "loss": 3.0426,
+      "step": 5091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.640331110609803e-06,
+      "loss": 2.876,
+      "step": 5092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6383967409170608e-06,
+      "loss": 2.9816,
+      "step": 5093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.636462591315938e-06,
+      "loss": 2.7753,
+      "step": 5094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6345286621190756e-06,
+      "loss": 2.7943,
+      "step": 5095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6325949536390766e-06,
+      "loss": 2.9667,
+      "step": 5096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.630661466188507e-06,
+      "loss": 3.192,
+      "step": 5097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6287282000798975e-06,
+      "loss": 3.0407,
+      "step": 5098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6267951556257486e-06,
+      "loss": 3.0214,
+      "step": 5099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6248623331385173e-06,
+      "loss": 3.1336,
+      "step": 5100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.622929732930629e-06,
+      "loss": 3.1478,
+      "step": 5101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.620997355314471e-06,
+      "loss": 2.9884,
+      "step": 5102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6190652006023973e-06,
+      "loss": 2.8854,
+      "step": 5103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6171332691067244e-06,
+      "loss": 2.9246,
+      "step": 5104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.615201561139732e-06,
+      "loss": 2.8451,
+      "step": 5105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.613270077013663e-06,
+      "loss": 2.9252,
+      "step": 5106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.611338817040726e-06,
+      "loss": 2.9044,
+      "step": 5107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.609407781533094e-06,
+      "loss": 3.0009,
+      "step": 5108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.607476970802901e-06,
+      "loss": 2.9362,
+      "step": 5109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6055463851622484e-06,
+      "loss": 2.9091,
+      "step": 5110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6036160249231988e-06,
+      "loss": 3.0172,
+      "step": 5111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6016858903977764e-06,
+      "loss": 3.0843,
+      "step": 5112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.599755981897972e-06,
+      "loss": 3.0603,
+      "step": 5113
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5978262997357386e-06,
+      "loss": 3.002,
+      "step": 5114
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5958968442229947e-06,
+      "loss": 2.9803,
+      "step": 5115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5939676156716185e-06,
+      "loss": 3.026,
+      "step": 5116
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5920386143934525e-06,
+      "loss": 2.8409,
+      "step": 5117
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.590109840700307e-06,
+      "loss": 2.8825,
+      "step": 5118
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5881812949039497e-06,
+      "loss": 3.1005,
+      "step": 5119
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5862529773161143e-06,
+      "loss": 2.9229,
+      "step": 5120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5843248882484947e-06,
+      "loss": 3.1133,
+      "step": 5121
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5823970280127542e-06,
+      "loss": 2.8951,
+      "step": 5122
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5804693969205124e-06,
+      "loss": 3.059,
+      "step": 5123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5785419952833546e-06,
+      "loss": 2.8853,
+      "step": 5124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.57661482341283e-06,
+      "loss": 3.1727,
+      "step": 5125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5746878816204476e-06,
+      "loss": 3.0461,
+      "step": 5126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5727611702176823e-06,
+      "loss": 2.8817,
+      "step": 5127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.570834689515971e-06,
+      "loss": 2.8844,
+      "step": 5128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5689084398267144e-06,
+      "loss": 2.9425,
+      "step": 5129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5669824214612725e-06,
+      "loss": 3.0105,
+      "step": 5130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.56505663473097e-06,
+      "loss": 2.8169,
+      "step": 5131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5631310799470942e-06,
+      "loss": 3.0962,
+      "step": 5132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.561205757420894e-06,
+      "loss": 2.9624,
+      "step": 5133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.559280667463583e-06,
+      "loss": 2.9313,
+      "step": 5134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5573558103863347e-06,
+      "loss": 2.7648,
+      "step": 5135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5554311865002866e-06,
+      "loss": 3.0098,
+      "step": 5136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.553506796116535e-06,
+      "loss": 2.9987,
+      "step": 5137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5515826395461455e-06,
+      "loss": 3.0314,
+      "step": 5138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.54965871710014e-06,
+      "loss": 2.9725,
+      "step": 5139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.547735029089503e-06,
+      "loss": 2.9134,
+      "step": 5140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5458115758251843e-06,
+      "loss": 2.9041,
+      "step": 5141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5438883576180927e-06,
+      "loss": 3.0403,
+      "step": 5142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5419653747791006e-06,
+      "loss": 2.8521,
+      "step": 5143
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5400426276190415e-06,
+      "loss": 2.977,
+      "step": 5144
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.53812011644871e-06,
+      "loss": 2.9114,
+      "step": 5145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.536197841578866e-06,
+      "loss": 2.8684,
+      "step": 5146
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.534275803320226e-06,
+      "loss": 2.9006,
+      "step": 5147
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5323540019834753e-06,
+      "loss": 3.056,
+      "step": 5148
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5304324378792544e-06,
+      "loss": 2.81,
+      "step": 5149
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.528511111318168e-06,
+      "loss": 3.0042,
+      "step": 5150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5265900226107826e-06,
+      "loss": 2.9736,
+      "step": 5151
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.524669172067624e-06,
+      "loss": 2.9689,
+      "step": 5152
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.522748559999184e-06,
+      "loss": 3.0038,
+      "step": 5153
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.520828186715913e-06,
+      "loss": 2.886,
+      "step": 5154
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5189080525282215e-06,
+      "loss": 2.9158,
+      "step": 5155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.516988157746484e-06,
+      "loss": 2.8297,
+      "step": 5156
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5150685026810326e-06,
+      "loss": 2.9103,
+      "step": 5157
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.513149087642168e-06,
+      "loss": 2.8886,
+      "step": 5158
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.511229912940144e-06,
+      "loss": 3.1961,
+      "step": 5159
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.50931097888518e-06,
+      "loss": 3.0776,
+      "step": 5160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5073922857874564e-06,
+      "loss": 3.0388,
+      "step": 5161
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5054738339571116e-06,
+      "loss": 2.8419,
+      "step": 5162
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5035556237042485e-06,
+      "loss": 3.0547,
+      "step": 5163
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.5016376553389275e-06,
+      "loss": 3.0915,
+      "step": 5164
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4997199291711755e-06,
+      "loss": 2.9392,
+      "step": 5165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4978024455109727e-06,
+      "loss": 2.9807,
+      "step": 5166
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.495885204668266e-06,
+      "loss": 2.939,
+      "step": 5167
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4939682069529625e-06,
+      "loss": 2.9251,
+      "step": 5168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4920514526749274e-06,
+      "loss": 2.8403,
+      "step": 5169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.490134942143987e-06,
+      "loss": 3.0379,
+      "step": 5170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4882186756699298e-06,
+      "loss": 3.0386,
+      "step": 5171
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4863026535625033e-06,
+      "loss": 3.0593,
+      "step": 5172
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.484386876131417e-06,
+      "loss": 3.0262,
+      "step": 5173
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.48247134368634e-06,
+      "loss": 2.8342,
+      "step": 5174
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4805560565369024e-06,
+      "loss": 2.841,
+      "step": 5175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4786410149926904e-06,
+      "loss": 3.1164,
+      "step": 5176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.47672621936326e-06,
+      "loss": 2.9908,
+      "step": 5177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.474811669958118e-06,
+      "loss": 3.0242,
+      "step": 5178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4728973670867363e-06,
+      "loss": 3.0066,
+      "step": 5179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4709833110585457e-06,
+      "loss": 3.0093,
+      "step": 5180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4690695021829367e-06,
+      "loss": 2.823,
+      "step": 5181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.46715594076926e-06,
+      "loss": 2.9655,
+      "step": 5182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.465242627126828e-06,
+      "loss": 2.7471,
+      "step": 5183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4633295615649095e-06,
+      "loss": 3.0146,
+      "step": 5184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.461416744392737e-06,
+      "loss": 3.0164,
+      "step": 5185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.459504175919499e-06,
+      "loss": 2.8652,
+      "step": 5186
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4575918564543487e-06,
+      "loss": 2.9362,
+      "step": 5187
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4556797863063963e-06,
+      "loss": 2.901,
+      "step": 5188
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.45376796578471e-06,
+      "loss": 3.0058,
+      "step": 5189
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4518563951983207e-06,
+      "loss": 2.9381,
+      "step": 5190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4499450748562152e-06,
+      "loss": 2.963,
+      "step": 5191
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4480340050673453e-06,
+      "loss": 3.0021,
+      "step": 5192
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4461231861406176e-06,
+      "loss": 2.9971,
+      "step": 5193
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4442126183849e-06,
+      "loss": 2.9091,
+      "step": 5194
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.44230230210902e-06,
+      "loss": 3.0956,
+      "step": 5195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4403922376217614e-06,
+      "loss": 3.1998,
+      "step": 5196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4384824252318743e-06,
+      "loss": 2.9683,
+      "step": 5197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.43657286524806e-06,
+      "loss": 3.0045,
+      "step": 5198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.434663557978986e-06,
+      "loss": 2.8249,
+      "step": 5199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4327545037332737e-06,
+      "loss": 3.0131,
+      "step": 5200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.430845702819505e-06,
+      "loss": 2.9035,
+      "step": 5201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4289371555462224e-06,
+      "loss": 3.0149,
+      "step": 5202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4270288622219254e-06,
+      "loss": 2.983,
+      "step": 5203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4251208231550747e-06,
+      "loss": 2.9427,
+      "step": 5204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4232130386540883e-06,
+      "loss": 2.9909,
+      "step": 5205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4213055090273407e-06,
+      "loss": 2.9687,
+      "step": 5206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.419398234583173e-06,
+      "loss": 3.1105,
+      "step": 5207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.417491215629877e-06,
+      "loss": 3.0542,
+      "step": 5208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4155844524757054e-06,
+      "loss": 2.9242,
+      "step": 5209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.413677945428871e-06,
+      "loss": 2.8954,
+      "step": 5210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4117716947975456e-06,
+      "loss": 2.9489,
+      "step": 5211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.409865700889858e-06,
+      "loss": 2.912,
+      "step": 5212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.407959964013895e-06,
+      "loss": 2.8777,
+      "step": 5213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4060544844777032e-06,
+      "loss": 2.989,
+      "step": 5214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.404149262589286e-06,
+      "loss": 3.0093,
+      "step": 5215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4022442986566095e-06,
+      "loss": 3.1034,
+      "step": 5216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4003395929875925e-06,
+      "loss": 2.9162,
+      "step": 5217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3984351458901154e-06,
+      "loss": 2.8885,
+      "step": 5218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3965309576720164e-06,
+      "loss": 2.9565,
+      "step": 5219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.394627028641091e-06,
+      "loss": 3.0503,
+      "step": 5220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3927233591050925e-06,
+      "loss": 3.0401,
+      "step": 5221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.390819949371732e-06,
+      "loss": 2.9813,
+      "step": 5222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.388916799748682e-06,
+      "loss": 2.9302,
+      "step": 5223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3870139105435696e-06,
+      "loss": 2.9817,
+      "step": 5224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3851112820639786e-06,
+      "loss": 2.8528,
+      "step": 5225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.383208914617456e-06,
+      "loss": 2.8777,
+      "step": 5226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.381306808511503e-06,
+      "loss": 3.0104,
+      "step": 5227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3794049640535774e-06,
+      "loss": 2.9298,
+      "step": 5228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3775033815510955e-06,
+      "loss": 2.8994,
+      "step": 5229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.375602061311435e-06,
+      "loss": 2.9109,
+      "step": 5230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.373701003641926e-06,
+      "loss": 3.0364,
+      "step": 5231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.371800208849858e-06,
+      "loss": 2.916,
+      "step": 5232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3698996772424802e-06,
+      "loss": 2.9297,
+      "step": 5233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3679994091269953e-06,
+      "loss": 2.9143,
+      "step": 5234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3660994048105663e-06,
+      "loss": 2.979,
+      "step": 5235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.364199664600314e-06,
+      "loss": 2.9996,
+      "step": 5236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3623001888033136e-06,
+      "loss": 3.0111,
+      "step": 5237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.360400977726601e-06,
+      "loss": 2.8523,
+      "step": 5238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3585020316771664e-06,
+      "loss": 3.0864,
+      "step": 5239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.356603350961959e-06,
+      "loss": 3.0905,
+      "step": 5240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3547049358878837e-06,
+      "loss": 3.0983,
+      "step": 5241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.352806786761802e-06,
+      "loss": 2.8795,
+      "step": 5242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3509089038905363e-06,
+      "loss": 2.9591,
+      "step": 5243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.349011287580862e-06,
+      "loss": 2.8852,
+      "step": 5244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.347113938139511e-06,
+      "loss": 3.0153,
+      "step": 5245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3452168558731775e-06,
+      "loss": 3.0318,
+      "step": 5246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3433200410885065e-06,
+      "loss": 3.0218,
+      "step": 5247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.341423494092102e-06,
+      "loss": 2.7121,
+      "step": 5248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3395272151905245e-06,
+      "loss": 2.7251,
+      "step": 5249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3376312046902926e-06,
+      "loss": 2.8202,
+      "step": 5250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.33573546289788e-06,
+      "loss": 2.9062,
+      "step": 5251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3338399901197164e-06,
+      "loss": 3.0458,
+      "step": 5252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3319447866621903e-06,
+      "loss": 3.0054,
+      "step": 5253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3300498528316427e-06,
+      "loss": 2.8005,
+      "step": 5254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.328155188934375e-06,
+      "loss": 2.954,
+      "step": 5255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.326260795276644e-06,
+      "loss": 3.0022,
+      "step": 5256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3243666721646633e-06,
+      "loss": 2.9699,
+      "step": 5257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.322472819904601e-06,
+      "loss": 2.855,
+      "step": 5258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3205792388025814e-06,
+      "loss": 2.9589,
+      "step": 5259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3186859291646867e-06,
+      "loss": 2.9891,
+      "step": 5260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3167928912969527e-06,
+      "loss": 2.8837,
+      "step": 5261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.314900125505376e-06,
+      "loss": 3.0357,
+      "step": 5262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3130076320959036e-06,
+      "loss": 2.9759,
+      "step": 5263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3111154113744403e-06,
+      "loss": 2.907,
+      "step": 5264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3092234636468512e-06,
+      "loss": 2.9094,
+      "step": 5265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3073317892189518e-06,
+      "loss": 3.0237,
+      "step": 5266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3054403883965136e-06,
+      "loss": 2.9795,
+      "step": 5267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.303549261485267e-06,
+      "loss": 2.9037,
+      "step": 5268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3016584087908967e-06,
+      "loss": 2.9243,
+      "step": 5269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2997678306190427e-06,
+      "loss": 2.8619,
+      "step": 5270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2978775272753015e-06,
+      "loss": 2.9293,
+      "step": 5271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.295987499065224e-06,
+      "loss": 3.137,
+      "step": 5272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2940977462943164e-06,
+      "loss": 3.0588,
+      "step": 5273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2922082692680413e-06,
+      "loss": 2.9622,
+      "step": 5274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2903190682918183e-06,
+      "loss": 3.069,
+      "step": 5275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2884301436710207e-06,
+      "loss": 2.9715,
+      "step": 5276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.286541495710976e-06,
+      "loss": 2.9008,
+      "step": 5277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2846531247169688e-06,
+      "loss": 2.9655,
+      "step": 5278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2827650309942372e-06,
+      "loss": 2.999,
+      "step": 5279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.280877214847976e-06,
+      "loss": 3.0431,
+      "step": 5280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.278989676583335e-06,
+      "loss": 2.8783,
+      "step": 5281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.27710241650542e-06,
+      "loss": 2.8901,
+      "step": 5282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.275215434919288e-06,
+      "loss": 3.0985,
+      "step": 5283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2733287321299536e-06,
+      "loss": 2.9015,
+      "step": 5284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2714423084423895e-06,
+      "loss": 2.9787,
+      "step": 5285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2695561641615176e-06,
+      "loss": 3.0074,
+      "step": 5286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.267670299592217e-06,
+      "loss": 2.958,
+      "step": 5287
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.265784715039323e-06,
+      "loss": 3.0081,
+      "step": 5288
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.263899410807624e-06,
+      "loss": 2.968,
+      "step": 5289
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2620143872018633e-06,
+      "loss": 3.0945,
+      "step": 5290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2601296445267388e-06,
+      "loss": 2.8455,
+      "step": 5291
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2582451830869023e-06,
+      "loss": 2.8837,
+      "step": 5292
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2563610031869626e-06,
+      "loss": 2.8875,
+      "step": 5293
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2544771051314783e-06,
+      "loss": 3.0232,
+      "step": 5294
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.252593489224971e-06,
+      "loss": 2.972,
+      "step": 5295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2507101557719076e-06,
+      "loss": 3.018,
+      "step": 5296
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2488271050767138e-06,
+      "loss": 2.9629,
+      "step": 5297
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.246944337443768e-06,
+      "loss": 2.9994,
+      "step": 5298
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2450618531774043e-06,
+      "loss": 2.854,
+      "step": 5299
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2431796525819104e-06,
+      "loss": 2.9844,
+      "step": 5300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2412977359615282e-06,
+      "loss": 3.0047,
+      "step": 5301
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.239416103620453e-06,
+      "loss": 2.9896,
+      "step": 5302
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.237534755862833e-06,
+      "loss": 2.9324,
+      "step": 5303
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2356536929927763e-06,
+      "loss": 2.9738,
+      "step": 5304
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.233772915314338e-06,
+      "loss": 2.7991,
+      "step": 5305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2318924231315295e-06,
+      "loss": 2.8656,
+      "step": 5306
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2300122167483183e-06,
+      "loss": 2.9989,
+      "step": 5307
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2281322964686234e-06,
+      "loss": 3.0348,
+      "step": 5308
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2262526625963162e-06,
+      "loss": 3.064,
+      "step": 5309
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2243733154352266e-06,
+      "loss": 3.0039,
+      "step": 5310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.222494255289131e-06,
+      "loss": 2.8763,
+      "step": 5311
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.220615482461767e-06,
+      "loss": 2.9911,
+      "step": 5312
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.21873699725682e-06,
+      "loss": 2.9684,
+      "step": 5313
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2168587999779324e-06,
+      "loss": 2.9749,
+      "step": 5314
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2149808909287e-06,
+      "loss": 2.8183,
+      "step": 5315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.21310327041267e-06,
+      "loss": 2.9155,
+      "step": 5316
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2112259387333426e-06,
+      "loss": 2.9859,
+      "step": 5317
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2093488961941745e-06,
+      "loss": 2.9607,
+      "step": 5318
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2074721430985706e-06,
+      "loss": 2.9269,
+      "step": 5319
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2055956797498956e-06,
+      "loss": 2.9155,
+      "step": 5320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2037195064514624e-06,
+      "loss": 2.9709,
+      "step": 5321
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.2018436235065388e-06,
+      "loss": 2.8831,
+      "step": 5322
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1999680312183427e-06,
+      "loss": 2.9001,
+      "step": 5323
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1980927298900526e-06,
+      "loss": 3.1102,
+      "step": 5324
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1962177198247924e-06,
+      "loss": 3.014,
+      "step": 5325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.194343001325641e-06,
+      "loss": 2.9864,
+      "step": 5326
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1924685746956322e-06,
+      "loss": 3.0932,
+      "step": 5327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.190594440237751e-06,
+      "loss": 2.8419,
+      "step": 5328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.188720598254934e-06,
+      "loss": 2.9367,
+      "step": 5329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1868470490500736e-06,
+      "loss": 3.0048,
+      "step": 5330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1849737929260105e-06,
+      "loss": 3.0178,
+      "step": 5331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1831008301855437e-06,
+      "loss": 2.9919,
+      "step": 5332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.181228161131418e-06,
+      "loss": 3.01,
+      "step": 5333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.17935578606634e-06,
+      "loss": 3.0332,
+      "step": 5334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.177483705292959e-06,
+      "loss": 2.9739,
+      "step": 5335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.175611919113883e-06,
+      "loss": 2.9924,
+      "step": 5336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.173740427831669e-06,
+      "loss": 3.0159,
+      "step": 5337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1718692317488275e-06,
+      "loss": 2.9781,
+      "step": 5338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1699983311678228e-06,
+      "loss": 2.9349,
+      "step": 5339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1681277263910702e-06,
+      "loss": 2.9855,
+      "step": 5340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.166257417720936e-06,
+      "loss": 3.1425,
+      "step": 5341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.16438740545974e-06,
+      "loss": 2.8374,
+      "step": 5342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1625176899097522e-06,
+      "loss": 3.0158,
+      "step": 5343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1606482713731995e-06,
+      "loss": 3.0246,
+      "step": 5344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.158779150152256e-06,
+      "loss": 2.9622,
+      "step": 5345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.15691032654905e-06,
+      "loss": 3.0449,
+      "step": 5346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.155041800865659e-06,
+      "loss": 2.9637,
+      "step": 5347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1531735734041176e-06,
+      "loss": 2.9288,
+      "step": 5348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.151305644466406e-06,
+      "loss": 3.0962,
+      "step": 5349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1494380143544595e-06,
+      "loss": 2.9446,
+      "step": 5350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1475706833701663e-06,
+      "loss": 2.9364,
+      "step": 5351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1457036518153616e-06,
+      "loss": 2.8159,
+      "step": 5352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1438369199918384e-06,
+      "loss": 2.8881,
+      "step": 5353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.141970488201338e-06,
+      "loss": 2.9781,
+      "step": 5354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1401043567455516e-06,
+      "loss": 2.9568,
+      "step": 5355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.138238525926124e-06,
+      "loss": 2.917,
+      "step": 5356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1363729960446497e-06,
+      "loss": 3.0101,
+      "step": 5357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.134507767402679e-06,
+      "loss": 2.9553,
+      "step": 5358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.132642840301708e-06,
+      "loss": 2.9456,
+      "step": 5359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1307782150431864e-06,
+      "loss": 2.9792,
+      "step": 5360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1289138919285157e-06,
+      "loss": 2.9579,
+      "step": 5361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1270498712590456e-06,
+      "loss": 3.0551,
+      "step": 5362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1251861533360826e-06,
+      "loss": 2.9748,
+      "step": 5363
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.12332273846088e-06,
+      "loss": 2.7792,
+      "step": 5364
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.121459626934642e-06,
+      "loss": 2.8862,
+      "step": 5365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1195968190585263e-06,
+      "loss": 3.0107,
+      "step": 5366
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.117734315133639e-06,
+      "loss": 2.9514,
+      "step": 5367
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1158721154610373e-06,
+      "loss": 2.9326,
+      "step": 5368
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1140102203417303e-06,
+      "loss": 3.1398,
+      "step": 5369
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.112148630076679e-06,
+      "loss": 2.9623,
+      "step": 5370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.110287344966793e-06,
+      "loss": 2.9323,
+      "step": 5371
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.108426365312931e-06,
+      "loss": 2.9348,
+      "step": 5372
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.106565691415908e-06,
+      "loss": 2.9335,
+      "step": 5373
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1047053235764847e-06,
+      "loss": 2.8684,
+      "step": 5374
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.102845262095374e-06,
+      "loss": 2.9539,
+      "step": 5375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.1009855072732385e-06,
+      "loss": 2.9421,
+      "step": 5376
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0991260594106927e-06,
+      "loss": 2.9259,
+      "step": 5377
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0972669188082994e-06,
+      "loss": 2.9335,
+      "step": 5378
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0954080857665746e-06,
+      "loss": 2.872,
+      "step": 5379
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.093549560585981e-06,
+      "loss": 2.9725,
+      "step": 5380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.091691343566934e-06,
+      "loss": 3.0777,
+      "step": 5381
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.089833435009798e-06,
+      "loss": 2.9627,
+      "step": 5382
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.087975835214891e-06,
+      "loss": 3.073,
+      "step": 5383
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.086118544482475e-06,
+      "loss": 3.0376,
+      "step": 5384
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.084261563112767e-06,
+      "loss": 2.9819,
+      "step": 5385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0824048914059325e-06,
+      "loss": 3.1014,
+      "step": 5386
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0805485296620864e-06,
+      "loss": 2.9899,
+      "step": 5387
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0786924781812933e-06,
+      "loss": 2.8837,
+      "step": 5388
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0768367372635675e-06,
+      "loss": 3.0699,
+      "step": 5389
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.074981307208875e-06,
+      "loss": 3.0544,
+      "step": 5390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0731261883171317e-06,
+      "loss": 2.8358,
+      "step": 5391
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0712713808881967e-06,
+      "loss": 2.9227,
+      "step": 5392
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0694168852218904e-06,
+      "loss": 3.1225,
+      "step": 5393
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0675627016179734e-06,
+      "loss": 2.9333,
+      "step": 5394
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0657088303761574e-06,
+      "loss": 2.9275,
+      "step": 5395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.063855271796106e-06,
+      "loss": 2.8528,
+      "step": 5396
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.062002026177433e-06,
+      "loss": 2.8817,
+      "step": 5397
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0601490938196977e-06,
+      "loss": 2.9439,
+      "step": 5398
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0582964750224113e-06,
+      "loss": 2.8567,
+      "step": 5399
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0564441700850345e-06,
+      "loss": 3.0046,
+      "step": 5400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0545921793069745e-06,
+      "loss": 3.0613,
+      "step": 5401
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0527405029875936e-06,
+      "loss": 2.9663,
+      "step": 5402
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.050889141426196e-06,
+      "loss": 3.1207,
+      "step": 5403
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0490380949220422e-06,
+      "loss": 2.9199,
+      "step": 5404
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0471873637743353e-06,
+      "loss": 3.041,
+      "step": 5405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.045336948282232e-06,
+      "loss": 2.8896,
+      "step": 5406
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0434868487448354e-06,
+      "loss": 2.8629,
+      "step": 5407
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.041637065461197e-06,
+      "loss": 2.9301,
+      "step": 5408
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0397875987303218e-06,
+      "loss": 2.9895,
+      "step": 5409
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0379384488511587e-06,
+      "loss": 3.084,
+      "step": 5410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0360896161226043e-06,
+      "loss": 3.0503,
+      "step": 5411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0342411008435124e-06,
+      "loss": 2.9255,
+      "step": 5412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0323929033126764e-06,
+      "loss": 2.9995,
+      "step": 5413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0305450238288425e-06,
+      "loss": 2.9987,
+      "step": 5414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.028697462690704e-06,
+      "loss": 2.8894,
+      "step": 5415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0268502201969047e-06,
+      "loss": 2.8925,
+      "step": 5416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0250032966460354e-06,
+      "loss": 3.0305,
+      "step": 5417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0231566923366348e-06,
+      "loss": 3.1495,
+      "step": 5418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.021310407567191e-06,
+      "loss": 2.9412,
+      "step": 5419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0194644426361404e-06,
+      "loss": 3.155,
+      "step": 5420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.017618797841867e-06,
+      "loss": 2.9338,
+      "step": 5421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.015773473482705e-06,
+      "loss": 3.0113,
+      "step": 5422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.013928469856936e-06,
+      "loss": 3.1025,
+      "step": 5423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0120837872627874e-06,
+      "loss": 3.0074,
+      "step": 5424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.010239425998437e-06,
+      "loss": 2.8408,
+      "step": 5425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0083953863620107e-06,
+      "loss": 2.9449,
+      "step": 5426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0065516686515807e-06,
+      "loss": 2.975,
+      "step": 5427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0047082731651694e-06,
+      "loss": 3.009,
+      "step": 5428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.0028652002007464e-06,
+      "loss": 2.9681,
+      "step": 5429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.001022450056228e-06,
+      "loss": 2.9368,
+      "step": 5430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9991800230294764e-06,
+      "loss": 2.9629,
+      "step": 5431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.99733791941831e-06,
+      "loss": 2.9012,
+      "step": 5432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.995496139520486e-06,
+      "loss": 3.0442,
+      "step": 5433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9936546836337122e-06,
+      "loss": 3.0399,
+      "step": 5434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.991813552055646e-06,
+      "loss": 2.8508,
+      "step": 5435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.98997274508389e-06,
+      "loss": 3.1104,
+      "step": 5436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9881322630159946e-06,
+      "loss": 3.1509,
+      "step": 5437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.986292106149458e-06,
+      "loss": 3.0551,
+      "step": 5438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9844522747817257e-06,
+      "loss": 2.9524,
+      "step": 5439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9826127692101917e-06,
+      "loss": 2.9796,
+      "step": 5440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.980773589732195e-06,
+      "loss": 2.9881,
+      "step": 5441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.978934736645026e-06,
+      "loss": 2.8824,
+      "step": 5442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.977096210245919e-06,
+      "loss": 2.9733,
+      "step": 5443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.975258010832054e-06,
+      "loss": 2.8827,
+      "step": 5444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.973420138700563e-06,
+      "loss": 2.941,
+      "step": 5445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.97158259414852e-06,
+      "loss": 3.0449,
+      "step": 5446
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.96974537747295e-06,
+      "loss": 2.905,
+      "step": 5447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.967908488970824e-06,
+      "loss": 2.9624,
+      "step": 5448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.966071928939059e-06,
+      "loss": 2.7731,
+      "step": 5449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.964235697674516e-06,
+      "loss": 3.0433,
+      "step": 5450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.962399795474012e-06,
+      "loss": 2.9597,
+      "step": 5451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9605642226343024e-06,
+      "loss": 3.1229,
+      "step": 5452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.95872897945209e-06,
+      "loss": 2.9485,
+      "step": 5453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.95689406622403e-06,
+      "loss": 3.1026,
+      "step": 5454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9550594832467182e-06,
+      "loss": 2.9153,
+      "step": 5455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9532252308166996e-06,
+      "loss": 3.0655,
+      "step": 5456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.951391309230467e-06,
+      "loss": 2.9386,
+      "step": 5457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.949557718784455e-06,
+      "loss": 2.8992,
+      "step": 5458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9477244597750513e-06,
+      "loss": 3.1368,
+      "step": 5459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.945891532498585e-06,
+      "loss": 2.9612,
+      "step": 5460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.944058937251333e-06,
+      "loss": 3.0654,
+      "step": 5461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9422266743295208e-06,
+      "loss": 3.0278,
+      "step": 5462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.940394744029317e-06,
+      "loss": 2.9622,
+      "step": 5463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9385631466468368e-06,
+      "loss": 3.088,
+      "step": 5464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9367318824781438e-06,
+      "loss": 2.9479,
+      "step": 5465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.934900951819244e-06,
+      "loss": 2.8243,
+      "step": 5466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.933070354966094e-06,
+      "loss": 2.9506,
+      "step": 5467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9312400922145947e-06,
+      "loss": 2.9402,
+      "step": 5468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.929410163860591e-06,
+      "loss": 2.8629,
+      "step": 5469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9275805701998737e-06,
+      "loss": 2.9149,
+      "step": 5470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9257513115281853e-06,
+      "loss": 2.9492,
+      "step": 5471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.923922388141207e-06,
+      "loss": 2.9793,
+      "step": 5472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.92209380033457e-06,
+      "loss": 2.9382,
+      "step": 5473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9202655484038494e-06,
+      "loss": 3.025,
+      "step": 5474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9184376326445663e-06,
+      "loss": 2.8545,
+      "step": 5475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.91661005335219e-06,
+      "loss": 2.9843,
+      "step": 5476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.91478281082213e-06,
+      "loss": 2.9738,
+      "step": 5477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.912955905349748e-06,
+      "loss": 2.9268,
+      "step": 5478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9111293372303434e-06,
+      "loss": 2.8136,
+      "step": 5479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.909303106759168e-06,
+      "loss": 3.1069,
+      "step": 5480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.907477214231416e-06,
+      "loss": 3.0567,
+      "step": 5481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9056516599422285e-06,
+      "loss": 3.0016,
+      "step": 5482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9038264441866914e-06,
+      "loss": 2.8445,
+      "step": 5483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9020015672598326e-06,
+      "loss": 2.9892,
+      "step": 5484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.900177029456632e-06,
+      "loss": 2.9307,
+      "step": 5485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8983528310720056e-06,
+      "loss": 2.8948,
+      "step": 5486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8965289724008217e-06,
+      "loss": 3.053,
+      "step": 5487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8947054537378942e-06,
+      "loss": 3.0589,
+      "step": 5488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.892882275377976e-06,
+      "loss": 2.8068,
+      "step": 5489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8910594376157687e-06,
+      "loss": 2.9235,
+      "step": 5490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8892369407459196e-06,
+      "loss": 3.1146,
+      "step": 5491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.887414785063023e-06,
+      "loss": 3.1007,
+      "step": 5492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8855929708616088e-06,
+      "loss": 2.9751,
+      "step": 5493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.883771498436163e-06,
+      "loss": 2.8694,
+      "step": 5494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.881950368081107e-06,
+      "loss": 2.9878,
+      "step": 5495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8801295800908124e-06,
+      "loss": 2.9359,
+      "step": 5496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.878309134759597e-06,
+      "loss": 3.0353,
+      "step": 5497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.876489032381716e-06,
+      "loss": 2.9336,
+      "step": 5498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.874669273251376e-06,
+      "loss": 2.8998,
+      "step": 5499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.872849857662725e-06,
+      "loss": 3.1473,
+      "step": 5500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8710307859098585e-06,
+      "loss": 3.028,
+      "step": 5501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.86921205828681e-06,
+      "loss": 2.8747,
+      "step": 5502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.867393675087563e-06,
+      "loss": 2.986,
+      "step": 5503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.865575636606046e-06,
+      "loss": 3.0797,
+      "step": 5504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.863757943136125e-06,
+      "loss": 2.874,
+      "step": 5505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8619405949716206e-06,
+      "loss": 2.9626,
+      "step": 5506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8601235924062853e-06,
+      "loss": 3.0303,
+      "step": 5507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.858306935733825e-06,
+      "loss": 3.0336,
+      "step": 5508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8564906252478874e-06,
+      "loss": 2.8669,
+      "step": 5509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.854674661242063e-06,
+      "loss": 3.0141,
+      "step": 5510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8528590440098903e-06,
+      "loss": 3.011,
+      "step": 5511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8510437738448423e-06,
+      "loss": 2.9706,
+      "step": 5512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8492288510403475e-06,
+      "loss": 3.0325,
+      "step": 5513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.847414275889768e-06,
+      "loss": 2.9295,
+      "step": 5514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.845600048686418e-06,
+      "loss": 2.8583,
+      "step": 5515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8437861697235524e-06,
+      "loss": 2.9407,
+      "step": 5516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.841972639294366e-06,
+      "loss": 3.0661,
+      "step": 5517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.840159457692004e-06,
+      "loss": 2.8993,
+      "step": 5518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8383466252095467e-06,
+      "loss": 2.9363,
+      "step": 5519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8365341421400305e-06,
+      "loss": 3.0128,
+      "step": 5520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8347220087764217e-06,
+      "loss": 3.1055,
+      "step": 5521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.832910225411639e-06,
+      "loss": 2.9385,
+      "step": 5522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8310987923385436e-06,
+      "loss": 2.8508,
+      "step": 5523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.829287709849934e-06,
+      "loss": 2.9619,
+      "step": 5524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8274769782385606e-06,
+      "loss": 2.9146,
+      "step": 5525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8256665977971086e-06,
+      "loss": 3.0806,
+      "step": 5526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8238565688182133e-06,
+      "loss": 2.8572,
+      "step": 5527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.822046891594452e-06,
+      "loss": 2.8894,
+      "step": 5528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8202375664183367e-06,
+      "loss": 2.9701,
+      "step": 5529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8184285935823396e-06,
+      "loss": 2.9058,
+      "step": 5530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.816619973378858e-06,
+      "loss": 2.9115,
+      "step": 5531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.814811706100245e-06,
+      "loss": 2.8255,
+      "step": 5532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8130037920387875e-06,
+      "loss": 2.938,
+      "step": 5533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8111962314867215e-06,
+      "loss": 3.033,
+      "step": 5534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8093890247362247e-06,
+      "loss": 2.9292,
+      "step": 5535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8075821720794146e-06,
+      "loss": 2.9035,
+      "step": 5536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.805775673808356e-06,
+      "loss": 2.9413,
+      "step": 5537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8039695302150494e-06,
+      "loss": 3.0303,
+      "step": 5538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8021637415914498e-06,
+      "loss": 2.8633,
+      "step": 5539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8003583082294416e-06,
+      "loss": 2.8528,
+      "step": 5540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7985532304208608e-06,
+      "loss": 2.8651,
+      "step": 5541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.796748508457483e-06,
+      "loss": 2.8011,
+      "step": 5542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7949441426310242e-06,
+      "loss": 2.7791,
+      "step": 5543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7931401332331477e-06,
+      "loss": 2.8705,
+      "step": 5544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.791336480555452e-06,
+      "loss": 2.9253,
+      "step": 5545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7895331848894848e-06,
+      "loss": 3.0234,
+      "step": 5546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7877302465267366e-06,
+      "loss": 3.1311,
+      "step": 5547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7859276657586287e-06,
+      "loss": 2.9092,
+      "step": 5548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.784125442876543e-06,
+      "loss": 3.04,
+      "step": 5549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7823235781717874e-06,
+      "loss": 2.7972,
+      "step": 5550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7805220719356207e-06,
+      "loss": 2.9313,
+      "step": 5551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7787209244592382e-06,
+      "loss": 2.8645,
+      "step": 5552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.776920136033782e-06,
+      "loss": 2.905,
+      "step": 5553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7751197069503367e-06,
+      "loss": 3.0026,
+      "step": 5554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7733196374999207e-06,
+      "loss": 2.956,
+      "step": 5555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.771519927973506e-06,
+      "loss": 3.1102,
+      "step": 5556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.769720578661995e-06,
+      "loss": 2.8801,
+      "step": 5557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.76792158985624e-06,
+      "loss": 2.9055,
+      "step": 5558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.766122961847032e-06,
+      "loss": 2.7344,
+      "step": 5559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.764324694925104e-06,
+      "loss": 2.9495,
+      "step": 5560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7625267893811325e-06,
+      "loss": 2.9153,
+      "step": 5561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.76072924550573e-06,
+      "loss": 2.8967,
+      "step": 5562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7589320635894578e-06,
+      "loss": 2.9712,
+      "step": 5563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7571352439228118e-06,
+      "loss": 2.8942,
+      "step": 5564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7553387867962334e-06,
+      "loss": 2.9641,
+      "step": 5565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7535426925001076e-06,
+      "loss": 2.9784,
+      "step": 5566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.751746961324754e-06,
+      "loss": 2.9441,
+      "step": 5567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7499515935604393e-06,
+      "loss": 2.893,
+      "step": 5568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.748156589497368e-06,
+      "loss": 2.8107,
+      "step": 5569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7463619494256917e-06,
+      "loss": 3.0874,
+      "step": 5570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.744567673635493e-06,
+      "loss": 2.954,
+      "step": 5571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.742773762416804e-06,
+      "loss": 3.0076,
+      "step": 5572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7409802160595967e-06,
+      "loss": 2.9832,
+      "step": 5573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7391870348537793e-06,
+      "loss": 3.0783,
+      "step": 5574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.737394219089208e-06,
+      "loss": 2.8205,
+      "step": 5575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7356017690556726e-06,
+      "loss": 3.0672,
+      "step": 5576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7338096850429084e-06,
+      "loss": 3.0612,
+      "step": 5577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7320179673405923e-06,
+      "loss": 2.9813,
+      "step": 5578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7302266162383383e-06,
+      "loss": 3.0242,
+      "step": 5579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.728435632025707e-06,
+      "loss": 2.9677,
+      "step": 5580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.726645014992191e-06,
+      "loss": 2.9404,
+      "step": 5581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.724854765427232e-06,
+      "loss": 2.9568,
+      "step": 5582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7230648836202056e-06,
+      "loss": 2.795,
+      "step": 5583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7212753698604317e-06,
+      "loss": 2.9538,
+      "step": 5584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.719486224437173e-06,
+      "loss": 3.0152,
+      "step": 5585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7176974476396255e-06,
+      "loss": 3.0418,
+      "step": 5586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7159090397569315e-06,
+      "loss": 3.0558,
+      "step": 5587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7141210010781725e-06,
+      "loss": 3.0989,
+      "step": 5588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.712333331892372e-06,
+      "loss": 3.1007,
+      "step": 5589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.710546032488486e-06,
+      "loss": 2.9091,
+      "step": 5590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7087591031554193e-06,
+      "loss": 2.9583,
+      "step": 5591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.706972544182017e-06,
+      "loss": 2.928,
+      "step": 5592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.705186355857055e-06,
+      "loss": 2.9983,
+      "step": 5593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7034005384692608e-06,
+      "loss": 3.0707,
+      "step": 5594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.701615092307293e-06,
+      "loss": 2.9364,
+      "step": 5595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6998300176597547e-06,
+      "loss": 2.9678,
+      "step": 5596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6980453148151897e-06,
+      "loss": 2.9135,
+      "step": 5597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.696260984062078e-06,
+      "loss": 3.1615,
+      "step": 5598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.694477025688845e-06,
+      "loss": 3.0435,
+      "step": 5599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.692693439983848e-06,
+      "loss": 2.9916,
+      "step": 5600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6909102272353924e-06,
+      "loss": 2.9341,
+      "step": 5601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.689127387731716e-06,
+      "loss": 2.9176,
+      "step": 5602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6873449217610005e-06,
+      "loss": 2.9549,
+      "step": 5603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6855628296113697e-06,
+      "loss": 2.7821,
+      "step": 5604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.683781111570879e-06,
+      "loss": 2.972,
+      "step": 5605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6819997679275316e-06,
+      "loss": 2.7528,
+      "step": 5606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.680218798969262e-06,
+      "loss": 3.0182,
+      "step": 5607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.678438204983955e-06,
+      "loss": 2.9103,
+      "step": 5608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6766579862594246e-06,
+      "loss": 2.9227,
+      "step": 5609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6748781430834277e-06,
+      "loss": 3.0844,
+      "step": 5610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.673098675743664e-06,
+      "loss": 2.998,
+      "step": 5611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6713195845277664e-06,
+      "loss": 3.0073,
+      "step": 5612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6695408697233114e-06,
+      "loss": 3.064,
+      "step": 5613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6677625316178113e-06,
+      "loss": 3.0759,
+      "step": 5614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.665984570498721e-06,
+      "loss": 2.9094,
+      "step": 5615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.664206986653434e-06,
+      "loss": 2.9848,
+      "step": 5616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.662429780369277e-06,
+      "loss": 2.9817,
+      "step": 5617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.660652951933526e-06,
+      "loss": 2.9329,
+      "step": 5618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.658876501633386e-06,
+      "loss": 3.0012,
+      "step": 5619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.65710042975601e-06,
+      "loss": 2.9309,
+      "step": 5620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6553247365884793e-06,
+      "loss": 3.0365,
+      "step": 5621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6535494224178215e-06,
+      "loss": 3.07,
+      "step": 5622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.651774487531004e-06,
+      "loss": 2.9505,
+      "step": 5623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.649999932214926e-06,
+      "loss": 3.0646,
+      "step": 5624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6482257567564328e-06,
+      "loss": 3.063,
+      "step": 5625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6464519614423012e-06,
+      "loss": 2.9549,
+      "step": 5626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6446785465592524e-06,
+      "loss": 2.8939,
+      "step": 5627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.642905512393943e-06,
+      "loss": 3.0538,
+      "step": 5628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6411328592329695e-06,
+      "loss": 2.9868,
+      "step": 5629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.639360587362868e-06,
+      "loss": 2.9992,
+      "step": 5630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6375886970701076e-06,
+      "loss": 2.8273,
+      "step": 5631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.635817188641103e-06,
+      "loss": 2.8891,
+      "step": 5632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6340460623622e-06,
+      "loss": 2.8168,
+      "step": 5633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.632275318519687e-06,
+      "loss": 2.8587,
+      "step": 5634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6305049573997933e-06,
+      "loss": 3.0601,
+      "step": 5635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6287349792886776e-06,
+      "loss": 2.887,
+      "step": 5636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6269653844724443e-06,
+      "loss": 3.1309,
+      "step": 5637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6251961732371335e-06,
+      "loss": 3.0206,
+      "step": 5638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6234273458687248e-06,
+      "loss": 2.9141,
+      "step": 5639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.62165890265313e-06,
+      "loss": 3.0568,
+      "step": 5640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.619890843876208e-06,
+      "loss": 2.8332,
+      "step": 5641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6181231698237464e-06,
+      "loss": 3.0336,
+      "step": 5642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.616355880781475e-06,
+      "loss": 3.0065,
+      "step": 5643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6145889770350647e-06,
+      "loss": 2.9585,
+      "step": 5644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6128224588701156e-06,
+      "loss": 2.8811,
+      "step": 5645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6110563265721733e-06,
+      "loss": 3.0289,
+      "step": 5646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6092905804267177e-06,
+      "loss": 2.9632,
+      "step": 5647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6075252207191683e-06,
+      "loss": 3.0164,
+      "step": 5648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.605760247734877e-06,
+      "loss": 2.9135,
+      "step": 5649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.6039956617591387e-06,
+      "loss": 2.9106,
+      "step": 5650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.602231463077185e-06,
+      "loss": 2.9975,
+      "step": 5651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.600467651974181e-06,
+      "loss": 3.0548,
+      "step": 5652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.598704228735235e-06,
+      "loss": 2.9169,
+      "step": 5653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.596941193645385e-06,
+      "loss": 2.8456,
+      "step": 5654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.595178546989614e-06,
+      "loss": 2.8824,
+      "step": 5655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5934162890528368e-06,
+      "loss": 3.0352,
+      "step": 5656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.59165442011991e-06,
+      "loss": 2.8945,
+      "step": 5657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5898929404756244e-06,
+      "loss": 3.0712,
+      "step": 5658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.588131850404706e-06,
+      "loss": 2.932,
+      "step": 5659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.586371150191823e-06,
+      "loss": 2.9967,
+      "step": 5660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.584610840121575e-06,
+      "loss": 2.9975,
+      "step": 5661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5828509204785013e-06,
+      "loss": 2.9397,
+      "step": 5662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5810913915470816e-06,
+      "loss": 2.8549,
+      "step": 5663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.579332253611724e-06,
+      "loss": 3.0201,
+      "step": 5664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.577573506956782e-06,
+      "loss": 2.7872,
+      "step": 5665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5758151518665363e-06,
+      "loss": 2.9368,
+      "step": 5666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5740571886252193e-06,
+      "loss": 2.9363,
+      "step": 5667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.572299617516983e-06,
+      "loss": 3.0184,
+      "step": 5668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.570542438825927e-06,
+      "loss": 3.0302,
+      "step": 5669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5687856528360854e-06,
+      "loss": 3.0121,
+      "step": 5670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.567029259831425e-06,
+      "loss": 2.9274,
+      "step": 5671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.565273260095855e-06,
+      "loss": 3.0662,
+      "step": 5672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5635176539132147e-06,
+      "loss": 2.9033,
+      "step": 5673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5617624415672836e-06,
+      "loss": 2.9723,
+      "step": 5674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.56000762334178e-06,
+      "loss": 2.9232,
+      "step": 5675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5582531995203496e-06,
+      "loss": 2.894,
+      "step": 5676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5564991703865864e-06,
+      "loss": 3.0465,
+      "step": 5677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5547455362240104e-06,
+      "loss": 2.9098,
+      "step": 5678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5529922973160838e-06,
+      "loss": 3.0244,
+      "step": 5679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5512394539462003e-06,
+      "loss": 2.9372,
+      "step": 5680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.549487006397694e-06,
+      "loss": 3.0435,
+      "step": 5681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.547734954953834e-06,
+      "loss": 2.9776,
+      "step": 5682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5459832998978222e-06,
+      "loss": 3.071,
+      "step": 5683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.544232041512802e-06,
+      "loss": 2.9395,
+      "step": 5684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5424811800818426e-06,
+      "loss": 2.9149,
+      "step": 5685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5407307158879656e-06,
+      "loss": 3.0523,
+      "step": 5686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.538980649214111e-06,
+      "loss": 3.0385,
+      "step": 5687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5372309803431663e-06,
+      "loss": 3.0015,
+      "step": 5688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5354817095579505e-06,
+      "loss": 2.9227,
+      "step": 5689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.533732837141215e-06,
+      "loss": 3.0849,
+      "step": 5690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.531984363375655e-06,
+      "loss": 3.0204,
+      "step": 5691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5302362885438924e-06,
+      "loss": 2.8689,
+      "step": 5692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5284886129284898e-06,
+      "loss": 3.0316,
+      "step": 5693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5267413368119466e-06,
+      "loss": 2.9003,
+      "step": 5694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.524994460476689e-06,
+      "loss": 3.0848,
+      "step": 5695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5232479842050928e-06,
+      "loss": 2.9635,
+      "step": 5696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5215019082794557e-06,
+      "loss": 2.9765,
+      "step": 5697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.519756232982019e-06,
+      "loss": 3.0096,
+      "step": 5698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.518010958594953e-06,
+      "loss": 2.8862,
+      "step": 5699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5162660854003684e-06,
+      "loss": 2.9797,
+      "step": 5700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5145216136803108e-06,
+      "loss": 3.0005,
+      "step": 5701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5127775437167557e-06,
+      "loss": 3.0779,
+      "step": 5702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.511033875791621e-06,
+      "loss": 3.0288,
+      "step": 5703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.509290610186752e-06,
+      "loss": 3.0892,
+      "step": 5704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5075477471839347e-06,
+      "loss": 2.9795,
+      "step": 5705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5058052870648887e-06,
+      "loss": 3.0069,
+      "step": 5706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5040632301112668e-06,
+      "loss": 3.0099,
+      "step": 5707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5023215766046605e-06,
+      "loss": 2.8647,
+      "step": 5708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.5005803268265888e-06,
+      "loss": 2.8569,
+      "step": 5709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4988394810585144e-06,
+      "loss": 3.1361,
+      "step": 5710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4970990395818263e-06,
+      "loss": 2.8459,
+      "step": 5711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4953590026778534e-06,
+      "loss": 3.0123,
+      "step": 5712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.493619370627861e-06,
+      "loss": 2.9407,
+      "step": 5713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4918801437130407e-06,
+      "loss": 3.0135,
+      "step": 5714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.490141322214527e-06,
+      "loss": 3.047,
+      "step": 5715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4884029064133846e-06,
+      "loss": 2.965,
+      "step": 5716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4866648965906166e-06,
+      "loss": 2.8744,
+      "step": 5717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4849272930271524e-06,
+      "loss": 2.9194,
+      "step": 5718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.483190096003864e-06,
+      "loss": 2.9706,
+      "step": 5719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.481453305801556e-06,
+      "loss": 2.8209,
+      "step": 5720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.479716922700962e-06,
+      "loss": 2.9295,
+      "step": 5721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4779809469827572e-06,
+      "loss": 2.8657,
+      "step": 5722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4762453789275436e-06,
+      "loss": 3.0674,
+      "step": 5723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4745102188158636e-06,
+      "loss": 3.0195,
+      "step": 5724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4727754669281906e-06,
+      "loss": 2.9569,
+      "step": 5725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4710411235449322e-06,
+      "loss": 3.0378,
+      "step": 5726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4693071889464327e-06,
+      "loss": 2.87,
+      "step": 5727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4675736634129643e-06,
+      "loss": 2.8169,
+      "step": 5728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.46584054722474e-06,
+      "loss": 3.0124,
+      "step": 5729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4641078406618994e-06,
+      "loss": 2.9052,
+      "step": 5730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.462375544004523e-06,
+      "loss": 2.9192,
+      "step": 5731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.460643657532622e-06,
+      "loss": 3.0228,
+      "step": 5732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4589121815261387e-06,
+      "loss": 2.983,
+      "step": 5733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4571811162649524e-06,
+      "loss": 3.0879,
+      "step": 5734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4554504620288755e-06,
+      "loss": 3.0086,
+      "step": 5735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.453720219097656e-06,
+      "loss": 2.8647,
+      "step": 5736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4519903877509688e-06,
+      "loss": 3.0674,
+      "step": 5737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.450260968268428e-06,
+      "loss": 3.0555,
+      "step": 5738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.448531960929582e-06,
+      "loss": 2.9,
+      "step": 5739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4468033660139067e-06,
+      "loss": 2.9095,
+      "step": 5740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.445075183800818e-06,
+      "loss": 2.9501,
+      "step": 5741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.443347414569658e-06,
+      "loss": 2.9193,
+      "step": 5742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.441620058599708e-06,
+      "loss": 2.8212,
+      "step": 5743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.439893116170181e-06,
+      "loss": 2.9058,
+      "step": 5744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.438166587560222e-06,
+      "loss": 2.9789,
+      "step": 5745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.436440473048911e-06,
+      "loss": 2.7779,
+      "step": 5746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4347147729152576e-06,
+      "loss": 3.0047,
+      "step": 5747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4329894874382094e-06,
+      "loss": 2.8233,
+      "step": 5748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4312646168966397e-06,
+      "loss": 3.0657,
+      "step": 5749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4295401615693625e-06,
+      "loss": 2.9712,
+      "step": 5750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4278161217351227e-06,
+      "loss": 2.9511,
+      "step": 5751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.426092497672593e-06,
+      "loss": 2.9812,
+      "step": 5752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.424369289660386e-06,
+      "loss": 3.0048,
+      "step": 5753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.422646497977038e-06,
+      "loss": 2.9242,
+      "step": 5754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.420924122901032e-06,
+      "loss": 2.9396,
+      "step": 5755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.419202164710769e-06,
+      "loss": 3.0673,
+      "step": 5756
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.417480623684591e-06,
+      "loss": 2.8623,
+      "step": 5757
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.415759500100772e-06,
+      "loss": 3.101,
+      "step": 5758
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4140387942375144e-06,
+      "loss": 2.9912,
+      "step": 5759
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.412318506372958e-06,
+      "loss": 2.93,
+      "step": 5760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4105986367851703e-06,
+      "loss": 3.0608,
+      "step": 5761
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4088791857521544e-06,
+      "loss": 2.9613,
+      "step": 5762
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4071601535518472e-06,
+      "loss": 2.8942,
+      "step": 5763
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4054415404621105e-06,
+      "loss": 2.99,
+      "step": 5764
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4037233467607514e-06,
+      "loss": 2.9076,
+      "step": 5765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.4020055727254942e-06,
+      "loss": 2.8802,
+      "step": 5766
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.400288218634007e-06,
+      "loss": 3.0915,
+      "step": 5767
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3985712847638825e-06,
+      "loss": 2.7734,
+      "step": 5768
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.396854771392649e-06,
+      "loss": 2.87,
+      "step": 5769
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.39513867879777e-06,
+      "loss": 2.9619,
+      "step": 5770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3934230072566323e-06,
+      "loss": 2.9592,
+      "step": 5771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.391707757046563e-06,
+      "loss": 2.8149,
+      "step": 5772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.389992928444813e-06,
+      "loss": 2.9412,
+      "step": 5773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3882785217285775e-06,
+      "loss": 2.9204,
+      "step": 5774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3865645371749696e-06,
+      "loss": 2.793,
+      "step": 5775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3848509750610422e-06,
+      "loss": 3.0316,
+      "step": 5776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.38313783566378e-06,
+      "loss": 2.7733,
+      "step": 5777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3814251192600945e-06,
+      "loss": 3.0681,
+      "step": 5778
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3797128261268347e-06,
+      "loss": 3.0174,
+      "step": 5779
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3780009565407746e-06,
+      "loss": 3.1455,
+      "step": 5780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3762895107786256e-06,
+      "loss": 2.9502,
+      "step": 5781
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3745784891170298e-06,
+      "loss": 3.0875,
+      "step": 5782
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3728678918325553e-06,
+      "loss": 3.0043,
+      "step": 5783
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.371157719201709e-06,
+      "loss": 2.9488,
+      "step": 5784
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3694479715009245e-06,
+      "loss": 3.1445,
+      "step": 5785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.36773864900657e-06,
+      "loss": 2.8237,
+      "step": 5786
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3660297519949405e-06,
+      "loss": 3.0046,
+      "step": 5787
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3643212807422665e-06,
+      "loss": 2.9336,
+      "step": 5788
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3626132355247055e-06,
+      "loss": 2.9306,
+      "step": 5789
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.36090561661835e-06,
+      "loss": 3.0556,
+      "step": 5790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3591984242992237e-06,
+      "loss": 2.911,
+      "step": 5791
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3574916588432767e-06,
+      "loss": 2.9955,
+      "step": 5792
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.355785320526394e-06,
+      "loss": 2.9396,
+      "step": 5793
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.354079409624392e-06,
+      "loss": 2.9174,
+      "step": 5794
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.352373926413018e-06,
+      "loss": 3.0188,
+      "step": 5795
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3506688711679447e-06,
+      "loss": 2.9896,
+      "step": 5796
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3489642441647824e-06,
+      "loss": 3.051,
+      "step": 5797
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.347260045679071e-06,
+      "loss": 2.9081,
+      "step": 5798
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3455562759862766e-06,
+      "loss": 2.9848,
+      "step": 5799
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3438529353618023e-06,
+      "loss": 3.0905,
+      "step": 5800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.342150024080975e-06,
+      "loss": 3.0089,
+      "step": 5801
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3404475424190576e-06,
+      "loss": 2.8853,
+      "step": 5802
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.338745490651242e-06,
+      "loss": 2.9577,
+      "step": 5803
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3370438690526507e-06,
+      "loss": 3.0335,
+      "step": 5804
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3353426778983378e-06,
+      "loss": 3.1061,
+      "step": 5805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3336419174632833e-06,
+      "loss": 2.8929,
+      "step": 5806
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3319415880224038e-06,
+      "loss": 3.0417,
+      "step": 5807
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3302416898505397e-06,
+      "loss": 2.8735,
+      "step": 5808
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.328542223222467e-06,
+      "loss": 3.0877,
+      "step": 5809
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3268431884128924e-06,
+      "loss": 2.9644,
+      "step": 5810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3251445856964456e-06,
+      "loss": 2.8907,
+      "step": 5811
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3234464153476956e-06,
+      "loss": 2.9789,
+      "step": 5812
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.321748677641132e-06,
+      "loss": 2.9514,
+      "step": 5813
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.320051372851187e-06,
+      "loss": 3.0055,
+      "step": 5814
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3183545012522097e-06,
+      "loss": 3.0919,
+      "step": 5815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3166580631184866e-06,
+      "loss": 2.7647,
+      "step": 5816
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.314962058724235e-06,
+      "loss": 3.0782,
+      "step": 5817
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.313266488343596e-06,
+      "loss": 2.8698,
+      "step": 5818
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.311571352250647e-06,
+      "loss": 2.9026,
+      "step": 5819
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3098766507193896e-06,
+      "loss": 3.0036,
+      "step": 5820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.30818238402376e-06,
+      "loss": 2.9382,
+      "step": 5821
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3064885524376207e-06,
+      "loss": 2.9349,
+      "step": 5822
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3047951562347663e-06,
+      "loss": 2.8715,
+      "step": 5823
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3031021956889216e-06,
+      "loss": 2.9764,
+      "step": 5824
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3014096710737356e-06,
+      "loss": 2.8636,
+      "step": 5825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2997175826627938e-06,
+      "loss": 3.0021,
+      "step": 5826
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.298025930729605e-06,
+      "loss": 2.8765,
+      "step": 5827
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2963347155476112e-06,
+      "loss": 2.7598,
+      "step": 5828
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2946439373901858e-06,
+      "loss": 2.9051,
+      "step": 5829
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2929535965306248e-06,
+      "loss": 2.9399,
+      "step": 5830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.291263693242161e-06,
+      "loss": 2.9079,
+      "step": 5831
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2895742277979472e-06,
+      "loss": 2.9547,
+      "step": 5832
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2878852004710793e-06,
+      "loss": 2.9034,
+      "step": 5833
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.286196611534568e-06,
+      "loss": 2.8108,
+      "step": 5834
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2845084612613617e-06,
+      "loss": 3.0214,
+      "step": 5835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.282820749924337e-06,
+      "loss": 2.9755,
+      "step": 5836
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2811334777962947e-06,
+      "loss": 2.9112,
+      "step": 5837
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2794466451499713e-06,
+      "loss": 2.8409,
+      "step": 5838
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.277760252258026e-06,
+      "loss": 2.7699,
+      "step": 5839
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2760742993930514e-06,
+      "loss": 2.9948,
+      "step": 5840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2743887868275694e-06,
+      "loss": 3.019,
+      "step": 5841
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.272703714834023e-06,
+      "loss": 2.9426,
+      "step": 5842
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.271019083684797e-06,
+      "loss": 2.9861,
+      "step": 5843
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2693348936521936e-06,
+      "loss": 2.8361,
+      "step": 5844
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.26765114500845e-06,
+      "loss": 3.0835,
+      "step": 5845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2659678380257265e-06,
+      "loss": 2.9113,
+      "step": 5846
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2642849729761176e-06,
+      "loss": 3.0095,
+      "step": 5847
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2626025501316457e-06,
+      "loss": 2.913,
+      "step": 5848
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2609205697642573e-06,
+      "loss": 2.8709,
+      "step": 5849
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.259239032145833e-06,
+      "loss": 3.0259,
+      "step": 5850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2575579375481756e-06,
+      "loss": 3.0549,
+      "step": 5851
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2558772862430216e-06,
+      "loss": 3.0224,
+      "step": 5852
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2541970785020345e-06,
+      "loss": 2.9186,
+      "step": 5853
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.252517314596805e-06,
+      "loss": 2.8659,
+      "step": 5854
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.250837994798855e-06,
+      "loss": 2.9802,
+      "step": 5855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2491591193796288e-06,
+      "loss": 3.0407,
+      "step": 5856
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2474806886105057e-06,
+      "loss": 2.9506,
+      "step": 5857
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2458027027627854e-06,
+      "loss": 3.0006,
+      "step": 5858
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.244125162107704e-06,
+      "loss": 3.0526,
+      "step": 5859
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.242448066916421e-06,
+      "loss": 2.8864,
+      "step": 5860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2407714174600225e-06,
+      "loss": 3.0139,
+      "step": 5861
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2390952140095262e-06,
+      "loss": 3.0122,
+      "step": 5862
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2374194568358765e-06,
+      "loss": 2.8097,
+      "step": 5863
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2357441462099467e-06,
+      "loss": 3.026,
+      "step": 5864
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.234069282402533e-06,
+      "loss": 3.0116,
+      "step": 5865
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.232394865684365e-06,
+      "loss": 2.9677,
+      "step": 5866
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.230720896326099e-06,
+      "loss": 2.9396,
+      "step": 5867
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2290473745983147e-06,
+      "loss": 2.9028,
+      "step": 5868
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2273743007715276e-06,
+      "loss": 3.0644,
+      "step": 5869
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.22570167511617e-06,
+      "loss": 2.9355,
+      "step": 5870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.224029497902612e-06,
+      "loss": 2.8753,
+      "step": 5871
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.222357769401144e-06,
+      "loss": 2.901,
+      "step": 5872
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.220686489881989e-06,
+      "loss": 2.9548,
+      "step": 5873
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.219015659615296e-06,
+      "loss": 2.9558,
+      "step": 5874
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2173452788711375e-06,
+      "loss": 2.869,
+      "step": 5875
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2156753479195198e-06,
+      "loss": 2.8652,
+      "step": 5876
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2140058670303688e-06,
+      "loss": 3.0077,
+      "step": 5877
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.212336836473544e-06,
+      "loss": 2.9229,
+      "step": 5878
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2106682565188326e-06,
+      "loss": 2.8246,
+      "step": 5879
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2090001274359425e-06,
+      "loss": 3.0353,
+      "step": 5880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.207332449494514e-06,
+      "loss": 2.9648,
+      "step": 5881
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.205665222964113e-06,
+      "loss": 3.0778,
+      "step": 5882
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2039984481142345e-06,
+      "loss": 3.0047,
+      "step": 5883
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2023321252142956e-06,
+      "loss": 3.1406,
+      "step": 5884
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.2006662545336436e-06,
+      "loss": 3.0103,
+      "step": 5885
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.199000836341556e-06,
+      "loss": 2.7596,
+      "step": 5886
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1973358709072286e-06,
+      "loss": 3.0076,
+      "step": 5887
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.195671358499793e-06,
+      "loss": 3.0034,
+      "step": 5888
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1940072993883e-06,
+      "loss": 2.8603,
+      "step": 5889
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.192343693841732e-06,
+      "loss": 3.0259,
+      "step": 5890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.190680542128997e-06,
+      "loss": 3.0561,
+      "step": 5891
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.18901784451893e-06,
+      "loss": 2.8892,
+      "step": 5892
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1873556012802927e-06,
+      "loss": 2.9764,
+      "step": 5893
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.18569381268177e-06,
+      "loss": 2.8625,
+      "step": 5894
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1840324789919785e-06,
+      "loss": 2.8807,
+      "step": 5895
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1823716004794558e-06,
+      "loss": 2.935,
+      "step": 5896
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1807111774126706e-06,
+      "loss": 2.7878,
+      "step": 5897
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.179051210060018e-06,
+      "loss": 3.0435,
+      "step": 5898
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1773916986898137e-06,
+      "loss": 2.9941,
+      "step": 5899
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1757326435703074e-06,
+      "loss": 3.0632,
+      "step": 5900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1740740449696655e-06,
+      "loss": 3.0275,
+      "step": 5901
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.172415903155995e-06,
+      "loss": 3.0194,
+      "step": 5902
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1707582183973134e-06,
+      "loss": 3.0439,
+      "step": 5903
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1691009909615744e-06,
+      "loss": 3.0102,
+      "step": 5904
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.167444221116656e-06,
+      "loss": 3.1148,
+      "step": 5905
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1657879091303573e-06,
+      "loss": 3.0365,
+      "step": 5906
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.164132055270411e-06,
+      "loss": 2.8374,
+      "step": 5907
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1624766598044676e-06,
+      "loss": 2.9159,
+      "step": 5908
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1608217230001104e-06,
+      "loss": 2.8547,
+      "step": 5909
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1591672451248472e-06,
+      "loss": 2.912,
+      "step": 5910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.157513226446105e-06,
+      "loss": 2.9645,
+      "step": 5911
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.15585966723125e-06,
+      "loss": 2.9513,
+      "step": 5912
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.154206567747559e-06,
+      "loss": 2.8416,
+      "step": 5913
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1525539282622453e-06,
+      "loss": 2.9973,
+      "step": 5914
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1509017490424418e-06,
+      "loss": 2.8539,
+      "step": 5915
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.14925003035521e-06,
+      "loss": 3.0501,
+      "step": 5916
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1475987724675387e-06,
+      "loss": 3.005,
+      "step": 5917
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1459479756463354e-06,
+      "loss": 2.9556,
+      "step": 5918
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1442976401584416e-06,
+      "loss": 2.9088,
+      "step": 5919
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.142647766270614e-06,
+      "loss": 2.9529,
+      "step": 5920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1409983542495487e-06,
+      "loss": 2.8447,
+      "step": 5921
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.139349404361853e-06,
+      "loss": 2.8244,
+      "step": 5922
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.137700916874068e-06,
+      "loss": 2.8457,
+      "step": 5923
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1360528920526596e-06,
+      "loss": 2.9415,
+      "step": 5924
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1344053301640126e-06,
+      "loss": 2.9509,
+      "step": 5925
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.132758231474447e-06,
+      "loss": 2.879,
+      "step": 5926
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.131111596250196e-06,
+      "loss": 3.0792,
+      "step": 5927
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1294654247574275e-06,
+      "loss": 2.9114,
+      "step": 5928
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1278197172622334e-06,
+      "loss": 2.8753,
+      "step": 5929
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1261744740306232e-06,
+      "loss": 2.979,
+      "step": 5930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1245296953285403e-06,
+      "loss": 2.9253,
+      "step": 5931
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.122885381421848e-06,
+      "loss": 2.8388,
+      "step": 5932
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1212415325763373e-06,
+      "loss": 2.9314,
+      "step": 5933
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.11959814905772e-06,
+      "loss": 3.0198,
+      "step": 5934
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1179552311316376e-06,
+      "loss": 2.7842,
+      "step": 5935
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1163127790636506e-06,
+      "loss": 2.9471,
+      "step": 5936
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1146707931192496e-06,
+      "loss": 3.0923,
+      "step": 5937
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1130292735638484e-06,
+      "loss": 2.8982,
+      "step": 5938
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.111388220662783e-06,
+      "loss": 3.0669,
+      "step": 5939
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1097476346813157e-06,
+      "loss": 3.0148,
+      "step": 5940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1081075158846336e-06,
+      "loss": 2.9018,
+      "step": 5941
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1064678645378506e-06,
+      "loss": 3.07,
+      "step": 5942
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1048286809059987e-06,
+      "loss": 2.7328,
+      "step": 5943
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1031899652540395e-06,
+      "loss": 3.0156,
+      "step": 5944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.1015517178468585e-06,
+      "loss": 2.8513,
+      "step": 5945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.099913938949263e-06,
+      "loss": 3.0193,
+      "step": 5946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.098276628825988e-06,
+      "loss": 2.9194,
+      "step": 5947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.096639787741687e-06,
+      "loss": 2.9711,
+      "step": 5948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0950034159609445e-06,
+      "loss": 2.9922,
+      "step": 5949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.093367513748265e-06,
+      "loss": 3.0808,
+      "step": 5950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.091732081368079e-06,
+      "loss": 2.9684,
+      "step": 5951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0900971190847426e-06,
+      "loss": 3.1082,
+      "step": 5952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.088462627162528e-06,
+      "loss": 3.0422,
+      "step": 5953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0868286058656422e-06,
+      "loss": 2.9996,
+      "step": 5954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0851950554582075e-06,
+      "loss": 2.8251,
+      "step": 5955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0835619762042734e-06,
+      "loss": 2.9002,
+      "step": 5956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.081929368367817e-06,
+      "loss": 3.0007,
+      "step": 5957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0802972322127314e-06,
+      "loss": 2.8752,
+      "step": 5958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0786655680028407e-06,
+      "loss": 3.0238,
+      "step": 5959
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0770343760018846e-06,
+      "loss": 2.9822,
+      "step": 5960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0754036564735386e-06,
+      "loss": 3.0081,
+      "step": 5961
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0737734096813892e-06,
+      "loss": 2.9326,
+      "step": 5962
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0721436358889535e-06,
+      "loss": 2.9425,
+      "step": 5963
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0705143353596734e-06,
+      "loss": 3.0492,
+      "step": 5964
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0688855083569064e-06,
+      "loss": 2.9559,
+      "step": 5965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0672571551439437e-06,
+      "loss": 2.9549,
+      "step": 5966
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0656292759839903e-06,
+      "loss": 2.9523,
+      "step": 5967
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.064001871140181e-06,
+      "loss": 2.9657,
+      "step": 5968
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0623749408755717e-06,
+      "loss": 2.9466,
+      "step": 5969
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0607484854531428e-06,
+      "loss": 2.9463,
+      "step": 5970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.059122505135798e-06,
+      "loss": 2.9953,
+      "step": 5971
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.05749700018636e-06,
+      "loss": 2.8061,
+      "step": 5972
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.055871970867581e-06,
+      "loss": 2.9472,
+      "step": 5973
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.05424741744213e-06,
+      "loss": 3.0079,
+      "step": 5974
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0526233401726046e-06,
+      "loss": 3.0684,
+      "step": 5975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0509997393215243e-06,
+      "loss": 3.0336,
+      "step": 5976
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.049376615151327e-06,
+      "loss": 2.8386,
+      "step": 5977
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0477539679243794e-06,
+      "loss": 3.0083,
+      "step": 5978
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0461317979029653e-06,
+      "loss": 2.9214,
+      "step": 5979
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0445101053493005e-06,
+      "loss": 2.996,
+      "step": 5980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0428888905255133e-06,
+      "loss": 2.9286,
+      "step": 5981
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0412681536936605e-06,
+      "loss": 2.9242,
+      "step": 5982
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0396478951157226e-06,
+      "loss": 2.9598,
+      "step": 5983
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0380281150535967e-06,
+      "loss": 2.9008,
+      "step": 5984
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0364088137691108e-06,
+      "loss": 3.0467,
+      "step": 5985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0347899915240065e-06,
+      "loss": 3.0549,
+      "step": 5986
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0331716485799558e-06,
+      "loss": 2.9512,
+      "step": 5987
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.031553785198551e-06,
+      "loss": 2.8456,
+      "step": 5988
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.029936401641301e-06,
+      "loss": 3.035,
+      "step": 5989
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.028319498169649e-06,
+      "loss": 2.9795,
+      "step": 5990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.026703075044949e-06,
+      "loss": 2.9778,
+      "step": 5991
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.025087132528485e-06,
+      "loss": 2.8911,
+      "step": 5992
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0234716708814565e-06,
+      "loss": 2.9759,
+      "step": 5993
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0218566903649923e-06,
+      "loss": 2.8869,
+      "step": 5994
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.020242191240141e-06,
+      "loss": 3.0036,
+      "step": 5995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0186281737678693e-06,
+      "loss": 2.9662,
+      "step": 5996
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0170146382090734e-06,
+      "loss": 2.914,
+      "step": 5997
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0154015848245637e-06,
+      "loss": 2.9816,
+      "step": 5998
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0137890138750786e-06,
+      "loss": 2.9879,
+      "step": 5999
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0121769256212762e-06,
+      "loss": 3.0004,
+      "step": 6000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0105653203237375e-06,
+      "loss": 2.9383,
+      "step": 6001
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0089541982429657e-06,
+      "loss": 2.7688,
+      "step": 6002
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0073435596393824e-06,
+      "loss": 2.8501,
+      "step": 6003
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0057334047733366e-06,
+      "loss": 2.9702,
+      "step": 6004
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.004123733905093e-06,
+      "loss": 3.0,
+      "step": 6005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0025145472948427e-06,
+      "loss": 2.9906,
+      "step": 6006
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.0009058452026993e-06,
+      "loss": 3.0064,
+      "step": 6007
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.99929762788869e-06,
+      "loss": 3.0148,
+      "step": 6008
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9976898956127765e-06,
+      "loss": 2.9217,
+      "step": 6009
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9960826486348304e-06,
+      "loss": 2.9868,
+      "step": 6010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.994475887214653e-06,
+      "loss": 3.0014,
+      "step": 6011
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9928696116119593e-06,
+      "loss": 3.0297,
+      "step": 6012
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9912638220863933e-06,
+      "loss": 2.8228,
+      "step": 6013
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9896585188975177e-06,
+      "loss": 3.0284,
+      "step": 6014
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9880537023048133e-06,
+      "loss": 2.9435,
+      "step": 6015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.986449372567689e-06,
+      "loss": 2.7904,
+      "step": 6016
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.984845529945466e-06,
+      "loss": 2.97,
+      "step": 6017
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.983242174697395e-06,
+      "loss": 3.0512,
+      "step": 6018
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9816393070826443e-06,
+      "loss": 3.0298,
+      "step": 6019
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.980036927360305e-06,
+      "loss": 3.0211,
+      "step": 6020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.978435035789388e-06,
+      "loss": 3.004,
+      "step": 6021
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9768336326288227e-06,
+      "loss": 2.8318,
+      "step": 6022
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.975232718137467e-06,
+      "loss": 3.0069,
+      "step": 6023
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9736322925740896e-06,
+      "loss": 3.0572,
+      "step": 6024
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9720323561973896e-06,
+      "loss": 2.8685,
+      "step": 6025
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9704329092659837e-06,
+      "loss": 3.0451,
+      "step": 6026
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9688339520384053e-06,
+      "loss": 2.9471,
+      "step": 6027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.967235484773114e-06,
+      "loss": 2.8032,
+      "step": 6028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9656375077284895e-06,
+      "loss": 2.8708,
+      "step": 6029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.964040021162832e-06,
+      "loss": 3.0096,
+      "step": 6030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9624430253343584e-06,
+      "loss": 2.9128,
+      "step": 6031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9608465205012115e-06,
+      "loss": 3.0554,
+      "step": 6032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9592505069214544e-06,
+      "loss": 2.9004,
+      "step": 6033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9576549848530656e-06,
+      "loss": 2.935,
+      "step": 6034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9560599545539515e-06,
+      "loss": 2.9736,
+      "step": 6035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.954465416281931e-06,
+      "loss": 3.0372,
+      "step": 6036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9528713702947506e-06,
+      "loss": 3.0679,
+      "step": 6037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.951277816850073e-06,
+      "loss": 2.9764,
+      "step": 6038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9496847562054833e-06,
+      "loss": 3.0008,
+      "step": 6039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.948092188618488e-06,
+      "loss": 2.8716,
+      "step": 6040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9465001143465086e-06,
+      "loss": 2.9677,
+      "step": 6041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9449085336468936e-06,
+      "loss": 2.9228,
+      "step": 6042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9433174467769055e-06,
+      "loss": 2.9684,
+      "step": 6043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9417268539937317e-06,
+      "loss": 2.8354,
+      "step": 6044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.940136755554479e-06,
+      "loss": 2.9872,
+      "step": 6045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9385471517161712e-06,
+      "loss": 2.8625,
+      "step": 6046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.936958042735757e-06,
+      "loss": 2.7807,
+      "step": 6047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.935369428870097e-06,
+      "loss": 2.9527,
+      "step": 6048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.933781310375984e-06,
+      "loss": 2.8716,
+      "step": 6049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9321936875101192e-06,
+      "loss": 2.9192,
+      "step": 6050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9306065605291303e-06,
+      "loss": 3.0558,
+      "step": 6051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.929019929689564e-06,
+      "loss": 2.9452,
+      "step": 6052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9274337952478823e-06,
+      "loss": 3.0557,
+      "step": 6053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.925848157460474e-06,
+      "loss": 3.0088,
+      "step": 6054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9242630165836405e-06,
+      "loss": 2.885,
+      "step": 6055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9226783728736087e-06,
+      "loss": 2.9066,
+      "step": 6056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.921094226586522e-06,
+      "loss": 3.0553,
+      "step": 6057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.919510577978445e-06,
+      "loss": 2.8553,
+      "step": 6058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9179274273053617e-06,
+      "loss": 3.0504,
+      "step": 6059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9163447748231722e-06,
+      "loss": 2.9375,
+      "step": 6060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9147626207877027e-06,
+      "loss": 2.9941,
+      "step": 6061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9131809654546907e-06,
+      "loss": 2.8586,
+      "step": 6062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9115998090798e-06,
+      "loss": 2.8406,
+      "step": 6063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.910019151918612e-06,
+      "loss": 2.9314,
+      "step": 6064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9084389942266237e-06,
+      "loss": 2.898,
+      "step": 6065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9068593362592569e-06,
+      "loss": 2.8665,
+      "step": 6066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.905280178271845e-06,
+      "loss": 2.8751,
+      "step": 6067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.903701520519653e-06,
+      "loss": 2.948,
+      "step": 6068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9021233632578518e-06,
+      "loss": 3.0444,
+      "step": 6069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.9005457067415412e-06,
+      "loss": 2.8165,
+      "step": 6070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8989685512257316e-06,
+      "loss": 2.9659,
+      "step": 6071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8973918969653593e-06,
+      "loss": 2.7517,
+      "step": 6072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.895815744215278e-06,
+      "loss": 2.9214,
+      "step": 6073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8942400932302573e-06,
+      "loss": 2.9293,
+      "step": 6074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8926649442649902e-06,
+      "loss": 2.9239,
+      "step": 6075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8910902975740826e-06,
+      "loss": 3.1023,
+      "step": 6076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8895161534120653e-06,
+      "loss": 2.9562,
+      "step": 6077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.887942512033385e-06,
+      "loss": 3.0197,
+      "step": 6078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8863693736924077e-06,
+      "loss": 3.0117,
+      "step": 6079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8847967386434191e-06,
+      "loss": 3.1495,
+      "step": 6080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8832246071406191e-06,
+      "loss": 3.1324,
+      "step": 6081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8816529794381328e-06,
+      "loss": 2.9336,
+      "step": 6082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8800818557899969e-06,
+      "loss": 2.8901,
+      "step": 6083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8785112364501728e-06,
+      "loss": 3.0071,
+      "step": 6084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8769411216725386e-06,
+      "loss": 3.0427,
+      "step": 6085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8753715117108872e-06,
+      "loss": 3.0249,
+      "step": 6086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8738024068189337e-06,
+      "loss": 2.9641,
+      "step": 6087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8722338072503115e-06,
+      "loss": 2.9465,
+      "step": 6088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8706657132585727e-06,
+      "loss": 2.8683,
+      "step": 6089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8690981250971823e-06,
+      "loss": 2.8943,
+      "step": 6090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8675310430195309e-06,
+      "loss": 2.9979,
+      "step": 6091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8659644672789245e-06,
+      "loss": 2.8393,
+      "step": 6092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8643983981285834e-06,
+      "loss": 3.0156,
+      "step": 6093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8628328358216535e-06,
+      "loss": 2.9502,
+      "step": 6094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8612677806111907e-06,
+      "loss": 2.9671,
+      "step": 6095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8597032327501752e-06,
+      "loss": 3.0095,
+      "step": 6096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.858139192491502e-06,
+      "loss": 2.825,
+      "step": 6097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8565756600879853e-06,
+      "loss": 2.876,
+      "step": 6098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8550126357923593e-06,
+      "loss": 3.0599,
+      "step": 6099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.853450119857269e-06,
+      "loss": 2.9739,
+      "step": 6100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8518881125352866e-06,
+      "loss": 2.7875,
+      "step": 6101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8503266140788927e-06,
+      "loss": 3.0379,
+      "step": 6102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8487656247404927e-06,
+      "loss": 2.9137,
+      "step": 6103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8472051447724088e-06,
+      "loss": 3.0386,
+      "step": 6104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8456451744268755e-06,
+      "loss": 2.8516,
+      "step": 6105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8440857139560514e-06,
+      "loss": 2.9053,
+      "step": 6106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8425267636120092e-06,
+      "loss": 2.9268,
+      "step": 6107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8409683236467425e-06,
+      "loss": 3.0288,
+      "step": 6108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8394103943121555e-06,
+      "loss": 2.9878,
+      "step": 6109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8378529758600773e-06,
+      "loss": 2.9668,
+      "step": 6110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8362960685422515e-06,
+      "loss": 3.051,
+      "step": 6111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8347396726103367e-06,
+      "loss": 3.0593,
+      "step": 6112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8331837883159142e-06,
+      "loss": 2.912,
+      "step": 6113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8316284159104764e-06,
+      "loss": 2.984,
+      "step": 6114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8300735556454374e-06,
+      "loss": 3.1892,
+      "step": 6115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.828519207772127e-06,
+      "loss": 3.0886,
+      "step": 6116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.826965372541793e-06,
+      "loss": 2.8844,
+      "step": 6117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8254120502056005e-06,
+      "loss": 3.0781,
+      "step": 6118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8238592410146283e-06,
+      "loss": 3.0037,
+      "step": 6119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8223069452198777e-06,
+      "loss": 2.8784,
+      "step": 6120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8207551630722609e-06,
+      "loss": 2.8973,
+      "step": 6121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.819203894822612e-06,
+      "loss": 2.8936,
+      "step": 6122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8176531407216825e-06,
+      "loss": 2.9284,
+      "step": 6123
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8161029010201347e-06,
+      "loss": 2.9584,
+      "step": 6124
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8145531759685553e-06,
+      "loss": 2.9423,
+      "step": 6125
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8130039658174392e-06,
+      "loss": 3.0233,
+      "step": 6126
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8114552708172094e-06,
+      "loss": 3.0103,
+      "step": 6127
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8099070912181949e-06,
+      "loss": 2.9489,
+      "step": 6128
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8083594272706474e-06,
+      "loss": 2.9708,
+      "step": 6129
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8068122792247345e-06,
+      "loss": 2.8643,
+      "step": 6130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8052656473305374e-06,
+      "loss": 2.9872,
+      "step": 6131
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8037195318380584e-06,
+      "loss": 2.9231,
+      "step": 6132
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8021739329972116e-06,
+      "loss": 3.0443,
+      "step": 6133
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.8006288510578307e-06,
+      "loss": 2.9842,
+      "step": 6134
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7990842862696671e-06,
+      "loss": 2.9519,
+      "step": 6135
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.797540238882382e-06,
+      "loss": 2.9589,
+      "step": 6136
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7959967091455633e-06,
+      "loss": 2.7524,
+      "step": 6137
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7944536973087052e-06,
+      "loss": 2.9919,
+      "step": 6138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.792911203621226e-06,
+      "loss": 2.9505,
+      "step": 6139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7913692283324525e-06,
+      "loss": 2.9011,
+      "step": 6140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7898277716916346e-06,
+      "loss": 3.0757,
+      "step": 6141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7882868339479365e-06,
+      "loss": 2.8265,
+      "step": 6142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.786746415350435e-06,
+      "loss": 3.0188,
+      "step": 6143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7852065161481291e-06,
+      "loss": 2.9767,
+      "step": 6144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.783667136589926e-06,
+      "loss": 3.0124,
+      "step": 6145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7821282769246567e-06,
+      "loss": 2.9054,
+      "step": 6146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7805899374010637e-06,
+      "loss": 3.0064,
+      "step": 6147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7790521182678066e-06,
+      "loss": 2.8506,
+      "step": 6148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7775148197734633e-06,
+      "loss": 2.9589,
+      "step": 6149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.77597804216652e-06,
+      "loss": 2.8754,
+      "step": 6150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7744417856953894e-06,
+      "loss": 2.9529,
+      "step": 6151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.772906050608389e-06,
+      "loss": 2.7223,
+      "step": 6152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7713708371537603e-06,
+      "loss": 2.8753,
+      "step": 6153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7698361455796586e-06,
+      "loss": 3.0779,
+      "step": 6154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.768301976134149e-06,
+      "loss": 2.8567,
+      "step": 6155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7667683290652238e-06,
+      "loss": 2.9773,
+      "step": 6156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7652352046207782e-06,
+      "loss": 2.8388,
+      "step": 6157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7637026030486332e-06,
+      "loss": 2.9569,
+      "step": 6158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.762170524596517e-06,
+      "loss": 3.0118,
+      "step": 6159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.760638969512079e-06,
+      "loss": 3.0925,
+      "step": 6160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7591079380428832e-06,
+      "loss": 2.9267,
+      "step": 6161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7575774304364047e-06,
+      "loss": 2.8662,
+      "step": 6162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7560474469400414e-06,
+      "loss": 2.9124,
+      "step": 6163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.754517987801097e-06,
+      "loss": 3.1684,
+      "step": 6164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7529890532667986e-06,
+      "loss": 2.843,
+      "step": 6165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7514606435842856e-06,
+      "loss": 2.8036,
+      "step": 6166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7499327590006115e-06,
+      "loss": 2.9227,
+      "step": 6167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7484053997627481e-06,
+      "loss": 2.9517,
+      "step": 6168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7468785661175774e-06,
+      "loss": 2.9802,
+      "step": 6169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.745352258311901e-06,
+      "loss": 3.0062,
+      "step": 6170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7438264765924318e-06,
+      "loss": 2.9483,
+      "step": 6171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7423012212058e-06,
+      "loss": 3.0312,
+      "step": 6172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.740776492398552e-06,
+      "loss": 2.8883,
+      "step": 6173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7392522904171444e-06,
+      "loss": 2.8782,
+      "step": 6174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7377286155079532e-06,
+      "loss": 2.8341,
+      "step": 6175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7362054679172668e-06,
+      "loss": 3.0801,
+      "step": 6176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.734682847891292e-06,
+      "loss": 2.904,
+      "step": 6177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7331607556761426e-06,
+      "loss": 2.7855,
+      "step": 6178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7316391915178548e-06,
+      "loss": 2.7955,
+      "step": 6179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7301181556623781e-06,
+      "loss": 2.8161,
+      "step": 6180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7285976483555705e-06,
+      "loss": 2.951,
+      "step": 6181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.727077669843214e-06,
+      "loss": 2.8227,
+      "step": 6182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7255582203709959e-06,
+      "loss": 2.9564,
+      "step": 6183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7240393001845246e-06,
+      "loss": 3.0105,
+      "step": 6184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.72252090952932e-06,
+      "loss": 2.9476,
+      "step": 6185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7210030486508173e-06,
+      "loss": 3.0352,
+      "step": 6186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.719485717794368e-06,
+      "loss": 2.8483,
+      "step": 6187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7179689172052322e-06,
+      "loss": 2.916,
+      "step": 6188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7164526471285909e-06,
+      "loss": 2.8034,
+      "step": 6189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714936907809533e-06,
+      "loss": 2.8606,
+      "step": 6190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713421699493067e-06,
+      "loss": 2.9654,
+      "step": 6191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7119070224241146e-06,
+      "loss": 2.8485,
+      "step": 6192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7103928768475075e-06,
+      "loss": 2.819,
+      "step": 6193
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7088792630079976e-06,
+      "loss": 3.0728,
+      "step": 6194
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7073661811502429e-06,
+      "loss": 3.0257,
+      "step": 6195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7058536315188268e-06,
+      "loss": 3.0239,
+      "step": 6196
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704341614358236e-06,
+      "loss": 3.1261,
+      "step": 6197
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7028301299128753e-06,
+      "loss": 3.1308,
+      "step": 6198
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.701319178427066e-06,
+      "loss": 3.002,
+      "step": 6199
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6998087601450375e-06,
+      "loss": 3.0201,
+      "step": 6200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6982988753109385e-06,
+      "loss": 2.9013,
+      "step": 6201
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6967895241688265e-06,
+      "loss": 2.9724,
+      "step": 6202
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.695280706962677e-06,
+      "loss": 2.9687,
+      "step": 6203
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6937724239363774e-06,
+      "loss": 2.768,
+      "step": 6204
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6922646753337286e-06,
+      "loss": 2.979,
+      "step": 6205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6907574613984468e-06,
+      "loss": 2.9421,
+      "step": 6206
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6892507823741578e-06,
+      "loss": 2.8411,
+      "step": 6207
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6877446385044065e-06,
+      "loss": 3.0632,
+      "step": 6208
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6862390300326448e-06,
+      "loss": 2.9615,
+      "step": 6209
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.684733957202243e-06,
+      "loss": 2.8857,
+      "step": 6210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6832294202564858e-06,
+      "loss": 2.9391,
+      "step": 6211
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6817254194385646e-06,
+      "loss": 3.0423,
+      "step": 6212
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.680221954991592e-06,
+      "loss": 2.9569,
+      "step": 6213
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6787190271585857e-06,
+      "loss": 2.9099,
+      "step": 6214
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6772166361824877e-06,
+      "loss": 2.9505,
+      "step": 6215
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.675714782306142e-06,
+      "loss": 3.0576,
+      "step": 6216
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.674213465772313e-06,
+      "loss": 2.8196,
+      "step": 6217
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6727126868236732e-06,
+      "loss": 3.0053,
+      "step": 6218
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6712124457028122e-06,
+      "loss": 2.793,
+      "step": 6219
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.669712742652233e-06,
+      "loss": 2.9246,
+      "step": 6220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6682135779143466e-06,
+      "loss": 3.0699,
+      "step": 6221
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6667149517314835e-06,
+      "loss": 3.062,
+      "step": 6222
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6652168643458805e-06,
+      "loss": 3.0278,
+      "step": 6223
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6637193159996928e-06,
+      "loss": 2.9509,
+      "step": 6224
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6622223069349858e-06,
+      "loss": 2.887,
+      "step": 6225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.660725837393739e-06,
+      "loss": 3.0658,
+      "step": 6226
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6592299076178452e-06,
+      "loss": 2.9916,
+      "step": 6227
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.657734517849105e-06,
+      "loss": 2.8913,
+      "step": 6228
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6562396683292398e-06,
+      "loss": 2.9005,
+      "step": 6229
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6547453592998758e-06,
+      "loss": 2.9182,
+      "step": 6230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6532515910025565e-06,
+      "loss": 2.9133,
+      "step": 6231
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6517583636787388e-06,
+      "loss": 3.0386,
+      "step": 6232
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6502656775697868e-06,
+      "loss": 2.9386,
+      "step": 6233
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6487735329169823e-06,
+      "loss": 3.1014,
+      "step": 6234
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6472819299615184e-06,
+      "loss": 2.8877,
+      "step": 6235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6457908689445007e-06,
+      "loss": 2.9781,
+      "step": 6236
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6443003501069443e-06,
+      "loss": 2.846,
+      "step": 6237
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.64281037368978e-06,
+      "loss": 2.9974,
+      "step": 6238
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.641320939933851e-06,
+      "loss": 3.0454,
+      "step": 6239
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6398320490799091e-06,
+      "loss": 3.0742,
+      "step": 6240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6383437013686244e-06,
+      "loss": 2.894,
+      "step": 6241
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6368558970405718e-06,
+      "loss": 2.8628,
+      "step": 6242
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6353686363362437e-06,
+      "loss": 2.9146,
+      "step": 6243
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6338819194960438e-06,
+      "loss": 3.0911,
+      "step": 6244
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6323957467602869e-06,
+      "loss": 2.8676,
+      "step": 6245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6309101183692022e-06,
+      "loss": 2.9865,
+      "step": 6246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6294250345629254e-06,
+      "loss": 2.9316,
+      "step": 6247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.627940495581511e-06,
+      "loss": 3.0682,
+      "step": 6248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6264565016649186e-06,
+      "loss": 3.0255,
+      "step": 6249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6249730530530256e-06,
+      "loss": 2.9596,
+      "step": 6250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6234901499856198e-06,
+      "loss": 2.7841,
+      "step": 6251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.622007792702397e-06,
+      "loss": 2.8986,
+      "step": 6252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6205259814429691e-06,
+      "loss": 3.1165,
+      "step": 6253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6190447164468587e-06,
+      "loss": 2.9346,
+      "step": 6254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6175639979535013e-06,
+      "loss": 2.9828,
+      "step": 6255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6160838262022383e-06,
+      "loss": 2.9398,
+      "step": 6256
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.61460420143233e-06,
+      "loss": 2.9968,
+      "step": 6257
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6131251238829455e-06,
+      "loss": 2.9664,
+      "step": 6258
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.611646593793163e-06,
+      "loss": 2.8913,
+      "step": 6259
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6101686114019767e-06,
+      "loss": 2.853,
+      "step": 6260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6086911769482872e-06,
+      "loss": 2.926,
+      "step": 6261
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6072142906709109e-06,
+      "loss": 3.0203,
+      "step": 6262
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6057379528085738e-06,
+      "loss": 2.9488,
+      "step": 6263
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6042621635999129e-06,
+      "loss": 2.9162,
+      "step": 6264
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6027869232834797e-06,
+      "loss": 2.9682,
+      "step": 6265
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6013122320977297e-06,
+      "loss": 2.921,
+      "step": 6266
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.599838090281039e-06,
+      "loss": 2.8096,
+      "step": 6267
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.598364498071686e-06,
+      "loss": 2.8087,
+      "step": 6268
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5968914557078658e-06,
+      "loss": 2.8497,
+      "step": 6269
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.595418963427685e-06,
+      "loss": 2.8098,
+      "step": 6270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.593947021469157e-06,
+      "loss": 2.8205,
+      "step": 6271
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.592475630070211e-06,
+      "loss": 2.9149,
+      "step": 6272
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5910047894686803e-06,
+      "loss": 2.8363,
+      "step": 6273
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5895344999023204e-06,
+      "loss": 2.8973,
+      "step": 6274
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5880647616087868e-06,
+      "loss": 2.9826,
+      "step": 6275
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.586595574825651e-06,
+      "loss": 2.727,
+      "step": 6276
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5851269397903968e-06,
+      "loss": 2.8364,
+      "step": 6277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5836588567404127e-06,
+      "loss": 2.9047,
+      "step": 6278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5821913259130056e-06,
+      "loss": 3.0401,
+      "step": 6279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5807243475453866e-06,
+      "loss": 2.9547,
+      "step": 6280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5792579218746812e-06,
+      "loss": 2.863,
+      "step": 6281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5777920491379266e-06,
+      "loss": 2.997,
+      "step": 6282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5763267295720635e-06,
+      "loss": 2.8762,
+      "step": 6283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5748619634139555e-06,
+      "loss": 2.9956,
+      "step": 6284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5733977509003644e-06,
+      "loss": 2.9002,
+      "step": 6285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.571934092267971e-06,
+      "loss": 3.1841,
+      "step": 6286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5704709877533597e-06,
+      "loss": 2.9679,
+      "step": 6287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.569008437593031e-06,
+      "loss": 2.9026,
+      "step": 6288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5675464420233954e-06,
+      "loss": 2.9659,
+      "step": 6289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5660850012807688e-06,
+      "loss": 2.8604,
+      "step": 6290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5646241156013836e-06,
+      "loss": 2.8991,
+      "step": 6291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5631637852213754e-06,
+      "loss": 3.0076,
+      "step": 6292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5617040103768e-06,
+      "loss": 2.7884,
+      "step": 6293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5602447913036133e-06,
+      "loss": 2.7629,
+      "step": 6294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5587861282376864e-06,
+      "loss": 2.9917,
+      "step": 6295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5573280214148028e-06,
+      "loss": 2.9072,
+      "step": 6296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.555870471070649e-06,
+      "loss": 2.8934,
+      "step": 6297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5544134774408287e-06,
+      "loss": 3.0019,
+      "step": 6298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5529570407608501e-06,
+      "loss": 2.9939,
+      "step": 6299
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5515011612661345e-06,
+      "loss": 2.9037,
+      "step": 6300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5500458391920153e-06,
+      "loss": 2.9895,
+      "step": 6301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.548591074773727e-06,
+      "loss": 2.8579,
+      "step": 6302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5471368682464272e-06,
+      "loss": 2.8645,
+      "step": 6303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5456832198451704e-06,
+      "loss": 2.8254,
+      "step": 6304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.54423012980493e-06,
+      "loss": 2.9665,
+      "step": 6305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5427775983605835e-06,
+      "loss": 2.8543,
+      "step": 6306
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5413256257469205e-06,
+      "loss": 3.0964,
+      "step": 6307
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.539874212198642e-06,
+      "loss": 2.8726,
+      "step": 6308
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5384233579503543e-06,
+      "loss": 2.9087,
+      "step": 6309
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.536973063236578e-06,
+      "loss": 2.9408,
+      "step": 6310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5355233282917381e-06,
+      "loss": 3.0077,
+      "step": 6311
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.534074153350174e-06,
+      "loss": 2.9118,
+      "step": 6312
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5326255386461314e-06,
+      "loss": 2.7492,
+      "step": 6313
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5311774844137679e-06,
+      "loss": 2.9173,
+      "step": 6314
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5297299908871494e-06,
+      "loss": 2.9276,
+      "step": 6315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5282830583002488e-06,
+      "loss": 2.9846,
+      "step": 6316
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5268366868869532e-06,
+      "loss": 2.9955,
+      "step": 6317
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5253908768810532e-06,
+      "loss": 2.9911,
+      "step": 6318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5239456285162535e-06,
+      "loss": 2.916,
+      "step": 6319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5225009420261677e-06,
+      "loss": 3.0415,
+      "step": 6320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5210568176443141e-06,
+      "loss": 2.9897,
+      "step": 6321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5196132556041244e-06,
+      "loss": 3.0814,
+      "step": 6322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5181702561389384e-06,
+      "loss": 3.0727,
+      "step": 6323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5167278194820062e-06,
+      "loss": 3.0212,
+      "step": 6324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.515285945866482e-06,
+      "loss": 2.9668,
+      "step": 6325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.513844635525435e-06,
+      "loss": 3.0328,
+      "step": 6326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5124038886918413e-06,
+      "loss": 2.9869,
+      "step": 6327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5109637055985831e-06,
+      "loss": 3.0128,
+      "step": 6328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5095240864784561e-06,
+      "loss": 2.9624,
+      "step": 6329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5080850315641604e-06,
+      "loss": 2.9748,
+      "step": 6330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5066465410883075e-06,
+      "loss": 2.856,
+      "step": 6331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5052086152834177e-06,
+      "loss": 2.859,
+      "step": 6332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.5037712543819193e-06,
+      "loss": 2.9757,
+      "step": 6333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.502334458616152e-06,
+      "loss": 2.7439,
+      "step": 6334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.500898228218357e-06,
+      "loss": 2.9763,
+      "step": 6335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4994625634206934e-06,
+      "loss": 2.9404,
+      "step": 6336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49802746445522e-06,
+      "loss": 2.8781,
+      "step": 6337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49659293155391e-06,
+      "loss": 2.8726,
+      "step": 6338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4951589649486459e-06,
+      "loss": 2.9959,
+      "step": 6339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4937255648712123e-06,
+      "loss": 3.0519,
+      "step": 6340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922927315533087e-06,
+      "loss": 2.9891,
+      "step": 6341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.490860465226539e-06,
+      "loss": 2.9098,
+      "step": 6342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4894287661224194e-06,
+      "loss": 2.9428,
+      "step": 6343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.487997634472369e-06,
+      "loss": 2.927,
+      "step": 6344
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.486567070507719e-06,
+      "loss": 2.8636,
+      "step": 6345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.48513707445971e-06,
+      "loss": 2.9759,
+      "step": 6346
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4837076465594852e-06,
+      "loss": 2.8984,
+      "step": 6347
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4822787870381027e-06,
+      "loss": 2.8726,
+      "step": 6348
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.480850496126523e-06,
+      "loss": 2.982,
+      "step": 6349
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4794227740556177e-06,
+      "loss": 3.0118,
+      "step": 6350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.477995621056167e-06,
+      "loss": 2.9498,
+      "step": 6351
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4765690373588576e-06,
+      "loss": 2.9137,
+      "step": 6352
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.475143023194286e-06,
+      "loss": 2.967,
+      "step": 6353
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4737175787929524e-06,
+      "loss": 2.986,
+      "step": 6354
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4722927043852713e-06,
+      "loss": 2.8905,
+      "step": 6355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4708684002015584e-06,
+      "loss": 2.8327,
+      "step": 6356
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4694446664720414e-06,
+      "loss": 3.1354,
+      "step": 6357
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4680215034268563e-06,
+      "loss": 3.0582,
+      "step": 6358
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4665989112960427e-06,
+      "loss": 2.7923,
+      "step": 6359
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4651768903095537e-06,
+      "loss": 2.7821,
+      "step": 6360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4637554406972422e-06,
+      "loss": 3.0201,
+      "step": 6361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4623345626888792e-06,
+      "loss": 2.9997,
+      "step": 6362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4609142565141332e-06,
+      "loss": 2.8843,
+      "step": 6363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.459494522402588e-06,
+      "loss": 2.8253,
+      "step": 6364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4580753605837284e-06,
+      "loss": 2.9359,
+      "step": 6365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4566567712869512e-06,
+      "loss": 2.9166,
+      "step": 6366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4552387547415609e-06,
+      "loss": 3.0295,
+      "step": 6367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4538213111767653e-06,
+      "loss": 2.9307,
+      "step": 6368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4524044408216843e-06,
+      "loss": 2.8768,
+      "step": 6369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4509881439053408e-06,
+      "loss": 2.9464,
+      "step": 6370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.449572420656668e-06,
+      "loss": 2.926,
+      "step": 6371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.448157271304506e-06,
+      "loss": 2.9005,
+      "step": 6372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4467426960776015e-06,
+      "loss": 2.8912,
+      "step": 6373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4453286952046103e-06,
+      "loss": 2.882,
+      "step": 6374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4439152689140906e-06,
+      "loss": 3.083,
+      "step": 6375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4425024174345132e-06,
+      "loss": 2.9061,
+      "step": 6376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4410901409942512e-06,
+      "loss": 3.0845,
+      "step": 6377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4396784398215891e-06,
+      "loss": 3.0422,
+      "step": 6378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4382673141447167e-06,
+      "loss": 3.0139,
+      "step": 6379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.436856764191728e-06,
+      "loss": 3.0129,
+      "step": 6380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4354467901906283e-06,
+      "loss": 2.9965,
+      "step": 6381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4340373923693274e-06,
+      "loss": 2.8937,
+      "step": 6382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4326285709556443e-06,
+      "loss": 2.8781,
+      "step": 6383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4312203261773e-06,
+      "loss": 2.8792,
+      "step": 6384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4298126582619265e-06,
+      "loss": 2.9013,
+      "step": 6385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4284055674370638e-06,
+      "loss": 2.9612,
+      "step": 6386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4269990539301526e-06,
+      "loss": 3.0511,
+      "step": 6387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4255931179685467e-06,
+      "loss": 3.0098,
+      "step": 6388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4241877597795007e-06,
+      "loss": 2.9292,
+      "step": 6389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4227829795901803e-06,
+      "loss": 2.9268,
+      "step": 6390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4213787776276572e-06,
+      "loss": 2.9208,
+      "step": 6391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4199751541189078e-06,
+      "loss": 2.9301,
+      "step": 6392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4185721092908178e-06,
+      "loss": 3.0915,
+      "step": 6393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4171696433701742e-06,
+      "loss": 2.988,
+      "step": 6394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4157677565836768e-06,
+      "loss": 2.7925,
+      "step": 6395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4143664491579262e-06,
+      "loss": 2.953,
+      "step": 6396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4129657213194332e-06,
+      "loss": 2.8853,
+      "step": 6397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.411565573294615e-06,
+      "loss": 2.9305,
+      "step": 6398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.41016600530979e-06,
+      "loss": 2.8699,
+      "step": 6399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4087670175911888e-06,
+      "loss": 2.9457,
+      "step": 6400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4073686103649453e-06,
+      "loss": 2.9176,
+      "step": 6401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4059707838571024e-06,
+      "loss": 2.9617,
+      "step": 6402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4045735382936032e-06,
+      "loss": 2.9953,
+      "step": 6403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4031768739003027e-06,
+      "loss": 2.9743,
+      "step": 6404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4017807909029613e-06,
+      "loss": 3.0702,
+      "step": 6405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4003852895272402e-06,
+      "loss": 2.908,
+      "step": 6406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3989903699987145e-06,
+      "loss": 2.8497,
+      "step": 6407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3975960325428572e-06,
+      "loss": 3.0126,
+      "step": 6408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.396202277385053e-06,
+      "loss": 2.8146,
+      "step": 6409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3948091047505902e-06,
+      "loss": 2.9931,
+      "step": 6410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.393416514864664e-06,
+      "loss": 2.9284,
+      "step": 6411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.392024507952376e-06,
+      "loss": 2.8224,
+      "step": 6412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3906330842387295e-06,
+      "loss": 3.016,
+      "step": 6413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3892422439486386e-06,
+      "loss": 2.9618,
+      "step": 6414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3878519873069185e-06,
+      "loss": 2.908,
+      "step": 6415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3864623145382937e-06,
+      "loss": 2.8969,
+      "step": 6416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3850732258673948e-06,
+      "loss": 2.9423,
+      "step": 6417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3836847215187526e-06,
+      "loss": 3.0778,
+      "step": 6418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.382296801716811e-06,
+      "loss": 2.9696,
+      "step": 6419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3809094666859097e-06,
+      "loss": 2.8328,
+      "step": 6420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3795227166503067e-06,
+      "loss": 2.8994,
+      "step": 6421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3781365518341533e-06,
+      "loss": 2.9408,
+      "step": 6422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.376750972461513e-06,
+      "loss": 2.9873,
+      "step": 6423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3753659787563545e-06,
+      "loss": 3.1045,
+      "step": 6424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3739815709425475e-06,
+      "loss": 3.1568,
+      "step": 6425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.372597749243872e-06,
+      "loss": 3.0777,
+      "step": 6426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3712145138840088e-06,
+      "loss": 2.9529,
+      "step": 6427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3698318650865478e-06,
+      "loss": 2.8637,
+      "step": 6428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3684498030749832e-06,
+      "loss": 2.9087,
+      "step": 6429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3670683280727092e-06,
+      "loss": 3.068,
+      "step": 6430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3656874403030363e-06,
+      "loss": 3.1301,
+      "step": 6431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3643071399891682e-06,
+      "loss": 2.9527,
+      "step": 6432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3629274273542215e-06,
+      "loss": 3.0305,
+      "step": 6433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3615483026212128e-06,
+      "loss": 2.9597,
+      "step": 6434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3601697660130669e-06,
+      "loss": 2.966,
+      "step": 6435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3587918177526143e-06,
+      "loss": 3.0457,
+      "step": 6436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3574144580625854e-06,
+      "loss": 2.8566,
+      "step": 6437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3560376871656217e-06,
+      "loss": 2.9491,
+      "step": 6438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.354661505284262e-06,
+      "loss": 2.9605,
+      "step": 6439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3532859126409608e-06,
+      "loss": 3.0577,
+      "step": 6440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3519109094580663e-06,
+      "loss": 2.8833,
+      "step": 6441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3505364959578377e-06,
+      "loss": 2.9503,
+      "step": 6442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3491626723624385e-06,
+      "loss": 2.9912,
+      "step": 6443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3477894388939334e-06,
+      "loss": 3.0604,
+      "step": 6444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3464167957742957e-06,
+      "loss": 3.0623,
+      "step": 6445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3450447432253993e-06,
+      "loss": 2.937,
+      "step": 6446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3436732814690262e-06,
+      "loss": 3.0451,
+      "step": 6447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3423024107268628e-06,
+      "loss": 2.9835,
+      "step": 6448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3409321312204953e-06,
+      "loss": 3.0007,
+      "step": 6449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3395624431714226e-06,
+      "loss": 2.9611,
+      "step": 6450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3381933468010388e-06,
+      "loss": 2.8327,
+      "step": 6451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.33682484233065e-06,
+      "loss": 2.9835,
+      "step": 6452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3354569299814608e-06,
+      "loss": 2.9489,
+      "step": 6453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3340896099745832e-06,
+      "loss": 2.9684,
+      "step": 6454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3327228825310352e-06,
+      "loss": 3.0338,
+      "step": 6455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3313567478717333e-06,
+      "loss": 3.011,
+      "step": 6456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3299912062175041e-06,
+      "loss": 3.0337,
+      "step": 6457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3286262577890736e-06,
+      "loss": 3.0144,
+      "step": 6458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3272619028070755e-06,
+      "loss": 2.9394,
+      "step": 6459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3258981414920463e-06,
+      "loss": 2.8528,
+      "step": 6460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.324534974064426e-06,
+      "loss": 2.9066,
+      "step": 6461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3231724007445607e-06,
+      "loss": 2.9057,
+      "step": 6462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3218104217526955e-06,
+      "loss": 2.9657,
+      "step": 6463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3204490373089868e-06,
+      "loss": 2.9815,
+      "step": 6464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.319088247633487e-06,
+      "loss": 2.962,
+      "step": 6465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3177280529461583e-06,
+      "loss": 2.8924,
+      "step": 6466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3163684534668653e-06,
+      "loss": 2.8624,
+      "step": 6467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3150094494153737e-06,
+      "loss": 2.9885,
+      "step": 6468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3136510410113557e-06,
+      "loss": 2.8316,
+      "step": 6469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3122932284743866e-06,
+      "loss": 2.904,
+      "step": 6470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3109360120239473e-06,
+      "loss": 2.9576,
+      "step": 6471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3095793918794163e-06,
+      "loss": 2.8911,
+      "step": 6472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3082233682600826e-06,
+      "loss": 3.019,
+      "step": 6473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3068679413851366e-06,
+      "loss": 3.0031,
+      "step": 6474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3055131114736691e-06,
+      "loss": 3.0101,
+      "step": 6475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.304158878744679e-06,
+      "loss": 2.9219,
+      "step": 6476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.302805243417064e-06,
+      "loss": 2.8727,
+      "step": 6477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3014522057096297e-06,
+      "loss": 2.9133,
+      "step": 6478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3000997658410818e-06,
+      "loss": 2.9316,
+      "step": 6479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2987479240300328e-06,
+      "loss": 2.92,
+      "step": 6480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2973966804949962e-06,
+      "loss": 2.8466,
+      "step": 6481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.296046035454387e-06,
+      "loss": 3.1512,
+      "step": 6482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2946959891265281e-06,
+      "loss": 2.9286,
+      "step": 6483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2933465417296405e-06,
+      "loss": 2.9921,
+      "step": 6484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2919976934818517e-06,
+      "loss": 2.9151,
+      "step": 6485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2906494446011935e-06,
+      "loss": 2.8039,
+      "step": 6486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2893017953055965e-06,
+      "loss": 2.8463,
+      "step": 6487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2879547458128971e-06,
+      "loss": 2.8879,
+      "step": 6488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.286608296340835e-06,
+      "loss": 2.8649,
+      "step": 6489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.285262447107054e-06,
+      "loss": 2.9565,
+      "step": 6490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.283917198329096e-06,
+      "loss": 2.907,
+      "step": 6491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2825725502244108e-06,
+      "loss": 2.9213,
+      "step": 6492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2812285030103506e-06,
+      "loss": 2.9933,
+      "step": 6493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.279885056904166e-06,
+      "loss": 3.0749,
+      "step": 6494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2785422121230169e-06,
+      "loss": 2.9021,
+      "step": 6495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2771999688839603e-06,
+      "loss": 2.8948,
+      "step": 6496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2758583274039594e-06,
+      "loss": 2.9782,
+      "step": 6497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.274517287899879e-06,
+      "loss": 3.0815,
+      "step": 6498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2731768505884872e-06,
+      "loss": 2.9139,
+      "step": 6499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2718370156864556e-06,
+      "loss": 2.8527,
+      "step": 6500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2704977834103539e-06,
+      "loss": 2.9462,
+      "step": 6501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2691591539766612e-06,
+      "loss": 3.0191,
+      "step": 6502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2678211276017528e-06,
+      "loss": 2.8979,
+      "step": 6503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2664837045019096e-06,
+      "loss": 2.9897,
+      "step": 6504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2651468848933174e-06,
+      "loss": 2.9702,
+      "step": 6505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2638106689920581e-06,
+      "loss": 2.9966,
+      "step": 6506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.262475057014123e-06,
+      "loss": 2.9332,
+      "step": 6507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2611400491753979e-06,
+      "loss": 2.9107,
+      "step": 6508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2598056456916808e-06,
+      "loss": 3.0487,
+      "step": 6509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2584718467786632e-06,
+      "loss": 3.0575,
+      "step": 6510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2571386526519452e-06,
+      "loss": 3.0099,
+      "step": 6511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.255806063527023e-06,
+      "loss": 2.9823,
+      "step": 6512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2544740796193e-06,
+      "loss": 2.8569,
+      "step": 6513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2531427011440817e-06,
+      "loss": 2.8147,
+      "step": 6514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2518119283165708e-06,
+      "loss": 2.9127,
+      "step": 6515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2504817613518788e-06,
+      "loss": 2.9474,
+      "step": 6516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2491522004650131e-06,
+      "loss": 2.9564,
+      "step": 6517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2478232458708867e-06,
+      "loss": 2.911,
+      "step": 6518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2464948977843138e-06,
+      "loss": 2.8054,
+      "step": 6519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.245167156420012e-06,
+      "loss": 2.8748,
+      "step": 6520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2438400219925994e-06,
+      "loss": 2.9365,
+      "step": 6521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2425134947165935e-06,
+      "loss": 2.9777,
+      "step": 6522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2411875748064195e-06,
+      "loss": 2.8913,
+      "step": 6523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2398622624763979e-06,
+      "loss": 2.9682,
+      "step": 6524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2385375579407555e-06,
+      "loss": 3.0028,
+      "step": 6525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2372134614136217e-06,
+      "loss": 2.9588,
+      "step": 6526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2358899731090202e-06,
+      "loss": 2.9432,
+      "step": 6527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2345670932408883e-06,
+      "loss": 2.8636,
+      "step": 6528
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2332448220230531e-06,
+      "loss": 3.0641,
+      "step": 6529
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2319231596692522e-06,
+      "loss": 2.8434,
+      "step": 6530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2306021063931179e-06,
+      "loss": 2.9019,
+      "step": 6531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2292816624081881e-06,
+      "loss": 2.9655,
+      "step": 6532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2279618279279043e-06,
+      "loss": 2.8886,
+      "step": 6533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2266426031656027e-06,
+      "loss": 2.9669,
+      "step": 6534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2253239883345275e-06,
+      "loss": 3.0067,
+      "step": 6535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2240059836478185e-06,
+      "loss": 2.84,
+      "step": 6536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2226885893185225e-06,
+      "loss": 2.8761,
+      "step": 6537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2213718055595847e-06,
+      "loss": 2.8802,
+      "step": 6538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2200556325838513e-06,
+      "loss": 2.9986,
+      "step": 6539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2187400706040731e-06,
+      "loss": 2.9141,
+      "step": 6540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2174251198328957e-06,
+      "loss": 2.9751,
+      "step": 6541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2161107804828732e-06,
+      "loss": 2.9052,
+      "step": 6542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2147970527664537e-06,
+      "loss": 2.9981,
+      "step": 6543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2134839368959928e-06,
+      "loss": 3.0395,
+      "step": 6544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2121714330837453e-06,
+      "loss": 2.8733,
+      "step": 6545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2108595415418633e-06,
+      "loss": 3.049,
+      "step": 6546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2095482624824051e-06,
+      "loss": 2.8508,
+      "step": 6547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2082375961173271e-06,
+      "loss": 3.0048,
+      "step": 6548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2069275426584887e-06,
+      "loss": 2.9926,
+      "step": 6549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2056181023176473e-06,
+      "loss": 3.0022,
+      "step": 6550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2043092753064622e-06,
+      "loss": 3.0446,
+      "step": 6551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2030010618364979e-06,
+      "loss": 2.9844,
+      "step": 6552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2016934621192117e-06,
+      "loss": 2.9731,
+      "step": 6553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2003864763659689e-06,
+      "loss": 2.99,
+      "step": 6554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1990801047880307e-06,
+      "loss": 2.9719,
+      "step": 6555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1977743475965615e-06,
+      "loss": 2.9253,
+      "step": 6556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1964692050026261e-06,
+      "loss": 2.8398,
+      "step": 6557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1951646772171904e-06,
+      "loss": 3.0141,
+      "step": 6558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1938607644511207e-06,
+      "loss": 2.8856,
+      "step": 6559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1925574669151813e-06,
+      "loss": 2.7933,
+      "step": 6560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1912547848200418e-06,
+      "loss": 3.1583,
+      "step": 6561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1899527183762665e-06,
+      "loss": 2.7817,
+      "step": 6562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1886512677943258e-06,
+      "loss": 2.9772,
+      "step": 6563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1873504332845882e-06,
+      "loss": 2.8615,
+      "step": 6564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1860502150573211e-06,
+      "loss": 3.0112,
+      "step": 6565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1847506133226954e-06,
+      "loss": 2.9127,
+      "step": 6566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1834516282907776e-06,
+      "loss": 2.9077,
+      "step": 6567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1821532601715423e-06,
+      "loss": 2.8634,
+      "step": 6568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1808555091748558e-06,
+      "loss": 2.9844,
+      "step": 6569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.17955837551049e-06,
+      "loss": 2.8796,
+      "step": 6570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1782618593881173e-06,
+      "loss": 3.1059,
+      "step": 6571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.176965961017305e-06,
+      "loss": 3.1196,
+      "step": 6572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1756706806075274e-06,
+      "loss": 3.0475,
+      "step": 6573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1743760183681529e-06,
+      "loss": 2.8603,
+      "step": 6574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1730819745084536e-06,
+      "loss": 3.031,
+      "step": 6575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1717885492376012e-06,
+      "loss": 2.891,
+      "step": 6576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.170495742764667e-06,
+      "loss": 2.9373,
+      "step": 6577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1692035552986226e-06,
+      "loss": 3.019,
+      "step": 6578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1679119870483368e-06,
+      "loss": 2.9649,
+      "step": 6579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1666210382225846e-06,
+      "loss": 2.924,
+      "step": 6580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1653307090300326e-06,
+      "loss": 3.0808,
+      "step": 6581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1640409996792535e-06,
+      "loss": 2.8996,
+      "step": 6582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.16275191037872e-06,
+      "loss": 3.0422,
+      "step": 6583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1614634413367986e-06,
+      "loss": 2.9843,
+      "step": 6584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1601755927617625e-06,
+      "loss": 3.0256,
+      "step": 6585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1588883648617776e-06,
+      "loss": 2.996,
+      "step": 6586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1576017578449188e-06,
+      "loss": 2.9893,
+      "step": 6587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1563157719191508e-06,
+      "loss": 2.9584,
+      "step": 6588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1550304072923446e-06,
+      "loss": 2.9416,
+      "step": 6589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1537456641722682e-06,
+      "loss": 2.871,
+      "step": 6590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1524615427665886e-06,
+      "loss": 2.9196,
+      "step": 6591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1511780432828744e-06,
+      "loss": 2.9549,
+      "step": 6592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1498951659285907e-06,
+      "loss": 3.0311,
+      "step": 6593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1486129109111044e-06,
+      "loss": 2.8421,
+      "step": 6594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1473312784376829e-06,
+      "loss": 3.0468,
+      "step": 6595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1460502687154861e-06,
+      "loss": 2.9647,
+      "step": 6596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1447698819515858e-06,
+      "loss": 2.8668,
+      "step": 6597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.14349011835294e-06,
+      "loss": 2.976,
+      "step": 6598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1422109781264146e-06,
+      "loss": 2.9495,
+      "step": 6599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1409324614787698e-06,
+      "loss": 2.8908,
+      "step": 6600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1396545686166677e-06,
+      "loss": 2.867,
+      "step": 6601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1383772997466702e-06,
+      "loss": 2.9791,
+      "step": 6602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1371006550752346e-06,
+      "loss": 3.0383,
+      "step": 6603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1358246348087214e-06,
+      "loss": 2.9209,
+      "step": 6604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1345492391533874e-06,
+      "loss": 2.9439,
+      "step": 6605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1332744683153895e-06,
+      "loss": 2.9044,
+      "step": 6606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1320003225007841e-06,
+      "loss": 3.0979,
+      "step": 6607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1307268019155265e-06,
+      "loss": 2.9435,
+      "step": 6608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.129453906765471e-06,
+      "loss": 2.9601,
+      "step": 6609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1281816372563681e-06,
+      "loss": 2.7556,
+      "step": 6610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1269099935938716e-06,
+      "loss": 2.9242,
+      "step": 6611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.12563897598353e-06,
+      "loss": 2.9748,
+      "step": 6612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1243685846307934e-06,
+      "loss": 3.0097,
+      "step": 6613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1230988197410108e-06,
+      "loss": 3.0511,
+      "step": 6614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1218296815194275e-06,
+      "loss": 2.991,
+      "step": 6615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1205611701711888e-06,
+      "loss": 2.9367,
+      "step": 6616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1192932859013395e-06,
+      "loss": 2.9512,
+      "step": 6617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1180260289148232e-06,
+      "loss": 2.9297,
+      "step": 6618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1167593994164794e-06,
+      "loss": 3.1272,
+      "step": 6619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1154933976110488e-06,
+      "loss": 3.0146,
+      "step": 6620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1142280237031716e-06,
+      "loss": 2.99,
+      "step": 6621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1129632778973815e-06,
+      "loss": 2.8045,
+      "step": 6622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.111699160398117e-06,
+      "loss": 2.9389,
+      "step": 6623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1104356714097087e-06,
+      "loss": 2.8997,
+      "step": 6624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1091728111363914e-06,
+      "loss": 2.9689,
+      "step": 6625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1079105797822948e-06,
+      "loss": 2.9777,
+      "step": 6626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1066489775514483e-06,
+      "loss": 2.9717,
+      "step": 6627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.10538800464778e-06,
+      "loss": 2.7338,
+      "step": 6628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1041276612751133e-06,
+      "loss": 2.9502,
+      "step": 6629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1028679476371752e-06,
+      "loss": 3.0261,
+      "step": 6630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1016088639375838e-06,
+      "loss": 2.8455,
+      "step": 6631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.1003504103798612e-06,
+      "loss": 3.0282,
+      "step": 6632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.099092587167428e-06,
+      "loss": 2.9589,
+      "step": 6633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0978353945035964e-06,
+      "loss": 2.9127,
+      "step": 6634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0965788325915834e-06,
+      "loss": 2.9759,
+      "step": 6635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0953229016345008e-06,
+      "loss": 3.0512,
+      "step": 6636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0940676018353614e-06,
+      "loss": 2.8967,
+      "step": 6637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0928129333970706e-06,
+      "loss": 2.9881,
+      "step": 6638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.091558896522436e-06,
+      "loss": 2.9782,
+      "step": 6639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0903054914141636e-06,
+      "loss": 3.0022,
+      "step": 6640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0890527182748534e-06,
+      "loss": 2.8635,
+      "step": 6641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0878005773070078e-06,
+      "loss": 3.0645,
+      "step": 6642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.086549068713022e-06,
+      "loss": 2.8965,
+      "step": 6643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0852981926951934e-06,
+      "loss": 2.9541,
+      "step": 6644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0840479494557155e-06,
+      "loss": 2.9914,
+      "step": 6645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.082798339196679e-06,
+      "loss": 3.0752,
+      "step": 6646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.081549362120075e-06,
+      "loss": 2.9421,
+      "step": 6647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0803010184277867e-06,
+      "loss": 2.9748,
+      "step": 6648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0790533083216008e-06,
+      "loss": 3.0209,
+      "step": 6649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0778062320031974e-06,
+      "loss": 2.8491,
+      "step": 6650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0765597896741564e-06,
+      "loss": 3.019,
+      "step": 6651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.075313981535956e-06,
+      "loss": 3.0106,
+      "step": 6652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0740688077899675e-06,
+      "loss": 2.7453,
+      "step": 6653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0728242686374663e-06,
+      "loss": 3.0419,
+      "step": 6654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.071580364279617e-06,
+      "loss": 2.9716,
+      "step": 6655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0703370949174913e-06,
+      "loss": 2.9401,
+      "step": 6656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0690944607520497e-06,
+      "loss": 2.9587,
+      "step": 6657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0678524619841557e-06,
+      "loss": 2.9652,
+      "step": 6658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0666110988145657e-06,
+      "loss": 2.9876,
+      "step": 6659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.065370371443936e-06,
+      "loss": 3.0782,
+      "step": 6660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.064130280072822e-06,
+      "loss": 2.8106,
+      "step": 6661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0628908249016706e-06,
+      "loss": 3.0187,
+      "step": 6662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0616520061308322e-06,
+      "loss": 2.8117,
+      "step": 6663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0604138239605482e-06,
+      "loss": 2.7943,
+      "step": 6664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0591762785909627e-06,
+      "loss": 2.9466,
+      "step": 6665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.057939370222113e-06,
+      "loss": 2.7996,
+      "step": 6666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0567030990539362e-06,
+      "loss": 2.9039,
+      "step": 6667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.055467465286265e-06,
+      "loss": 2.9686,
+      "step": 6668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0542324691188277e-06,
+      "loss": 3.1271,
+      "step": 6669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0529981107512533e-06,
+      "loss": 2.8854,
+      "step": 6670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.051764390383062e-06,
+      "loss": 2.8622,
+      "step": 6671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.050531308213676e-06,
+      "loss": 2.9778,
+      "step": 6672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.049298864442414e-06,
+      "loss": 2.9466,
+      "step": 6673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0480670592684866e-06,
+      "loss": 2.8591,
+      "step": 6674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0468358928910088e-06,
+      "loss": 2.9288,
+      "step": 6675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0456053655089854e-06,
+      "loss": 2.9191,
+      "step": 6676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0443754773213228e-06,
+      "loss": 2.9715,
+      "step": 6677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.043146228526819e-06,
+      "loss": 2.8987,
+      "step": 6678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.041917619324173e-06,
+      "loss": 2.8904,
+      "step": 6679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0406896499119818e-06,
+      "loss": 3.0376,
+      "step": 6680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0394623204887317e-06,
+      "loss": 2.9048,
+      "step": 6681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0382356312528135e-06,
+      "loss": 2.8876,
+      "step": 6682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.037009582402509e-06,
+      "loss": 3.0891,
+      "step": 6683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0357841741359993e-06,
+      "loss": 2.9355,
+      "step": 6684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345594066513615e-06,
+      "loss": 2.6994,
+      "step": 6685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0333352801465685e-06,
+      "loss": 2.9244,
+      "step": 6686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032111794819492e-06,
+      "loss": 2.975,
+      "step": 6687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0308889508678944e-06,
+      "loss": 3.0354,
+      "step": 6688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0296667484894418e-06,
+      "loss": 3.0276,
+      "step": 6689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0284451878816892e-06,
+      "loss": 2.9926,
+      "step": 6690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0272242692420936e-06,
+      "loss": 3.0457,
+      "step": 6691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0260039927680072e-06,
+      "loss": 2.9936,
+      "step": 6692
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0247843586566747e-06,
+      "loss": 2.9772,
+      "step": 6693
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0235653671052403e-06,
+      "loss": 2.8455,
+      "step": 6694
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0223470183107449e-06,
+      "loss": 2.8418,
+      "step": 6695
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.021129312470125e-06,
+      "loss": 3.0453,
+      "step": 6696
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0199122497802095e-06,
+      "loss": 3.0736,
+      "step": 6697
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0186958304377276e-06,
+      "loss": 2.8519,
+      "step": 6698
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.017480054639305e-06,
+      "loss": 2.8044,
+      "step": 6699
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.016264922581458e-06,
+      "loss": 2.9372,
+      "step": 6700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.015050434460606e-06,
+      "loss": 3.0355,
+      "step": 6701
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0138365904730578e-06,
+      "loss": 2.9431,
+      "step": 6702
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0126233908150223e-06,
+      "loss": 2.9846,
+      "step": 6703
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0114108356826025e-06,
+      "loss": 2.9413,
+      "step": 6704
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0101989252717982e-06,
+      "loss": 2.9013,
+      "step": 6705
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0089876597785052e-06,
+      "loss": 2.9252,
+      "step": 6706
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.007777039398512e-06,
+      "loss": 2.9893,
+      "step": 6707
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0065670643275082e-06,
+      "loss": 2.9818,
+      "step": 6708
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0053577347610726e-06,
+      "loss": 3.0369,
+      "step": 6709
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.004149050894685e-06,
+      "loss": 2.8529,
+      "step": 6710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.00294101292372e-06,
+      "loss": 3.0094,
+      "step": 6711
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0017336210434436e-06,
+      "loss": 2.9162,
+      "step": 6712
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0005268754490238e-06,
+      "loss": 2.9058,
+      "step": 6713
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.993207763355162e-07,
+      "loss": 2.9433,
+      "step": 6714
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.981153238978825e-07,
+      "loss": 2.9548,
+      "step": 6715
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.96910518330969e-07,
+      "loss": 2.8869,
+      "step": 6716
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.95706359829524e-07,
+      "loss": 3.0105,
+      "step": 6717
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.945028485881919e-07,
+      "loss": 2.9289,
+      "step": 6718
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.932999848015056e-07,
+      "loss": 2.7943,
+      "step": 6719
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.920977686639016e-07,
+      "loss": 2.8564,
+      "step": 6720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.908962003697054e-07,
+      "loss": 2.9234,
+      "step": 6721
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.89695280113141e-07,
+      "loss": 2.826,
+      "step": 6722
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.884950080883276e-07,
+      "loss": 2.9397,
+      "step": 6723
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.872953844892785e-07,
+      "loss": 2.8266,
+      "step": 6724
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.86096409509904e-07,
+      "loss": 2.9997,
+      "step": 6725
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.84898083344006e-07,
+      "loss": 2.8053,
+      "step": 6726
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.837004061852861e-07,
+      "loss": 3.0131,
+      "step": 6727
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.825033782273363e-07,
+      "loss": 2.9205,
+      "step": 6728
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.813069996636464e-07,
+      "loss": 2.8808,
+      "step": 6729
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.801112706876037e-07,
+      "loss": 2.9022,
+      "step": 6730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.789161914924845e-07,
+      "loss": 3.0418,
+      "step": 6731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.777217622714647e-07,
+      "loss": 3.0771,
+      "step": 6732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.76527983217611e-07,
+      "loss": 2.9441,
+      "step": 6733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.753348545238929e-07,
+      "loss": 2.9388,
+      "step": 6734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.741423763831648e-07,
+      "loss": 3.0451,
+      "step": 6735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.72950548988182e-07,
+      "loss": 2.9066,
+      "step": 6736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.717593725315955e-07,
+      "loss": 2.8513,
+      "step": 6737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.705688472059455e-07,
+      "loss": 3.0561,
+      "step": 6738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.693789732036736e-07,
+      "loss": 2.9623,
+      "step": 6739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.681897507171095e-07,
+      "loss": 3.0463,
+      "step": 6740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.670011799384826e-07,
+      "loss": 3.1053,
+      "step": 6741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.658132610599164e-07,
+      "loss": 3.1285,
+      "step": 6742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.646259942734242e-07,
+      "loss": 2.938,
+      "step": 6743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63439379770923e-07,
+      "loss": 2.9909,
+      "step": 6744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.622534177442144e-07,
+      "loss": 2.9501,
+      "step": 6745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.610681083850027e-07,
+      "loss": 2.8823,
+      "step": 6746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.598834518848798e-07,
+      "loss": 2.9418,
+      "step": 6747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.586994484353368e-07,
+      "loss": 2.9709,
+      "step": 6748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.575160982277588e-07,
+      "loss": 2.7997,
+      "step": 6749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.563334014534227e-07,
+      "loss": 3.0509,
+      "step": 6750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.551513583035026e-07,
+      "loss": 2.8936,
+      "step": 6751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.53969968969064e-07,
+      "loss": 2.9548,
+      "step": 6752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.527892336410704e-07,
+      "loss": 2.8708,
+      "step": 6753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.516091525103765e-07,
+      "loss": 2.9979,
+      "step": 6754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.504297257677331e-07,
+      "loss": 3.1378,
+      "step": 6755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.492509536037852e-07,
+      "loss": 2.9994,
+      "step": 6756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.480728362090691e-07,
+      "loss": 2.9624,
+      "step": 6757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.468953737740205e-07,
+      "loss": 2.8547,
+      "step": 6758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.457185664889629e-07,
+      "loss": 2.9557,
+      "step": 6759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.44542414544119e-07,
+      "loss": 2.8615,
+      "step": 6760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.433669181296046e-07,
+      "loss": 2.9305,
+      "step": 6761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.421920774354255e-07,
+      "loss": 2.9254,
+      "step": 6762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.410178926514895e-07,
+      "loss": 3.1095,
+      "step": 6763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.398443639675897e-07,
+      "loss": 2.9364,
+      "step": 6764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.386714915734196e-07,
+      "loss": 2.8476,
+      "step": 6765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.374992756585621e-07,
+      "loss": 2.8911,
+      "step": 6766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.363277164124963e-07,
+      "loss": 2.7923,
+      "step": 6767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.35156814024597e-07,
+      "loss": 2.963,
+      "step": 6768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.339865686841271e-07,
+      "loss": 2.9644,
+      "step": 6769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.328169805802506e-07,
+      "loss": 3.0849,
+      "step": 6770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.316480499020181e-07,
+      "loss": 2.9783,
+      "step": 6771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.304797768383789e-07,
+      "loss": 2.9715,
+      "step": 6772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.293121615781742e-07,
+      "loss": 2.8879,
+      "step": 6773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.281452043101391e-07,
+      "loss": 2.9776,
+      "step": 6774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.269789052229039e-07,
+      "loss": 2.9575,
+      "step": 6775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.258132645049883e-07,
+      "loss": 3.0027,
+      "step": 6776
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.24648282344811e-07,
+      "loss": 3.0789,
+      "step": 6777
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.234839589306788e-07,
+      "loss": 2.893,
+      "step": 6778
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.223202944507952e-07,
+      "loss": 2.7943,
+      "step": 6779
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.211572890932591e-07,
+      "loss": 2.9166,
+      "step": 6780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.199949430460575e-07,
+      "loss": 2.9696,
+      "step": 6781
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.188332564970753e-07,
+      "loss": 2.911,
+      "step": 6782
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.176722296340884e-07,
+      "loss": 2.9714,
+      "step": 6783
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.165118626447694e-07,
+      "loss": 2.9085,
+      "step": 6784
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.153521557166783e-07,
+      "loss": 2.9067,
+      "step": 6785
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.141931090372735e-07,
+      "loss": 3.0184,
+      "step": 6786
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.130347227939062e-07,
+      "loss": 2.9866,
+      "step": 6787
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.118769971738178e-07,
+      "loss": 2.8098,
+      "step": 6788
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.107199323641469e-07,
+      "loss": 3.062,
+      "step": 6789
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.0956352855192e-07,
+      "loss": 2.7771,
+      "step": 6790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.084077859240626e-07,
+      "loss": 3.1827,
+      "step": 6791
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.072527046673896e-07,
+      "loss": 2.8271,
+      "step": 6792
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.060982849686123e-07,
+      "loss": 2.8569,
+      "step": 6793
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.049445270143292e-07,
+      "loss": 3.0254,
+      "step": 6794
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.037914309910378e-07,
+      "loss": 2.9941,
+      "step": 6795
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.026389970851268e-07,
+      "loss": 3.0731,
+      "step": 6796
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.014872254828755e-07,
+      "loss": 3.0507,
+      "step": 6797
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.003361163704599e-07,
+      "loss": 3.1157,
+      "step": 6798
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.991856699339447e-07,
+      "loss": 2.9259,
+      "step": 6799
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.980358863592914e-07,
+      "loss": 3.0044,
+      "step": 6800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.968867658323538e-07,
+      "loss": 2.8679,
+      "step": 6801
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.957383085388733e-07,
+      "loss": 2.964,
+      "step": 6802
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.945905146644939e-07,
+      "loss": 2.9724,
+      "step": 6803
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.934433843947432e-07,
+      "loss": 2.8675,
+      "step": 6804
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.922969179150465e-07,
+      "loss": 2.7768,
+      "step": 6805
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.911511154107188e-07,
+      "loss": 2.9384,
+      "step": 6806
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.900059770669706e-07,
+      "loss": 3.0025,
+      "step": 6807
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.888615030689052e-07,
+      "loss": 2.9739,
+      "step": 6808
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.877176936015136e-07,
+      "loss": 3.0235,
+      "step": 6809
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.865745488496858e-07,
+      "loss": 2.9111,
+      "step": 6810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.854320689981983e-07,
+      "loss": 3.0324,
+      "step": 6811
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.842902542317284e-07,
+      "loss": 3.0159,
+      "step": 6812
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.831491047348356e-07,
+      "loss": 2.7685,
+      "step": 6813
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.820086206919798e-07,
+      "loss": 2.9921,
+      "step": 6814
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.808688022875112e-07,
+      "loss": 2.866,
+      "step": 6815
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.797296497056695e-07,
+      "loss": 2.9134,
+      "step": 6816
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.785911631305911e-07,
+      "loss": 2.9936,
+      "step": 6817
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.774533427463005e-07,
+      "loss": 2.7464,
+      "step": 6818
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.763161887367183e-07,
+      "loss": 2.9574,
+      "step": 6819
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.751797012856561e-07,
+      "loss": 3.0713,
+      "step": 6820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.740438805768148e-07,
+      "loss": 3.0707,
+      "step": 6821
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.729087267937946e-07,
+      "loss": 2.8552,
+      "step": 6822
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.717742401200796e-07,
+      "loss": 3.108,
+      "step": 6823
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.706404207390528e-07,
+      "loss": 3.0228,
+      "step": 6824
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.695072688339845e-07,
+      "loss": 2.9372,
+      "step": 6825
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.683747845880391e-07,
+      "loss": 3.057,
+      "step": 6826
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.672429681842748e-07,
+      "loss": 3.0118,
+      "step": 6827
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.661118198056384e-07,
+      "loss": 3.1016,
+      "step": 6828
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.649813396349722e-07,
+      "loss": 2.8703,
+      "step": 6829
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.638515278550058e-07,
+      "loss": 2.9522,
+      "step": 6830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.627223846483657e-07,
+      "loss": 2.998,
+      "step": 6831
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.615939101975684e-07,
+      "loss": 2.8087,
+      "step": 6832
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.604661046850215e-07,
+      "loss": 3.0068,
+      "step": 6833
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.593389682930275e-07,
+      "loss": 2.9428,
+      "step": 6834
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.582125012037751e-07,
+      "loss": 2.9449,
+      "step": 6835
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.570867035993507e-07,
+      "loss": 2.9039,
+      "step": 6836
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.559615756617273e-07,
+      "loss": 2.9884,
+      "step": 6837
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.548371175727743e-07,
+      "loss": 2.8283,
+      "step": 6838
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.537133295142519e-07,
+      "loss": 2.8956,
+      "step": 6839
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.525902116678075e-07,
+      "loss": 2.9418,
+      "step": 6840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.51467764214986e-07,
+      "loss": 2.8167,
+      "step": 6841
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.503459873372205e-07,
+      "loss": 2.9639,
+      "step": 6842
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.492248812158388e-07,
+      "loss": 2.9995,
+      "step": 6843
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.481044460320553e-07,
+      "loss": 2.9987,
+      "step": 6844
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.469846819669802e-07,
+      "loss": 2.9872,
+      "step": 6845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.458655892016148e-07,
+      "loss": 2.8429,
+      "step": 6846
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.447471679168489e-07,
+      "loss": 3.0034,
+      "step": 6847
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.436294182934678e-07,
+      "loss": 2.9134,
+      "step": 6848
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.425123405121443e-07,
+      "loss": 2.9161,
+      "step": 6849
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.413959347534456e-07,
+      "loss": 2.8299,
+      "step": 6850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.402802011978295e-07,
+      "loss": 2.8014,
+      "step": 6851
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.391651400256439e-07,
+      "loss": 2.901,
+      "step": 6852
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.380507514171316e-07,
+      "loss": 3.0848,
+      "step": 6853
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.369370355524203e-07,
+      "loss": 2.9125,
+      "step": 6854
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.358239926115363e-07,
+      "loss": 2.9855,
+      "step": 6855
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.347116227743901e-07,
+      "loss": 2.9744,
+      "step": 6856
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.335999262207889e-07,
+      "loss": 2.9531,
+      "step": 6857
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.324889031304295e-07,
+      "loss": 3.0168,
+      "step": 6858
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.313785536828972e-07,
+      "loss": 2.9595,
+      "step": 6859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.302688780576723e-07,
+      "loss": 2.9388,
+      "step": 6860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.29159876434124e-07,
+      "loss": 2.8309,
+      "step": 6861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.280515489915142e-07,
+      "loss": 2.8441,
+      "step": 6862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.269438959089926e-07,
+      "loss": 2.9055,
+      "step": 6863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.258369173656023e-07,
+      "loss": 3.116,
+      "step": 6864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.247306135402788e-07,
+      "loss": 3.102,
+      "step": 6865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.23624984611845e-07,
+      "loss": 3.0509,
+      "step": 6866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.22520030759017e-07,
+      "loss": 3.1044,
+      "step": 6867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.214157521604005e-07,
+      "loss": 2.7671,
+      "step": 6868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.203121489944932e-07,
+      "loss": 2.981,
+      "step": 6869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.192092214396835e-07,
+      "loss": 2.882,
+      "step": 6870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.1810696967425e-07,
+      "loss": 2.9851,
+      "step": 6871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.17005393876364e-07,
+      "loss": 2.985,
+      "step": 6872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.159044942240829e-07,
+      "loss": 2.974,
+      "step": 6873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.148042708953608e-07,
+      "loss": 2.9493,
+      "step": 6874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.137047240680368e-07,
+      "loss": 2.9527,
+      "step": 6875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.126058539198439e-07,
+      "loss": 3.0367,
+      "step": 6876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.115076606284072e-07,
+      "loss": 2.9334,
+      "step": 6877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.104101443712381e-07,
+      "loss": 2.8896,
+      "step": 6878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.093133053257424e-07,
+      "loss": 2.9098,
+      "step": 6879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.082171436692116e-07,
+      "loss": 2.9657,
+      "step": 6880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.071216595788356e-07,
+      "loss": 2.7387,
+      "step": 6881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.060268532316867e-07,
+      "loss": 3.1198,
+      "step": 6882
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.049327248047323e-07,
+      "loss": 3.0048,
+      "step": 6883
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.038392744748297e-07,
+      "loss": 3.196,
+      "step": 6884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.027465024187243e-07,
+      "loss": 2.9371,
+      "step": 6885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.016544088130552e-07,
+      "loss": 2.9354,
+      "step": 6886
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.005629938343479e-07,
+      "loss": 2.9457,
+      "step": 6887
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.994722576590219e-07,
+      "loss": 2.8997,
+      "step": 6888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.983822004633856e-07,
+      "loss": 3.0298,
+      "step": 6889
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.972928224236348e-07,
+      "loss": 2.8726,
+      "step": 6890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.962041237158629e-07,
+      "loss": 3.0756,
+      "step": 6891
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.951161045160455e-07,
+      "loss": 2.8731,
+      "step": 6892
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.940287650000534e-07,
+      "loss": 3.0337,
+      "step": 6893
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.929421053436437e-07,
+      "loss": 3.035,
+      "step": 6894
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.918561257224667e-07,
+      "loss": 3.0591,
+      "step": 6895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.907708263120634e-07,
+      "loss": 2.9708,
+      "step": 6896
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.896862072878608e-07,
+      "loss": 2.8835,
+      "step": 6897
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.886022688251805e-07,
+      "loss": 2.9735,
+      "step": 6898
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.875190110992298e-07,
+      "loss": 3.0706,
+      "step": 6899
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.864364342851089e-07,
+      "loss": 2.8305,
+      "step": 6900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.853545385578082e-07,
+      "loss": 2.9527,
+      "step": 6901
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.842733240922057e-07,
+      "loss": 2.9922,
+      "step": 6902
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.831927910630721e-07,
+      "loss": 2.7921,
+      "step": 6903
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.82112939645065e-07,
+      "loss": 2.8823,
+      "step": 6904
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.810337700127341e-07,
+      "loss": 3.1134,
+      "step": 6905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.799552823405165e-07,
+      "loss": 3.0205,
+      "step": 6906
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.788774768027418e-07,
+      "loss": 2.9945,
+      "step": 6907
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.778003535736295e-07,
+      "loss": 2.9896,
+      "step": 6908
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.76723912827283e-07,
+      "loss": 3.0067,
+      "step": 6909
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.756481547377054e-07,
+      "loss": 3.0654,
+      "step": 6910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.745730794787798e-07,
+      "loss": 3.0407,
+      "step": 6911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.734986872242855e-07,
+      "loss": 2.9758,
+      "step": 6912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.724249781478871e-07,
+      "loss": 2.8424,
+      "step": 6913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.71351952423141e-07,
+      "loss": 2.9866,
+      "step": 6914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.702796102234939e-07,
+      "loss": 2.9578,
+      "step": 6915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.692079517222789e-07,
+      "loss": 2.7593,
+      "step": 6916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.681369770927227e-07,
+      "loss": 2.8406,
+      "step": 6917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.670666865079369e-07,
+      "loss": 2.9094,
+      "step": 6918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.659970801409261e-07,
+      "loss": 2.8004,
+      "step": 6919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.649281581645829e-07,
+      "loss": 2.9207,
+      "step": 6920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.638599207516901e-07,
+      "loss": 2.8039,
+      "step": 6921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.627923680749194e-07,
+      "loss": 2.8771,
+      "step": 6922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.617255003068297e-07,
+      "loss": 2.9927,
+      "step": 6923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.60659317619874e-07,
+      "loss": 2.8153,
+      "step": 6924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.59593820186389e-07,
+      "loss": 3.006,
+      "step": 6925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.58529008178604e-07,
+      "loss": 2.9864,
+      "step": 6926
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.574648817686392e-07,
+      "loss": 2.9829,
+      "step": 6927
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.564014411284976e-07,
+      "loss": 3.0244,
+      "step": 6928
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.553386864300783e-07,
+      "loss": 3.1435,
+      "step": 6929
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.542766178451649e-07,
+      "loss": 2.8608,
+      "step": 6930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.532152355454342e-07,
+      "loss": 2.9461,
+      "step": 6931
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.521545397024471e-07,
+      "loss": 3.053,
+      "step": 6932
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.510945304876566e-07,
+      "loss": 2.9482,
+      "step": 6933
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.50035208072406e-07,
+      "loss": 3.0008,
+      "step": 6934
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.48976572627923e-07,
+      "loss": 2.9953,
+      "step": 6935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.4791862432533e-07,
+      "loss": 3.0044,
+      "step": 6936
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.468613633356325e-07,
+      "loss": 3.0586,
+      "step": 6937
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.458047898297294e-07,
+      "loss": 2.8826,
+      "step": 6938
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.447489039784062e-07,
+      "loss": 3.0315,
+      "step": 6939
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.436937059523397e-07,
+      "loss": 3.0506,
+      "step": 6940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.426391959220907e-07,
+      "loss": 2.8858,
+      "step": 6941
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.415853740581136e-07,
+      "loss": 2.9559,
+      "step": 6942
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.405322405307507e-07,
+      "loss": 2.9307,
+      "step": 6943
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.3947979551023e-07,
+      "loss": 2.9558,
+      "step": 6944
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.38428039166672e-07,
+      "loss": 3.101,
+      "step": 6945
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.373769716700829e-07,
+      "loss": 2.8868,
+      "step": 6946
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.363265931903591e-07,
+      "loss": 2.9021,
+      "step": 6947
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.352769038972868e-07,
+      "loss": 2.7108,
+      "step": 6948
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.34227903960536e-07,
+      "loss": 3.024,
+      "step": 6949
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.331795935496727e-07,
+      "loss": 3.0732,
+      "step": 6950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.321319728341447e-07,
+      "loss": 2.9333,
+      "step": 6951
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.310850419832933e-07,
+      "loss": 2.7642,
+      "step": 6952
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.300388011663434e-07,
+      "loss": 2.9988,
+      "step": 6953
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.289932505524122e-07,
+      "loss": 3.0971,
+      "step": 6954
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.279483903105051e-07,
+      "loss": 3.0233,
+      "step": 6955
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.269042206095122e-07,
+      "loss": 3.0043,
+      "step": 6956
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.258607416182178e-07,
+      "loss": 3.1765,
+      "step": 6957
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.248179535052879e-07,
+      "loss": 3.0452,
+      "step": 6958
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.237758564392844e-07,
+      "loss": 2.9837,
+      "step": 6959
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.227344505886502e-07,
+      "loss": 2.993,
+      "step": 6960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.216937361217213e-07,
+      "loss": 2.7816,
+      "step": 6961
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.20653713206721e-07,
+      "loss": 2.8995,
+      "step": 6962
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.196143820117579e-07,
+      "loss": 2.8987,
+      "step": 6963
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.18575742704834e-07,
+      "loss": 2.9338,
+      "step": 6964
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.175377954538332e-07,
+      "loss": 2.9838,
+      "step": 6965
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.16500540426533e-07,
+      "loss": 2.9809,
+      "step": 6966
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.154639777905975e-07,
+      "loss": 2.9253,
+      "step": 6967
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.144281077135751e-07,
+      "loss": 2.9683,
+      "step": 6968
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.133929303629095e-07,
+      "loss": 2.9721,
+      "step": 6969
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.123584459059258e-07,
+      "loss": 2.946,
+      "step": 6970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.113246545098407e-07,
+      "loss": 2.8093,
+      "step": 6971
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.102915563417567e-07,
+      "loss": 2.9317,
+      "step": 6972
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.092591515686659e-07,
+      "loss": 2.985,
+      "step": 6973
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.08227440357449e-07,
+      "loss": 2.8151,
+      "step": 6974
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.071964228748707e-07,
+      "loss": 2.8214,
+      "step": 6975
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.061660992875896e-07,
+      "loss": 2.9921,
+      "step": 6976
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.051364697621454e-07,
+      "loss": 3.0068,
+      "step": 6977
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.041075344649706e-07,
+      "loss": 2.9319,
+      "step": 6978
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.030792935623837e-07,
+      "loss": 2.9346,
+      "step": 6979
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.020517472205917e-07,
+      "loss": 2.9251,
+      "step": 6980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.010248956056886e-07,
+      "loss": 2.8544,
+      "step": 6981
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.999987388836554e-07,
+      "loss": 2.913,
+      "step": 6982
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.989732772203634e-07,
+      "loss": 2.8902,
+      "step": 6983
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.97948510781567e-07,
+      "loss": 3.0019,
+      "step": 6984
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.969244397329128e-07,
+      "loss": 3.0095,
+      "step": 6985
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.959010642399339e-07,
+      "loss": 2.9912,
+      "step": 6986
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.948783844680484e-07,
+      "loss": 2.7878,
+      "step": 6987
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.938564005825649e-07,
+      "loss": 2.9826,
+      "step": 6988
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.928351127486777e-07,
+      "loss": 2.9586,
+      "step": 6989
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.918145211314714e-07,
+      "loss": 2.9019,
+      "step": 6990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.907946258959136e-07,
+      "loss": 2.9999,
+      "step": 6991
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.897754272068618e-07,
+      "loss": 3.034,
+      "step": 6992
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.887569252290632e-07,
+      "loss": 2.8364,
+      "step": 6993
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.877391201271472e-07,
+      "loss": 2.8098,
+      "step": 6994
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.867220120656359e-07,
+      "loss": 2.8776,
+      "step": 6995
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.857056012089336e-07,
+      "loss": 2.9488,
+      "step": 6996
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.846898877213359e-07,
+      "loss": 2.8789,
+      "step": 6997
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.836748717670238e-07,
+      "loss": 2.8322,
+      "step": 6998
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.826605535100666e-07,
+      "loss": 3.0782,
+      "step": 6999
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.81646933114421e-07,
+      "loss": 2.9685,
+      "step": 7000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.80634010743928e-07,
+      "loss": 3.0893,
+      "step": 7001
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.796217865623206e-07,
+      "loss": 2.8224,
+      "step": 7002
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.786102607332129e-07,
+      "loss": 2.9462,
+      "step": 7003
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.77599433420112e-07,
+      "loss": 3.1035,
+      "step": 7004
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.765893047864097e-07,
+      "loss": 2.9508,
+      "step": 7005
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.755798749953824e-07,
+      "loss": 3.0619,
+      "step": 7006
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.745711442101977e-07,
+      "loss": 2.9059,
+      "step": 7007
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.73563112593908e-07,
+      "loss": 2.9475,
+      "step": 7008
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.72555780309454e-07,
+      "loss": 2.9068,
+      "step": 7009
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.715491475196606e-07,
+      "loss": 2.9943,
+      "step": 7010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.70543214387242e-07,
+      "loss": 2.9509,
+      "step": 7011
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.69537981074801e-07,
+      "loss": 3.0878,
+      "step": 7012
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.685334477448213e-07,
+      "loss": 2.8814,
+      "step": 7013
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.675296145596804e-07,
+      "loss": 2.9002,
+      "step": 7014
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.665264816816369e-07,
+      "loss": 3.0744,
+      "step": 7015
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.655240492728399e-07,
+      "loss": 3.0274,
+      "step": 7016
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.645223174953241e-07,
+      "loss": 2.9509,
+      "step": 7017
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.635212865110113e-07,
+      "loss": 3.0734,
+      "step": 7018
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.625209564817098e-07,
+      "loss": 2.9315,
+      "step": 7019
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.615213275691135e-07,
+      "loss": 2.8982,
+      "step": 7020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.605223999348054e-07,
+      "loss": 3.0038,
+      "step": 7021
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.595241737402513e-07,
+      "loss": 3.0435,
+      "step": 7022
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.585266491468074e-07,
+      "loss": 3.0398,
+      "step": 7023
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.575298263157165e-07,
+      "loss": 2.9566,
+      "step": 7024
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.565337054081034e-07,
+      "loss": 2.8779,
+      "step": 7025
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.555382865849863e-07,
+      "loss": 2.9056,
+      "step": 7026
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.545435700072616e-07,
+      "loss": 2.9544,
+      "step": 7027
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.535495558357213e-07,
+      "loss": 2.9056,
+      "step": 7028
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.525562442310368e-07,
+      "loss": 3.006,
+      "step": 7029
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.515636353537696e-07,
+      "loss": 2.9578,
+      "step": 7030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.505717293643665e-07,
+      "loss": 2.7809,
+      "step": 7031
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.495805264231597e-07,
+      "loss": 2.8062,
+      "step": 7032
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.485900266903706e-07,
+      "loss": 2.865,
+      "step": 7033
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.47600230326102e-07,
+      "loss": 2.973,
+      "step": 7034
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.466111374903489e-07,
+      "loss": 3.1051,
+      "step": 7035
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.456227483429889e-07,
+      "loss": 2.8493,
+      "step": 7036
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.446350630437853e-07,
+      "loss": 2.9633,
+      "step": 7037
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.436480817523922e-07,
+      "loss": 2.9474,
+      "step": 7038
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.426618046283439e-07,
+      "loss": 3.054,
+      "step": 7039
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.416762318310654e-07,
+      "loss": 2.9796,
+      "step": 7040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.406913635198653e-07,
+      "loss": 2.7714,
+      "step": 7041
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.39707199853939e-07,
+      "loss": 2.9325,
+      "step": 7042
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.387237409923697e-07,
+      "loss": 2.9681,
+      "step": 7043
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.377409870941231e-07,
+      "loss": 2.9099,
+      "step": 7044
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.367589383180556e-07,
+      "loss": 3.0619,
+      "step": 7045
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.357775948229028e-07,
+      "loss": 2.9321,
+      "step": 7046
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.347969567672957e-07,
+      "loss": 2.854,
+      "step": 7047
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.33817024309743e-07,
+      "loss": 3.0303,
+      "step": 7048
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.328377976086425e-07,
+      "loss": 2.9594,
+      "step": 7049
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.318592768222803e-07,
+      "loss": 3.0478,
+      "step": 7050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.308814621088227e-07,
+      "loss": 3.0239,
+      "step": 7051
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.299043536263283e-07,
+      "loss": 2.9566,
+      "step": 7052
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.289279515327356e-07,
+      "loss": 2.9199,
+      "step": 7053
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.279522559858731e-07,
+      "loss": 2.8961,
+      "step": 7054
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.269772671434544e-07,
+      "loss": 2.9551,
+      "step": 7055
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.26002985163075e-07,
+      "loss": 2.8396,
+      "step": 7056
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.250294102022241e-07,
+      "loss": 2.8313,
+      "step": 7057
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.240565424182681e-07,
+      "loss": 2.9676,
+      "step": 7058
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.230843819684656e-07,
+      "loss": 2.9481,
+      "step": 7059
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.221129290099553e-07,
+      "loss": 3.1099,
+      "step": 7060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.211421836997661e-07,
+      "loss": 2.8612,
+      "step": 7061
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.201721461948107e-07,
+      "loss": 3.0767,
+      "step": 7062
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.192028166518865e-07,
+      "loss": 3.026,
+      "step": 7063
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.182341952276788e-07,
+      "loss": 2.8953,
+      "step": 7064
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.172662820787551e-07,
+      "loss": 3.0414,
+      "step": 7065
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.162990773615718e-07,
+      "loss": 2.9999,
+      "step": 7066
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.153325812324684e-07,
+      "loss": 2.8227,
+      "step": 7067
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.143667938476716e-07,
+      "loss": 2.9658,
+      "step": 7068
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.134017153632937e-07,
+      "loss": 2.9973,
+      "step": 7069
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.124373459353289e-07,
+      "loss": 2.9419,
+      "step": 7070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.114736857196624e-07,
+      "loss": 3.2229,
+      "step": 7071
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.105107348720584e-07,
+      "loss": 2.941,
+      "step": 7072
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.095484935481715e-07,
+      "loss": 2.8586,
+      "step": 7073
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.085869619035406e-07,
+      "loss": 2.8302,
+      "step": 7074
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.076261400935874e-07,
+      "loss": 3.0127,
+      "step": 7075
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.066660282736215e-07,
+      "loss": 2.8871,
+      "step": 7076
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.057066265988365e-07,
+      "loss": 3.0527,
+      "step": 7077
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.047479352243135e-07,
+      "loss": 3.0602,
+      "step": 7078
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.037899543050135e-07,
+      "loss": 3.0326,
+      "step": 7079
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.028326839957877e-07,
+      "loss": 2.8986,
+      "step": 7080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.018761244513721e-07,
+      "loss": 3.0484,
+      "step": 7081
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.009202758263838e-07,
+      "loss": 2.8435,
+      "step": 7082
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.999651382753297e-07,
+      "loss": 3.1106,
+      "step": 7083
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.990107119525978e-07,
+      "loss": 2.9257,
+      "step": 7084
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.980569970124639e-07,
+      "loss": 2.848,
+      "step": 7085
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.971039936090883e-07,
+      "loss": 2.8256,
+      "step": 7086
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.961517018965169e-07,
+      "loss": 2.7801,
+      "step": 7087
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.952001220286768e-07,
+      "loss": 2.834,
+      "step": 7088
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.942492541593847e-07,
+      "loss": 2.9697,
+      "step": 7089
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.932990984423415e-07,
+      "loss": 3.1411,
+      "step": 7090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.923496550311286e-07,
+      "loss": 3.0799,
+      "step": 7091
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.914009240792185e-07,
+      "loss": 3.0675,
+      "step": 7092
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.904529057399633e-07,
+      "loss": 3.1204,
+      "step": 7093
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.895056001666027e-07,
+      "loss": 3.1205,
+      "step": 7094
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.885590075122616e-07,
+      "loss": 2.9975,
+      "step": 7095
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.876131279299474e-07,
+      "loss": 2.9113,
+      "step": 7096
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.866679615725557e-07,
+      "loss": 2.8997,
+      "step": 7097
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.857235085928614e-07,
+      "loss": 3.0315,
+      "step": 7098
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.847797691435302e-07,
+      "loss": 2.9971,
+      "step": 7099
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.838367433771081e-07,
+      "loss": 2.9688,
+      "step": 7100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.828944314460267e-07,
+      "loss": 2.9388,
+      "step": 7101
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.819528335026048e-07,
+      "loss": 2.953,
+      "step": 7102
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.810119496990413e-07,
+      "loss": 2.9763,
+      "step": 7103
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.800717801874245e-07,
+      "loss": 2.7489,
+      "step": 7104
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.791323251197212e-07,
+      "loss": 3.0869,
+      "step": 7105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.781935846477909e-07,
+      "loss": 2.9693,
+      "step": 7106
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.772555589233697e-07,
+      "loss": 2.8737,
+      "step": 7107
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.763182480980828e-07,
+      "loss": 2.9495,
+      "step": 7108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.753816523234396e-07,
+      "loss": 2.8315,
+      "step": 7109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.744457717508301e-07,
+      "loss": 2.8494,
+      "step": 7110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.735106065315344e-07,
+      "loss": 2.8937,
+      "step": 7111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.725761568167115e-07,
+      "loss": 3.0682,
+      "step": 7112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.716424227574086e-07,
+      "loss": 3.0416,
+      "step": 7113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.707094045045564e-07,
+      "loss": 2.9024,
+      "step": 7114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.697771022089671e-07,
+      "loss": 2.91,
+      "step": 7115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.68845516021343e-07,
+      "loss": 2.8763,
+      "step": 7116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.679146460922636e-07,
+      "loss": 2.8736,
+      "step": 7117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.66984492572199e-07,
+      "loss": 3.0205,
+      "step": 7118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.660550556114986e-07,
+      "loss": 2.958,
+      "step": 7119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.651263353603987e-07,
+      "loss": 3.0244,
+      "step": 7120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.641983319690198e-07,
+      "loss": 3.0055,
+      "step": 7121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.632710455873642e-07,
+      "loss": 2.9281,
+      "step": 7122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.623444763653219e-07,
+      "loss": 2.8873,
+      "step": 7123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.614186244526626e-07,
+      "loss": 2.8576,
+      "step": 7124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.604934899990439e-07,
+      "loss": 2.9372,
+      "step": 7125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.595690731540048e-07,
+      "loss": 2.9123,
+      "step": 7126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.586453740669712e-07,
+      "loss": 2.9102,
+      "step": 7127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.57722392887251e-07,
+      "loss": 3.053,
+      "step": 7128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.568001297640347e-07,
+      "loss": 2.9517,
+      "step": 7129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.558785848463999e-07,
+      "loss": 2.8918,
+      "step": 7130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.549577582833048e-07,
+      "loss": 2.9478,
+      "step": 7131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.540376502235945e-07,
+      "loss": 2.8668,
+      "step": 7132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.531182608159973e-07,
+      "loss": 2.9689,
+      "step": 7133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.521995902091232e-07,
+      "loss": 2.7544,
+      "step": 7134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.512816385514675e-07,
+      "loss": 3.0207,
+      "step": 7135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.503644059914104e-07,
+      "loss": 2.9172,
+      "step": 7136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.494478926772151e-07,
+      "loss": 2.914,
+      "step": 7137
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.485320987570264e-07,
+      "loss": 2.9707,
+      "step": 7138
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.476170243788753e-07,
+      "loss": 2.9317,
+      "step": 7139
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.467026696906774e-07,
+      "loss": 2.9313,
+      "step": 7140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.457890348402284e-07,
+      "loss": 3.0241,
+      "step": 7141
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.448761199752106e-07,
+      "loss": 2.9005,
+      "step": 7142
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.439639252431884e-07,
+      "loss": 2.9882,
+      "step": 7143
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.430524507916101e-07,
+      "loss": 2.972,
+      "step": 7144
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.421416967678084e-07,
+      "loss": 2.9953,
+      "step": 7145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.41231663318999e-07,
+      "loss": 2.965,
+      "step": 7146
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.403223505922822e-07,
+      "loss": 2.9867,
+      "step": 7147
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.394137587346381e-07,
+      "loss": 2.902,
+      "step": 7148
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.385058878929355e-07,
+      "loss": 2.8038,
+      "step": 7149
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.375987382139219e-07,
+      "loss": 2.9884,
+      "step": 7150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.366923098442311e-07,
+      "loss": 2.8274,
+      "step": 7151
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.357866029303805e-07,
+      "loss": 2.7758,
+      "step": 7152
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.348816176187682e-07,
+      "loss": 2.941,
+      "step": 7153
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.339773540556787e-07,
+      "loss": 3.0025,
+      "step": 7154
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.330738123872776e-07,
+      "loss": 2.9861,
+      "step": 7155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.321709927596164e-07,
+      "loss": 2.9081,
+      "step": 7156
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.312688953186268e-07,
+      "loss": 2.9403,
+      "step": 7157
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.303675202101249e-07,
+      "loss": 3.0757,
+      "step": 7158
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.29466867579812e-07,
+      "loss": 2.7152,
+      "step": 7159
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.285669375732694e-07,
+      "loss": 3.0088,
+      "step": 7160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.27667730335964e-07,
+      "loss": 3.0176,
+      "step": 7161
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.267692460132445e-07,
+      "loss": 2.8069,
+      "step": 7162
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.258714847503432e-07,
+      "loss": 2.8133,
+      "step": 7163
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.249744466923756e-07,
+      "loss": 3.0741,
+      "step": 7164
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.240781319843408e-07,
+      "loss": 3.018,
+      "step": 7165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.231825407711216e-07,
+      "loss": 2.871,
+      "step": 7166
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.222876731974801e-07,
+      "loss": 2.8523,
+      "step": 7167
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.213935294080663e-07,
+      "loss": 3.0956,
+      "step": 7168
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.205001095474088e-07,
+      "loss": 2.9567,
+      "step": 7169
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.196074137599233e-07,
+      "loss": 2.8734,
+      "step": 7170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.18715442189906e-07,
+      "loss": 3.0104,
+      "step": 7171
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.178241949815355e-07,
+      "loss": 2.9621,
+      "step": 7172
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.169336722788764e-07,
+      "loss": 2.9586,
+      "step": 7173
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.160438742258706e-07,
+      "loss": 2.9897,
+      "step": 7174
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.15154800966351e-07,
+      "loss": 2.924,
+      "step": 7175
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.142664526440256e-07,
+      "loss": 2.8281,
+      "step": 7176
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.133788294024889e-07,
+      "loss": 2.8461,
+      "step": 7177
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.12491931385219e-07,
+      "loss": 3.0103,
+      "step": 7178
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.116057587355733e-07,
+      "loss": 2.9822,
+      "step": 7179
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.107203115967963e-07,
+      "loss": 2.9212,
+      "step": 7180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.098355901120111e-07,
+      "loss": 2.9258,
+      "step": 7181
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.089515944242263e-07,
+      "loss": 2.8375,
+      "step": 7182
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.08068324676333e-07,
+      "loss": 2.9722,
+      "step": 7183
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.071857810111019e-07,
+      "loss": 2.8652,
+      "step": 7184
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.063039635711925e-07,
+      "loss": 2.9852,
+      "step": 7185
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.054228724991395e-07,
+      "loss": 3.0649,
+      "step": 7186
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.045425079373667e-07,
+      "loss": 3.025,
+      "step": 7187
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.036628700281748e-07,
+      "loss": 2.999,
+      "step": 7188
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.027839589137512e-07,
+      "loss": 3.0106,
+      "step": 7189
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.019057747361655e-07,
+      "loss": 3.0616,
+      "step": 7190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.010283176373665e-07,
+      "loss": 3.0251,
+      "step": 7191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.001515877591895e-07,
+      "loss": 3.0233,
+      "step": 7192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.992755852433484e-07,
+      "loss": 2.9114,
+      "step": 7193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.984003102314444e-07,
+      "loss": 2.9813,
+      "step": 7194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.97525762864956e-07,
+      "loss": 2.999,
+      "step": 7195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.966519432852473e-07,
+      "loss": 2.9434,
+      "step": 7196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.957788516335644e-07,
+      "loss": 2.9082,
+      "step": 7197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.949064880510335e-07,
+      "loss": 2.9384,
+      "step": 7198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.94034852678667e-07,
+      "loss": 2.8808,
+      "step": 7199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.93163945657355e-07,
+      "loss": 3.0841,
+      "step": 7200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.922937671278733e-07,
+      "loss": 2.9443,
+      "step": 7201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.914243172308797e-07,
+      "loss": 3.0787,
+      "step": 7202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.905555961069108e-07,
+      "loss": 3.0474,
+      "step": 7203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.896876038963922e-07,
+      "loss": 3.0455,
+      "step": 7204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.888203407396236e-07,
+      "loss": 3.0409,
+      "step": 7205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.879538067767937e-07,
+      "loss": 3.0017,
+      "step": 7206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.870880021479674e-07,
+      "loss": 2.7945,
+      "step": 7207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.862229269930962e-07,
+      "loss": 2.9172,
+      "step": 7208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.853585814520135e-07,
+      "loss": 2.7397,
+      "step": 7209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.844949656644305e-07,
+      "loss": 2.9264,
+      "step": 7210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.836320797699467e-07,
+      "loss": 2.8643,
+      "step": 7211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.827699239080369e-07,
+      "loss": 3.0498,
+      "step": 7212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.819084982180632e-07,
+      "loss": 2.8545,
+      "step": 7213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.810478028392674e-07,
+      "loss": 2.8844,
+      "step": 7214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.801878379107739e-07,
+      "loss": 2.9972,
+      "step": 7215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.793286035715894e-07,
+      "loss": 3.0615,
+      "step": 7216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.784700999605995e-07,
+      "loss": 2.8506,
+      "step": 7217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.776123272165772e-07,
+      "loss": 2.8464,
+      "step": 7218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7675528547817094e-07,
+      "loss": 2.8467,
+      "step": 7219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.75898974883916e-07,
+      "loss": 2.9733,
+      "step": 7220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7504339557222833e-07,
+      "loss": 2.9784,
+      "step": 7221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7418854768140345e-07,
+      "loss": 2.9017,
+      "step": 7222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.733344313496213e-07,
+      "loss": 3.0439,
+      "step": 7223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.724810467149421e-07,
+      "loss": 2.9712,
+      "step": 7224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.7162839391530935e-07,
+      "loss": 3.111,
+      "step": 7225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.707764730885456e-07,
+      "loss": 2.8658,
+      "step": 7226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6992528437235676e-07,
+      "loss": 3.046,
+      "step": 7227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6907482790433223e-07,
+      "loss": 2.9668,
+      "step": 7228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.682251038219382e-07,
+      "loss": 2.9067,
+      "step": 7229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.673761122625281e-07,
+      "loss": 2.7338,
+      "step": 7230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6652785336333217e-07,
+      "loss": 3.0458,
+      "step": 7231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.656803272614646e-07,
+      "loss": 2.8895,
+      "step": 7232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.648335340939214e-07,
+      "loss": 2.9257,
+      "step": 7233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6398747399758025e-07,
+      "loss": 3.0111,
+      "step": 7234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.631421471091979e-07,
+      "loss": 3.005,
+      "step": 7235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6229755356541505e-07,
+      "loss": 3.0134,
+      "step": 7236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.614536935027536e-07,
+      "loss": 3.0176,
+      "step": 7237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6061056705761495e-07,
+      "loss": 2.9705,
+      "step": 7238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5976817436628573e-07,
+      "loss": 2.9311,
+      "step": 7239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5892651556492863e-07,
+      "loss": 3.0494,
+      "step": 7240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5808559078959203e-07,
+      "loss": 2.8097,
+      "step": 7241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.57245400176205e-07,
+      "loss": 2.9073,
+      "step": 7242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5640594386057615e-07,
+      "loss": 3.0522,
+      "step": 7243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.555672219783974e-07,
+      "loss": 3.0096,
+      "step": 7244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5472923466524043e-07,
+      "loss": 2.991,
+      "step": 7245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.538919820565596e-07,
+      "loss": 2.8635,
+      "step": 7246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5305546428768777e-07,
+      "loss": 3.1142,
+      "step": 7247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.522196814938423e-07,
+      "loss": 3.0012,
+      "step": 7248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.513846338101213e-07,
+      "loss": 3.0097,
+      "step": 7249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.505503213715012e-07,
+      "loss": 2.8089,
+      "step": 7250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4971674431284307e-07,
+      "loss": 2.9875,
+      "step": 7251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4888390276888574e-07,
+      "loss": 3.1384,
+      "step": 7252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.480517968742537e-07,
+      "loss": 3.0258,
+      "step": 7253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.472204267634472e-07,
+      "loss": 3.0466,
+      "step": 7254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4638979257085103e-07,
+      "loss": 2.9906,
+      "step": 7255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4555989443073157e-07,
+      "loss": 2.955,
+      "step": 7256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.447307324772332e-07,
+      "loss": 2.9759,
+      "step": 7257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4390230684438374e-07,
+      "loss": 2.913,
+      "step": 7258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4307461766608996e-07,
+      "loss": 2.9561,
+      "step": 7259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4224766507614205e-07,
+      "loss": 2.9476,
+      "step": 7260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.414214492082097e-07,
+      "loss": 3.0081,
+      "step": 7261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4059597019584175e-07,
+      "loss": 2.9504,
+      "step": 7262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.397712281724736e-07,
+      "loss": 2.9911,
+      "step": 7263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.389472232714154e-07,
+      "loss": 2.9641,
+      "step": 7264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3812395562586095e-07,
+      "loss": 2.7888,
+      "step": 7265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3730142536888397e-07,
+      "loss": 2.8904,
+      "step": 7266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.364796326334403e-07,
+      "loss": 2.791,
+      "step": 7267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3565857755236585e-07,
+      "loss": 2.8973,
+      "step": 7268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3483826025837674e-07,
+      "loss": 2.9832,
+      "step": 7269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3401868088407073e-07,
+      "loss": 2.8897,
+      "step": 7270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3319983956192466e-07,
+      "loss": 2.795,
+      "step": 7271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3238173642429824e-07,
+      "loss": 3.0461,
+      "step": 7272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.315643716034307e-07,
+      "loss": 2.8857,
+      "step": 7273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3074774523144237e-07,
+      "loss": 3.1465,
+      "step": 7274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2993185744033495e-07,
+      "loss": 2.9009,
+      "step": 7275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2911670836198794e-07,
+      "loss": 3.0303,
+      "step": 7276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2830229812816483e-07,
+      "loss": 2.8752,
+      "step": 7277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2748862687050584e-07,
+      "loss": 2.8319,
+      "step": 7278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2667569472053636e-07,
+      "loss": 2.8977,
+      "step": 7279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2586350180965965e-07,
+      "loss": 2.9153,
+      "step": 7280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.250520482691578e-07,
+      "loss": 3.0217,
+      "step": 7281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.242413342301982e-07,
+      "loss": 3.0239,
+      "step": 7282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2343135982382433e-07,
+      "loss": 2.8852,
+      "step": 7283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.226221251809626e-07,
+      "loss": 2.8205,
+      "step": 7284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.218136304324172e-07,
+      "loss": 2.9299,
+      "step": 7285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.2100587570887584e-07,
+      "loss": 2.7424,
+      "step": 7286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.201988611409058e-07,
+      "loss": 3.0382,
+      "step": 7287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1939258685895214e-07,
+      "loss": 2.9515,
+      "step": 7288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1858705299334443e-07,
+      "loss": 2.9566,
+      "step": 7289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1778225967428864e-07,
+      "loss": 2.9103,
+      "step": 7290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.169782070318734e-07,
+      "loss": 3.0418,
+      "step": 7291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1617489519606644e-07,
+      "loss": 3.0924,
+      "step": 7292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.153723242967178e-07,
+      "loss": 3.0138,
+      "step": 7293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.145704944635559e-07,
+      "loss": 2.8981,
+      "step": 7294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1376940582618873e-07,
+      "loss": 3.0109,
+      "step": 7295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1296905851410664e-07,
+      "loss": 2.9809,
+      "step": 7296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1216945265667716e-07,
+      "loss": 3.0412,
+      "step": 7297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.1137058838315146e-07,
+      "loss": 2.9756,
+      "step": 7298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.105724658226595e-07,
+      "loss": 3.0148,
+      "step": 7299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0977508510420874e-07,
+      "loss": 3.098,
+      "step": 7300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0897844635669095e-07,
+      "loss": 2.7718,
+      "step": 7301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.081825497088759e-07,
+      "loss": 2.8912,
+      "step": 7302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.073873952894136e-07,
+      "loss": 2.8938,
+      "step": 7303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0659298322683284e-07,
+      "loss": 2.9018,
+      "step": 7304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.057993136495447e-07,
+      "loss": 2.9036,
+      "step": 7305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0500638668583947e-07,
+      "loss": 2.8246,
+      "step": 7306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.042142024638851e-07,
+      "loss": 2.9749,
+      "step": 7307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.034227611117342e-07,
+      "loss": 2.8691,
+      "step": 7308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.026320627573138e-07,
+      "loss": 2.8116,
+      "step": 7309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0184210752843566e-07,
+      "loss": 3.0194,
+      "step": 7310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.010528955527876e-07,
+      "loss": 2.8438,
+      "step": 7311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.002644269579409e-07,
+      "loss": 2.9238,
+      "step": 7312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9947670187134414e-07,
+      "loss": 2.9893,
+      "step": 7313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.986897204203255e-07,
+      "loss": 3.0658,
+      "step": 7314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.97903482732096e-07,
+      "loss": 3.0263,
+      "step": 7315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9711798893374123e-07,
+      "loss": 2.9448,
+      "step": 7316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.963332391522318e-07,
+      "loss": 2.9128,
+      "step": 7317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9554923351441564e-07,
+      "loss": 3.0318,
+      "step": 7318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9476597214701986e-07,
+      "loss": 3.0189,
+      "step": 7319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9398345517665303e-07,
+      "loss": 2.9451,
+      "step": 7320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.932016827298002e-07,
+      "loss": 2.8907,
+      "step": 7321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.924206549328313e-07,
+      "loss": 3.0003,
+      "step": 7322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.91640371911991e-07,
+      "loss": 3.0242,
+      "step": 7323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.90860833793405e-07,
+      "loss": 2.9547,
+      "step": 7324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9008204070308144e-07,
+      "loss": 2.9061,
+      "step": 7325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.893039927669029e-07,
+      "loss": 3.0122,
+      "step": 7326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8852669011063606e-07,
+      "loss": 2.8513,
+      "step": 7327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.877501328599242e-07,
+      "loss": 2.7986,
+      "step": 7328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.86974321140291e-07,
+      "loss": 2.8846,
+      "step": 7329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8619925507714093e-07,
+      "loss": 2.9625,
+      "step": 7330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8542493479575614e-07,
+      "loss": 2.942,
+      "step": 7331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.846513604213004e-07,
+      "loss": 3.0524,
+      "step": 7332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8387853207881296e-07,
+      "loss": 2.9956,
+      "step": 7333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8310644989321687e-07,
+      "loss": 2.888,
+      "step": 7334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.823351139893117e-07,
+      "loss": 3.0382,
+      "step": 7335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.815645244917776e-07,
+      "loss": 2.8411,
+      "step": 7336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.807946815251745e-07,
+      "loss": 2.9896,
+      "step": 7337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.800255852139395e-07,
+      "loss": 2.993,
+      "step": 7338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.792572356823926e-07,
+      "loss": 2.8867,
+      "step": 7339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7848963305472774e-07,
+      "loss": 3.0498,
+      "step": 7340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.777227774550252e-07,
+      "loss": 2.9566,
+      "step": 7341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7695666900723795e-07,
+      "loss": 2.8474,
+      "step": 7342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.76191307835202e-07,
+      "loss": 2.9367,
+      "step": 7343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.754266940626322e-07,
+      "loss": 3.0824,
+      "step": 7344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.746628278131204e-07,
+      "loss": 3.0375,
+      "step": 7345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.738997092101404e-07,
+      "loss": 2.8542,
+      "step": 7346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.731373383770426e-07,
+      "loss": 2.9915,
+      "step": 7347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7237571543705775e-07,
+      "loss": 2.9805,
+      "step": 7348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.71614840513298e-07,
+      "loss": 2.9767,
+      "step": 7349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7085471372874825e-07,
+      "loss": 2.9734,
+      "step": 7350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.700953352062814e-07,
+      "loss": 2.8996,
+      "step": 7351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6933670506864086e-07,
+      "loss": 2.9687,
+      "step": 7352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6857882343845517e-07,
+      "loss": 2.9419,
+      "step": 7353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.67821690438227e-07,
+      "loss": 2.9449,
+      "step": 7354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.670653061903423e-07,
+      "loss": 3.0145,
+      "step": 7355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6630967081706436e-07,
+      "loss": 2.8916,
+      "step": 7356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.655547844405338e-07,
+      "loss": 2.8394,
+      "step": 7357
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.648006471827731e-07,
+      "loss": 2.9359,
+      "step": 7358
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.6404725916568075e-07,
+      "loss": 3.0243,
+      "step": 7359
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.6329462051103604e-07,
+      "loss": 3.0206,
+      "step": 7360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.625427313404967e-07,
+      "loss": 2.9865,
+      "step": 7361
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.617915917755993e-07,
+      "loss": 2.9203,
+      "step": 7362
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.6104120193776015e-07,
+      "loss": 3.0517,
+      "step": 7363
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.602915619482722e-07,
+      "loss": 3.0737,
+      "step": 7364
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.595426719283085e-07,
+      "loss": 3.0151,
+      "step": 7365
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.5879453199892054e-07,
+      "loss": 3.0221,
+      "step": 7366
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.580471422810394e-07,
+      "loss": 2.9165,
+      "step": 7367
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.573005028954746e-07,
+      "loss": 2.9226,
+      "step": 7368
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.565546139629128e-07,
+      "loss": 2.9344,
+      "step": 7369
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.5580947560392153e-07,
+      "loss": 2.7813,
+      "step": 7370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.550650879389456e-07,
+      "loss": 2.884,
+      "step": 7371
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.5432145108831026e-07,
+      "loss": 2.9597,
+      "step": 7372
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.535785651722162e-07,
+      "loss": 3.1322,
+      "step": 7373
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.528364303107468e-07,
+      "loss": 2.9075,
+      "step": 7374
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.520950466238593e-07,
+      "loss": 2.9078,
+      "step": 7375
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.5135441423139347e-07,
+      "loss": 2.9135,
+      "step": 7376
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.5061453325306747e-07,
+      "loss": 2.993,
+      "step": 7377
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4987540380847395e-07,
+      "loss": 2.9303,
+      "step": 7378
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4913702601708897e-07,
+      "loss": 2.9392,
+      "step": 7379
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4839939999826434e-07,
+      "loss": 2.9041,
+      "step": 7380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.476625258712324e-07,
+      "loss": 2.8536,
+      "step": 7381
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.469264037551007e-07,
+      "loss": 2.9353,
+      "step": 7382
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.461910337688573e-07,
+      "loss": 3.0122,
+      "step": 7383
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4545641603137e-07,
+      "loss": 2.913,
+      "step": 7384
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4472255066138214e-07,
+      "loss": 2.8836,
+      "step": 7385
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4398943777751824e-07,
+      "loss": 2.9173,
+      "step": 7386
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.43257077498278e-07,
+      "loss": 2.9639,
+      "step": 7387
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4252546994204216e-07,
+      "loss": 3.0,
+      "step": 7388
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4179461522706847e-07,
+      "loss": 2.8684,
+      "step": 7389
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4106451347149403e-07,
+      "loss": 2.8414,
+      "step": 7390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4033516479333394e-07,
+      "loss": 3.0038,
+      "step": 7391
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3960656931048e-07,
+      "loss": 3.0193,
+      "step": 7392
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3887872714070516e-07,
+      "loss": 2.9977,
+      "step": 7393
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.38151638401657e-07,
+      "loss": 3.0542,
+      "step": 7394
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3742530321086375e-07,
+      "loss": 3.0492,
+      "step": 7395
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.366997216857326e-07,
+      "loss": 2.9538,
+      "step": 7396
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3597489394354645e-07,
+      "loss": 2.9459,
+      "step": 7397
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3525082010146814e-07,
+      "loss": 3.0407,
+      "step": 7398
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.345275002765369e-07,
+      "loss": 3.0311,
+      "step": 7399
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.338049345856731e-07,
+      "loss": 2.958,
+      "step": 7400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.330831231456716e-07,
+      "loss": 3.0885,
+      "step": 7401
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3236206607320855e-07,
+      "loss": 2.8805,
+      "step": 7402
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.316417634848362e-07,
+      "loss": 2.8339,
+      "step": 7403
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3092221549698487e-07,
+      "loss": 3.0122,
+      "step": 7404
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3020342222596425e-07,
+      "loss": 3.0856,
+      "step": 7405
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2948538378796024e-07,
+      "loss": 3.1635,
+      "step": 7406
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.287681002990378e-07,
+      "loss": 3.1039,
+      "step": 7407
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.280515718751409e-07,
+      "loss": 2.9776,
+      "step": 7408
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.27335798632088e-07,
+      "loss": 2.8814,
+      "step": 7409
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2662078068558033e-07,
+      "loss": 3.0322,
+      "step": 7410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.259065181511928e-07,
+      "loss": 3.0112,
+      "step": 7411
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.251930111443813e-07,
+      "loss": 2.9536,
+      "step": 7412
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2448025978047637e-07,
+      "loss": 3.0216,
+      "step": 7413
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.237682641746892e-07,
+      "loss": 3.0095,
+      "step": 7414
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2305702444210883e-07,
+      "loss": 3.0012,
+      "step": 7415
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.223465406976989e-07,
+      "loss": 3.0065,
+      "step": 7416
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2163681305630524e-07,
+      "loss": 2.87,
+      "step": 7417
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.209278416326478e-07,
+      "loss": 2.9317,
+      "step": 7418
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.20219626541326e-07,
+      "loss": 2.8741,
+      "step": 7419
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1951216789681774e-07,
+      "loss": 3.0431,
+      "step": 7420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.188054658134765e-07,
+      "loss": 3.1133,
+      "step": 7421
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.180995204055365e-07,
+      "loss": 3.1459,
+      "step": 7422
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.173943317871053e-07,
+      "loss": 2.8772,
+      "step": 7423
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.166899000721735e-07,
+      "loss": 2.9678,
+      "step": 7424
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.159862253746038e-07,
+      "loss": 2.9179,
+      "step": 7425
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.152833078081408e-07,
+      "loss": 3.0934,
+      "step": 7426
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.145811474864052e-07,
+      "loss": 2.8905,
+      "step": 7427
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1387974452289305e-07,
+      "loss": 2.9645,
+      "step": 7428
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.131790990309841e-07,
+      "loss": 3.0705,
+      "step": 7429
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1247921112392886e-07,
+      "loss": 2.8616,
+      "step": 7430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1178008091485967e-07,
+      "loss": 2.9018,
+      "step": 7431
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1108170851678386e-07,
+      "loss": 3.0089,
+      "step": 7432
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1038409404258784e-07,
+      "loss": 3.0035,
+      "step": 7433
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.096872376050364e-07,
+      "loss": 2.9389,
+      "step": 7434
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0899113931676774e-07,
+      "loss": 3.082,
+      "step": 7435
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0829579929030295e-07,
+      "loss": 3.0378,
+      "step": 7436
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.076012176380361e-07,
+      "loss": 2.9809,
+      "step": 7437
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.069073944722406e-07,
+      "loss": 2.8518,
+      "step": 7438
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0621432990506796e-07,
+      "loss": 2.9866,
+      "step": 7439
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0552202404854513e-07,
+      "loss": 3.1202,
+      "step": 7440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0483047701457934e-07,
+      "loss": 2.9631,
+      "step": 7441
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.041396889149512e-07,
+      "loss": 2.9268,
+      "step": 7442
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.034496598613218e-07,
+      "loss": 3.0474,
+      "step": 7443
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.0276038996522816e-07,
+      "loss": 3.085,
+      "step": 7444
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.0207187933808446e-07,
+      "loss": 2.8534,
+      "step": 7445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.0138412809118443e-07,
+      "loss": 3.0059,
+      "step": 7446
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.0069713633569475e-07,
+      "loss": 2.8093,
+      "step": 7447
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.0001090418266323e-07,
+      "loss": 3.0612,
+      "step": 7448
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.993254317430133e-07,
+      "loss": 2.9618,
+      "step": 7449
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9864071912754643e-07,
+      "loss": 2.7527,
+      "step": 7450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.979567664469396e-07,
+      "loss": 2.883,
+      "step": 7451
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9727357381174773e-07,
+      "loss": 2.9499,
+      "step": 7452
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9659114133240463e-07,
+      "loss": 2.9515,
+      "step": 7453
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.959094691192177e-07,
+      "loss": 2.8544,
+      "step": 7454
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.952285572823754e-07,
+      "loss": 3.0636,
+      "step": 7455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9454840593194036e-07,
+      "loss": 2.9483,
+      "step": 7456
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9386901517785283e-07,
+      "loss": 3.0477,
+      "step": 7457
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9319038512993116e-07,
+      "loss": 2.8337,
+      "step": 7458
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.925125158978698e-07,
+      "loss": 2.9189,
+      "step": 7459
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.918354075912422e-07,
+      "loss": 2.92,
+      "step": 7460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.911590603194947e-07,
+      "loss": 2.9463,
+      "step": 7461
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9048347419195545e-07,
+      "loss": 2.9368,
+      "step": 7462
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.898086493178248e-07,
+      "loss": 3.062,
+      "step": 7463
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8913458580618394e-07,
+      "loss": 3.0106,
+      "step": 7464
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8846128376599e-07,
+      "loss": 2.7963,
+      "step": 7465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.87788743306075e-07,
+      "loss": 3.0067,
+      "step": 7466
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8711696453515017e-07,
+      "loss": 2.953,
+      "step": 7467
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.864459475618031e-07,
+      "loss": 3.1152,
+      "step": 7468
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.857756924944982e-07,
+      "loss": 2.9624,
+      "step": 7469
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8510619944157545e-07,
+      "loss": 3.0469,
+      "step": 7470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.844374685112539e-07,
+      "loss": 2.9523,
+      "step": 7471
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8376949981162807e-07,
+      "loss": 3.0293,
+      "step": 7472
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.831022934506683e-07,
+      "loss": 2.708,
+      "step": 7473
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.824358495362245e-07,
+      "loss": 2.9607,
+      "step": 7474
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.817701681760199e-07,
+      "loss": 2.858,
+      "step": 7475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8110524947765726e-07,
+      "loss": 3.1676,
+      "step": 7476
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.804410935486152e-07,
+      "loss": 2.8538,
+      "step": 7477
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.797777004962482e-07,
+      "loss": 3.0418,
+      "step": 7478
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7911507042778953e-07,
+      "loss": 2.8347,
+      "step": 7479
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7845320345034564e-07,
+      "loss": 2.8669,
+      "step": 7480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.777920996709038e-07,
+      "loss": 3.0778,
+      "step": 7481
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7713175919632463e-07,
+      "loss": 2.8841,
+      "step": 7482
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7647218213334603e-07,
+      "loss": 2.9929,
+      "step": 7483
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.75813368588585e-07,
+      "loss": 2.9734,
+      "step": 7484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.751553186685313e-07,
+      "loss": 2.9787,
+      "step": 7485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.744980324795538e-07,
+      "loss": 3.0472,
+      "step": 7486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.738415101278963e-07,
+      "loss": 3.0408,
+      "step": 7487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731857517196823e-07,
+      "loss": 2.7878,
+      "step": 7488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.725307573609076e-07,
+      "loss": 3.1451,
+      "step": 7489
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.718765271574475e-07,
+      "loss": 2.9992,
+      "step": 7490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7122306121505293e-07,
+      "loss": 3.0368,
+      "step": 7491
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7057035963934994e-07,
+      "loss": 2.8282,
+      "step": 7492
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.699184225358431e-07,
+      "loss": 2.8205,
+      "step": 7493
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6926725000991193e-07,
+      "loss": 2.8455,
+      "step": 7494
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.686168421668134e-07,
+      "loss": 2.8284,
+      "step": 7495
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.679671991116811e-07,
+      "loss": 2.9596,
+      "step": 7496
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.673183209495217e-07,
+      "loss": 3.0987,
+      "step": 7497
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.666702077852246e-07,
+      "loss": 3.061,
+      "step": 7498
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6602285972354925e-07,
+      "loss": 3.0249,
+      "step": 7499
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.653762768691348e-07,
+      "loss": 3.0035,
+      "step": 7500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6473045932649545e-07,
+      "loss": 2.7877,
+      "step": 7501
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6408540720002206e-07,
+      "loss": 2.9214,
+      "step": 7502
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.634411205939835e-07,
+      "loss": 2.8691,
+      "step": 7503
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6279759961252036e-07,
+      "loss": 2.9451,
+      "step": 7504
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.62154844359655e-07,
+      "loss": 3.0577,
+      "step": 7505
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6151285493928145e-07,
+      "loss": 3.015,
+      "step": 7506
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.608716314551729e-07,
+      "loss": 3.013,
+      "step": 7507
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.602311740109775e-07,
+      "loss": 2.9179,
+      "step": 7508
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5959148271021917e-07,
+      "loss": 2.9482,
+      "step": 7509
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.589525576563001e-07,
+      "loss": 2.9695,
+      "step": 7510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5831439895249554e-07,
+      "loss": 2.9973,
+      "step": 7511
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.576770067019596e-07,
+      "loss": 2.9387,
+      "step": 7512
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.570403810077204e-07,
+      "loss": 2.8478,
+      "step": 7513
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5640452197268296e-07,
+      "loss": 2.8115,
+      "step": 7514
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.557694296996299e-07,
+      "loss": 3.0238,
+      "step": 7515
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5513510429121704e-07,
+      "loss": 3.2292,
+      "step": 7516
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5450154584997835e-07,
+      "loss": 2.7459,
+      "step": 7517
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.538687544783225e-07,
+      "loss": 3.0245,
+      "step": 7518
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.532367302785366e-07,
+      "loss": 2.9082,
+      "step": 7519
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5260547335278e-07,
+      "loss": 2.835,
+      "step": 7520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.51974983803091e-07,
+      "loss": 2.9892,
+      "step": 7521
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.513452617313827e-07,
+      "loss": 2.9458,
+      "step": 7522
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5071630723944415e-07,
+      "loss": 2.835,
+      "step": 7523
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.500881204289418e-07,
+      "loss": 2.8876,
+      "step": 7524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4946070140141443e-07,
+      "loss": 2.9134,
+      "step": 7525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.488340502582803e-07,
+      "loss": 3.0157,
+      "step": 7526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.482081671008324e-07,
+      "loss": 2.9258,
+      "step": 7527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4758305203024025e-07,
+      "loss": 3.1627,
+      "step": 7528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4695870514754583e-07,
+      "loss": 2.8066,
+      "step": 7529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4633512655367174e-07,
+      "loss": 2.8597,
+      "step": 7530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4571231634941404e-07,
+      "loss": 3.0331,
+      "step": 7531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4509027463544334e-07,
+      "loss": 3.0668,
+      "step": 7532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4446900151230913e-07,
+      "loss": 3.0756,
+      "step": 7533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4384849708043335e-07,
+      "loss": 2.9613,
+      "step": 7534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.4322876144011577e-07,
+      "loss": 2.9602,
+      "step": 7535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.426097946915318e-07,
+      "loss": 2.8995,
+      "step": 7536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.419915969347314e-07,
+      "loss": 2.8914,
+      "step": 7537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.41374168269643e-07,
+      "loss": 2.9676,
+      "step": 7538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.407575087960662e-07,
+      "loss": 2.9913,
+      "step": 7539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.401416186136807e-07,
+      "loss": 2.7725,
+      "step": 7540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.395264978220385e-07,
+      "loss": 2.9564,
+      "step": 7541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3891214652056905e-07,
+      "loss": 2.8774,
+      "step": 7542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3829856480857783e-07,
+      "loss": 3.01,
+      "step": 7543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3768575278524385e-07,
+      "loss": 2.9137,
+      "step": 7544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3707371054962457e-07,
+      "loss": 2.8907,
+      "step": 7545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3646243820064918e-07,
+      "loss": 3.009,
+      "step": 7546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3585193583712697e-07,
+      "loss": 2.8441,
+      "step": 7547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3524220355773896e-07,
+      "loss": 2.9792,
+      "step": 7548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3463324146104361e-07,
+      "loss": 2.9792,
+      "step": 7549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3402504964547545e-07,
+      "loss": 3.0429,
+      "step": 7550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.33417628209342e-07,
+      "loss": 3.0082,
+      "step": 7551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3281097725082912e-07,
+      "loss": 2.8913,
+      "step": 7552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3220509686799565e-07,
+      "loss": 3.0393,
+      "step": 7553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3159998715877708e-07,
+      "loss": 2.8697,
+      "step": 7554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3099564822098518e-07,
+      "loss": 3.0859,
+      "step": 7555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.303920801523052e-07,
+      "loss": 3.0619,
+      "step": 7556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2978928305030014e-07,
+      "loss": 3.0626,
+      "step": 7557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.291872570124054e-07,
+      "loss": 2.9075,
+      "step": 7558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.285860021359354e-07,
+      "loss": 3.0819,
+      "step": 7559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.279855185180757e-07,
+      "loss": 2.8704,
+      "step": 7560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2738580625589035e-07,
+      "loss": 3.0045,
+      "step": 7561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2678686544631846e-07,
+      "loss": 3.0279,
+      "step": 7562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2618869618617202e-07,
+      "loss": 2.9529,
+      "step": 7563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.25591298572142e-07,
+      "loss": 3.0761,
+      "step": 7564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2499467270079068e-07,
+      "loss": 3.0094,
+      "step": 7565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2439881866855972e-07,
+      "loss": 3.0202,
+      "step": 7566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2380373657176212e-07,
+      "loss": 2.9773,
+      "step": 7567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2320942650658818e-07,
+      "loss": 2.8999,
+      "step": 7568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2261588856910432e-07,
+      "loss": 3.1424,
+      "step": 7569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.220231228552494e-07,
+      "loss": 3.0739,
+      "step": 7570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2143112946084012e-07,
+      "loss": 2.8469,
+      "step": 7571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.20839908481566e-07,
+      "loss": 3.0586,
+      "step": 7572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2024946001299397e-07,
+      "loss": 2.8152,
+      "step": 7573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1965978415056432e-07,
+      "loss": 2.9283,
+      "step": 7574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1907088098959251e-07,
+      "loss": 2.8864,
+      "step": 7575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1848275062527237e-07,
+      "loss": 2.8449,
+      "step": 7576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.178953931526673e-07,
+      "loss": 3.0861,
+      "step": 7577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1730880866672078e-07,
+      "loss": 2.8846,
+      "step": 7578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.16722997262247e-07,
+      "loss": 3.0203,
+      "step": 7579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1613795903393908e-07,
+      "loss": 2.9687,
+      "step": 7580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1555369407636362e-07,
+      "loss": 2.8995,
+      "step": 7581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1497020248396116e-07,
+      "loss": 2.9709,
+      "step": 7582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1438748435104905e-07,
+      "loss": 2.9868,
+      "step": 7583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.138055397718175e-07,
+      "loss": 2.8691,
+      "step": 7584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.132243688403335e-07,
+      "loss": 2.8678,
+      "step": 7585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1264397165053852e-07,
+      "loss": 2.9401,
+      "step": 7586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1206434829624922e-07,
+      "loss": 2.9629,
+      "step": 7587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1148549887115676e-07,
+      "loss": 3.0257,
+      "step": 7588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.109074234688263e-07,
+      "loss": 2.8982,
+      "step": 7589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1033012218270032e-07,
+      "loss": 2.8973,
+      "step": 7590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0975359510609305e-07,
+      "loss": 3.0939,
+      "step": 7591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0917784233219606e-07,
+      "loss": 2.912,
+      "step": 7592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0860286395407548e-07,
+      "loss": 2.8132,
+      "step": 7593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0802866006467025e-07,
+      "loss": 2.9849,
+      "step": 7594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0745523075679674e-07,
+      "loss": 2.946,
+      "step": 7595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.068825761231441e-07,
+      "loss": 3.0314,
+      "step": 7596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.063106962562783e-07,
+      "loss": 2.728,
+      "step": 7597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0573959124863761e-07,
+      "loss": 2.8803,
+      "step": 7598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0516926119253654e-07,
+      "loss": 3.0373,
+      "step": 7599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0459970618016466e-07,
+      "loss": 3.0139,
+      "step": 7600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0403092630358556e-07,
+      "loss": 3.006,
+      "step": 7601
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0346292165473736e-07,
+      "loss": 2.8944,
+      "step": 7602
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0289569232543327e-07,
+      "loss": 3.0652,
+      "step": 7603
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0232923840736107e-07,
+      "loss": 2.918,
+      "step": 7604
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0176355999208363e-07,
+      "loss": 3.048,
+      "step": 7605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0119865717103726e-07,
+      "loss": 2.8426,
+      "step": 7606
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0063453003553556e-07,
+      "loss": 2.9346,
+      "step": 7607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.0007117867676228e-07,
+      "loss": 3.0081,
+      "step": 7608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.995086031857807e-07,
+      "loss": 2.9571,
+      "step": 7609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.989468036535247e-07,
+      "loss": 2.8567,
+      "step": 7610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.98385780170805e-07,
+      "loss": 2.9861,
+      "step": 7611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9782553282830686e-07,
+      "loss": 2.9913,
+      "step": 7612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9726606171658892e-07,
+      "loss": 3.0639,
+      "step": 7613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9670736692608438e-07,
+      "loss": 3.0978,
+      "step": 7614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.961494485471027e-07,
+      "loss": 2.9554,
+      "step": 7615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9559230666982677e-07,
+      "loss": 2.8105,
+      "step": 7616
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9503594138431227e-07,
+      "loss": 2.9875,
+      "step": 7617
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9448035278049227e-07,
+      "loss": 3.0388,
+      "step": 7618
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9392554094817274e-07,
+      "loss": 3.0263,
+      "step": 7619
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9337150597703358e-07,
+      "loss": 2.9757,
+      "step": 7620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9281824795663152e-07,
+      "loss": 2.8554,
+      "step": 7621
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9226576697639386e-07,
+      "loss": 2.9316,
+      "step": 7622
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9171406312562534e-07,
+      "loss": 3.0376,
+      "step": 7623
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.911631364935046e-07,
+      "loss": 3.0448,
+      "step": 7624
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.906129871690843e-07,
+      "loss": 2.9378,
+      "step": 7625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9006361524129158e-07,
+      "loss": 3.0043,
+      "step": 7626
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.895150207989266e-07,
+      "loss": 2.9686,
+      "step": 7627
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8896720393066614e-07,
+      "loss": 2.9022,
+      "step": 7628
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.884201647250594e-07,
+      "loss": 3.0717,
+      "step": 7629
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8787390327053123e-07,
+      "loss": 3.0821,
+      "step": 7630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8732841965537984e-07,
+      "loss": 2.9639,
+      "step": 7631
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8678371396777805e-07,
+      "loss": 2.9314,
+      "step": 7632
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.862397862957732e-07,
+      "loss": 2.8699,
+      "step": 7633
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8569663672728555e-07,
+      "loss": 3.0884,
+      "step": 7634
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8515426535011204e-07,
+      "loss": 3.0815,
+      "step": 7635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8461267225192148e-07,
+      "loss": 2.9499,
+      "step": 7636
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8407185752025824e-07,
+      "loss": 2.914,
+      "step": 7637
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8353182124254076e-07,
+      "loss": 2.9492,
+      "step": 7638
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8299256350606032e-07,
+      "loss": 2.8883,
+      "step": 7639
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.824540843979844e-07,
+      "loss": 3.1672,
+      "step": 7640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8191638400535228e-07,
+      "loss": 2.9396,
+      "step": 7641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8137946241507998e-07,
+      "loss": 2.9156,
+      "step": 7642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8084331971395586e-07,
+      "loss": 2.9949,
+      "step": 7643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.803079559886417e-07,
+      "loss": 2.7863,
+      "step": 7644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7977337132567718e-07,
+      "loss": 3.0935,
+      "step": 7645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7923956581147095e-07,
+      "loss": 2.9369,
+      "step": 7646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7870653953230955e-07,
+      "loss": 2.8793,
+      "step": 7647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7817429257435125e-07,
+      "loss": 3.0647,
+      "step": 7648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7764282502362896e-07,
+      "loss": 3.0385,
+      "step": 7649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7711213696605169e-07,
+      "loss": 2.9169,
+      "step": 7650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7658222848739858e-07,
+      "loss": 2.9956,
+      "step": 7651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7605309967332672e-07,
+      "loss": 2.8808,
+      "step": 7652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.755247506093638e-07,
+      "loss": 3.0269,
+      "step": 7653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7499718138091316e-07,
+      "loss": 2.9105,
+      "step": 7654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7447039207325267e-07,
+      "loss": 2.9491,
+      "step": 7655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7394438277153258e-07,
+      "loss": 2.8957,
+      "step": 7656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7341915356077933e-07,
+      "loss": 2.9651,
+      "step": 7657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.728947045258894e-07,
+      "loss": 2.8713,
+      "step": 7658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7237103575163782e-07,
+      "loss": 2.9725,
+      "step": 7659
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7184814732266907e-07,
+      "loss": 2.9621,
+      "step": 7660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.71326039323505e-07,
+      "loss": 2.8527,
+      "step": 7661
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7080471183854031e-07,
+      "loss": 2.8803,
+      "step": 7662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7028416495204148e-07,
+      "loss": 2.899,
+      "step": 7663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.697643987481523e-07,
+      "loss": 3.013,
+      "step": 7664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.692454133108873e-07,
+      "loss": 2.86,
+      "step": 7665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6872720872413706e-07,
+      "loss": 2.9434,
+      "step": 7666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6820978507166408e-07,
+      "loss": 3.0173,
+      "step": 7667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6769314243710643e-07,
+      "loss": 3.0049,
+      "step": 7668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.67177280903974e-07,
+      "loss": 2.955,
+      "step": 7669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6666220055565173e-07,
+      "loss": 3.017,
+      "step": 7670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6614790147539862e-07,
+      "loss": 2.8568,
+      "step": 7671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6563438374634588e-07,
+      "loss": 2.9788,
+      "step": 7672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6512164745149993e-07,
+      "loss": 2.9426,
+      "step": 7673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.646096926737395e-07,
+      "loss": 3.0024,
+      "step": 7674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6409851949581947e-07,
+      "loss": 2.9478,
+      "step": 7675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.635881280003643e-07,
+      "loss": 2.9103,
+      "step": 7676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6307851826987632e-07,
+      "loss": 3.0628,
+      "step": 7677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6256969038672853e-07,
+      "loss": 3.0653,
+      "step": 7678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6206164443316907e-07,
+      "loss": 3.008,
+      "step": 7679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6155438049132e-07,
+      "loss": 3.0533,
+      "step": 7680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6104789864317406e-07,
+      "loss": 2.945,
+      "step": 7681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6054219897060186e-07,
+      "loss": 3.0505,
+      "step": 7682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6003728155534416e-07,
+      "loss": 2.8393,
+      "step": 7683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5953314647901786e-07,
+      "loss": 3.0068,
+      "step": 7684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5902979382311113e-07,
+      "loss": 2.8899,
+      "step": 7685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5852722366898666e-07,
+      "loss": 2.9691,
+      "step": 7686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5802543609788168e-07,
+      "loss": 2.9378,
+      "step": 7687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5752443119090466e-07,
+      "loss": 2.8294,
+      "step": 7688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5702420902903916e-07,
+      "loss": 3.0412,
+      "step": 7689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.565247696931421e-07,
+      "loss": 2.9697,
+      "step": 7690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5602611326394347e-07,
+      "loss": 2.9857,
+      "step": 7691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5552823982204757e-07,
+      "loss": 3.1606,
+      "step": 7692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5503114944792896e-07,
+      "loss": 2.9019,
+      "step": 7693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5453484222194114e-07,
+      "loss": 2.8619,
+      "step": 7694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5403931822430662e-07,
+      "loss": 2.9398,
+      "step": 7695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5354457753512187e-07,
+      "loss": 2.8974,
+      "step": 7696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5305062023435957e-07,
+      "loss": 3.0928,
+      "step": 7697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5255744640186143e-07,
+      "loss": 2.8317,
+      "step": 7698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5206505611734701e-07,
+      "loss": 2.8634,
+      "step": 7699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5157344946040486e-07,
+      "loss": 2.9368,
+      "step": 7700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5108262651050032e-07,
+      "loss": 2.8786,
+      "step": 7701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5059258734697102e-07,
+      "loss": 2.7759,
+      "step": 7702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.501033320490264e-07,
+      "loss": 2.9962,
+      "step": 7703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.496148606957515e-07,
+      "loss": 2.9273,
+      "step": 7704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4912717336610315e-07,
+      "loss": 2.8728,
+      "step": 7705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.486402701389128e-07,
+      "loss": 2.9175,
+      "step": 7706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4815415109288244e-07,
+      "loss": 2.947,
+      "step": 7707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4766881630658981e-07,
+      "loss": 3.0474,
+      "step": 7708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4718426585848656e-07,
+      "loss": 2.9214,
+      "step": 7709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4670049982689394e-07,
+      "loss": 2.878,
+      "step": 7710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4621751829000996e-07,
+      "loss": 2.7492,
+      "step": 7711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4573532132590384e-07,
+      "loss": 2.8389,
+      "step": 7712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.452539090125199e-07,
+      "loss": 2.9256,
+      "step": 7713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4477328142767256e-07,
+      "loss": 2.974,
+      "step": 7714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4429343864905242e-07,
+      "loss": 2.9251,
+      "step": 7715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4381438075422193e-07,
+      "loss": 2.8694,
+      "step": 7716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.433361078206158e-07,
+      "loss": 2.9245,
+      "step": 7717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.428586199255444e-07,
+      "loss": 2.7143,
+      "step": 7718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.423819171461882e-07,
+      "loss": 2.9688,
+      "step": 7719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4190599955960227e-07,
+      "loss": 2.8941,
+      "step": 7720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.414308672427156e-07,
+      "loss": 2.82,
+      "step": 7721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4095652027232786e-07,
+      "loss": 2.7417,
+      "step": 7722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4048295872511498e-07,
+      "loss": 2.8936,
+      "step": 7723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.400101826776229e-07,
+      "loss": 2.9655,
+      "step": 7724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3953819220627219e-07,
+      "loss": 2.9397,
+      "step": 7725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3906698738735626e-07,
+      "loss": 2.9728,
+      "step": 7726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3859656829704082e-07,
+      "loss": 2.9382,
+      "step": 7727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381269350113662e-07,
+      "loss": 3.0224,
+      "step": 7728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3765808760624332e-07,
+      "loss": 3.0125,
+      "step": 7729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.371900261574588e-07,
+      "loss": 2.9131,
+      "step": 7730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3672275074066932e-07,
+      "loss": 2.8491,
+      "step": 7731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3625626143140668e-07,
+      "loss": 2.9168,
+      "step": 7732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.357905583050756e-07,
+      "loss": 2.8903,
+      "step": 7733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3532564143695192e-07,
+      "loss": 2.8413,
+      "step": 7734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3486151090218612e-07,
+      "loss": 2.8215,
+      "step": 7735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3439816677580097e-07,
+      "loss": 3.0187,
+      "step": 7736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.339356091326921e-07,
+      "loss": 2.997,
+      "step": 7737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.334738380476275e-07,
+      "loss": 2.9515,
+      "step": 7738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3301285359524907e-07,
+      "loss": 2.9323,
+      "step": 7739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.325526558500717e-07,
+      "loss": 2.7349,
+      "step": 7740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3209324488648033e-07,
+      "loss": 3.0886,
+      "step": 7741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3163462077873724e-07,
+      "loss": 2.9387,
+      "step": 7742
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3117678360097363e-07,
+      "loss": 3.0712,
+      "step": 7743
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.307197334271959e-07,
+      "loss": 3.0736,
+      "step": 7744
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3026347033128106e-07,
+      "loss": 2.9075,
+      "step": 7745
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2980799438698177e-07,
+      "loss": 3.0112,
+      "step": 7746
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2935330566792082e-07,
+      "loss": 2.9849,
+      "step": 7747
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2889940424759496e-07,
+      "loss": 2.8793,
+      "step": 7748
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2844629019937438e-07,
+      "loss": 2.772,
+      "step": 7749
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.279939635964994e-07,
+      "loss": 2.9984,
+      "step": 7750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.27542424512086e-07,
+      "loss": 3.0254,
+      "step": 7751
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2709167301912128e-07,
+      "loss": 2.9979,
+      "step": 7752
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2664170919046537e-07,
+      "loss": 2.9653,
+      "step": 7753
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.261925330988517e-07,
+      "loss": 2.9358,
+      "step": 7754
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2574414481688502e-07,
+      "loss": 3.1343,
+      "step": 7755
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2529654441704398e-07,
+      "loss": 2.9546,
+      "step": 7756
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2484973197167905e-07,
+      "loss": 2.9459,
+      "step": 7757
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2440370755301357e-07,
+      "loss": 2.8885,
+      "step": 7758
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.239584712331443e-07,
+      "loss": 3.1688,
+      "step": 7759
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.235140230840387e-07,
+      "loss": 3.0533,
+      "step": 7760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.230703631775393e-07,
+      "loss": 2.8142,
+      "step": 7761
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2262749158535926e-07,
+      "loss": 3.1115,
+      "step": 7762
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2218540837908576e-07,
+      "loss": 3.0406,
+      "step": 7763
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2174411363017668e-07,
+      "loss": 2.99,
+      "step": 7764
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2130360740996439e-07,
+      "loss": 2.9665,
+      "step": 7765
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2086388978965301e-07,
+      "loss": 3.0192,
+      "step": 7766
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2042496084031853e-07,
+      "loss": 2.8916,
+      "step": 7767
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1998682063291078e-07,
+      "loss": 2.898,
+      "step": 7768
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.195494692382504e-07,
+      "loss": 2.9166,
+      "step": 7769
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1911290672703246e-07,
+      "loss": 2.8647,
+      "step": 7770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1867713316982388e-07,
+      "loss": 3.0055,
+      "step": 7771
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1824214863706273e-07,
+      "loss": 2.8961,
+      "step": 7772
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1780795319906168e-07,
+      "loss": 2.8473,
+      "step": 7773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1737454692600347e-07,
+      "loss": 2.8506,
+      "step": 7774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1694192988794595e-07,
+      "loss": 2.9832,
+      "step": 7775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.165101021548165e-07,
+      "loss": 2.9661,
+      "step": 7776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1607906379641708e-07,
+      "loss": 2.8753,
+      "step": 7777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1564881488242252e-07,
+      "loss": 2.9132,
+      "step": 7778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1521935548237718e-07,
+      "loss": 3.0058,
+      "step": 7779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1479068566570051e-07,
+      "loss": 2.8925,
+      "step": 7780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.143628055016821e-07,
+      "loss": 2.9804,
+      "step": 7781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.139357150594872e-07,
+      "loss": 2.9133,
+      "step": 7782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1350941440814944e-07,
+      "loss": 3.1744,
+      "step": 7783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1308390361657817e-07,
+      "loss": 2.9179,
+      "step": 7784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.126591827535528e-07,
+      "loss": 2.924,
+      "step": 7785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1223525188772566e-07,
+      "loss": 2.9709,
+      "step": 7786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1181211108762302e-07,
+      "loss": 2.976,
+      "step": 7787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1138976042163963e-07,
+      "loss": 2.7929,
+      "step": 7788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1096819995804698e-07,
+      "loss": 3.0257,
+      "step": 7789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1054742976498556e-07,
+      "loss": 3.0817,
+      "step": 7790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1012744991046931e-07,
+      "loss": 2.8818,
+      "step": 7791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0970826046238558e-07,
+      "loss": 2.9391,
+      "step": 7792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0928986148849186e-07,
+      "loss": 3.0163,
+      "step": 7793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0887225305641901e-07,
+      "loss": 3.0294,
+      "step": 7794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.084554352336692e-07,
+      "loss": 2.8401,
+      "step": 7795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0803940808761848e-07,
+      "loss": 3.0375,
+      "step": 7796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.076241716855142e-07,
+      "loss": 2.8972,
+      "step": 7797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0720972609447488e-07,
+      "loss": 2.8924,
+      "step": 7798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0679607138149306e-07,
+      "loss": 2.9839,
+      "step": 7799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0638320761343079e-07,
+      "loss": 2.9644,
+      "step": 7800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0597113485702693e-07,
+      "loss": 2.9436,
+      "step": 7801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0555985317888651e-07,
+      "loss": 2.921,
+      "step": 7802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0514936264549191e-07,
+      "loss": 2.7745,
+      "step": 7803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0473966332319452e-07,
+      "loss": 2.9679,
+      "step": 7804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0433075527821857e-07,
+      "loss": 2.9052,
+      "step": 7805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0392263857666175e-07,
+      "loss": 3.0854,
+      "step": 7806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0351531328449071e-07,
+      "loss": 3.0639,
+      "step": 7807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0310877946754782e-07,
+      "loss": 2.9677,
+      "step": 7808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0270303719154551e-07,
+      "loss": 2.9854,
+      "step": 7809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0229808652206741e-07,
+      "loss": 3.0541,
+      "step": 7810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0189392752457173e-07,
+      "loss": 3.0399,
+      "step": 7811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0149056026438675e-07,
+      "loss": 2.8308,
+      "step": 7812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0108798480671367e-07,
+      "loss": 2.8587,
+      "step": 7813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0068620121662487e-07,
+      "loss": 2.841,
+      "step": 7814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0028520955906618e-07,
+      "loss": 2.9919,
+      "step": 7815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.988500989885353e-08,
+      "loss": 2.8522,
+      "step": 7816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.948560230067573e-08,
+      "loss": 3.017,
+      "step": 7817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.908698682909501e-08,
+      "loss": 3.0801,
+      "step": 7818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.868916354854263e-08,
+      "loss": 3.012,
+      "step": 7819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.829213252332382e-08,
+      "loss": 2.9557,
+      "step": 7820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.78958938176161e-08,
+      "loss": 2.969,
+      "step": 7821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.750044749546717e-08,
+      "loss": 2.8577,
+      "step": 7822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.710579362079753e-08,
+      "loss": 3.0303,
+      "step": 7823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.671193225740006e-08,
+      "loss": 2.8408,
+      "step": 7824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.631886346893937e-08,
+      "loss": 2.9903,
+      "step": 7825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.592658731895077e-08,
+      "loss": 2.851,
+      "step": 7826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.553510387084352e-08,
+      "loss": 2.9698,
+      "step": 7827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.514441318789647e-08,
+      "loss": 2.9678,
+      "step": 7828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.47545153332624e-08,
+      "loss": 2.9365,
+      "step": 7829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.436541036996427e-08,
+      "loss": 2.9153,
+      "step": 7830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.397709836089786e-08,
+      "loss": 2.8368,
+      "step": 7831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.358957936883073e-08,
+      "loss": 2.8656,
+      "step": 7832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.320285345640168e-08,
+      "loss": 2.8734,
+      "step": 7833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.281692068612235e-08,
+      "loss": 3.1609,
+      "step": 7834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.243178112037454e-08,
+      "loss": 2.9793,
+      "step": 7835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.204743482141343e-08,
+      "loss": 3.0069,
+      "step": 7836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.166388185136544e-08,
+      "loss": 2.9448,
+      "step": 7837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.128112227222763e-08,
+      "loss": 2.7615,
+      "step": 7838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.089915614587164e-08,
+      "loss": 2.941,
+      "step": 7839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.051798353403696e-08,
+      "loss": 3.0168,
+      "step": 7840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.013760449833875e-08,
+      "loss": 2.7751,
+      "step": 7841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.975801910026061e-08,
+      "loss": 3.0592,
+      "step": 7842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.937922740116012e-08,
+      "loss": 3.0238,
+      "step": 7843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.900122946226607e-08,
+      "loss": 2.9821,
+      "step": 7844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.862402534467795e-08,
+      "loss": 2.8357,
+      "step": 7845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.824761510936808e-08,
+      "loss": 2.8059,
+      "step": 7846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.787199881717889e-08,
+      "loss": 2.9321,
+      "step": 7847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.749717652882683e-08,
+      "loss": 2.9631,
+      "step": 7848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.712314830489787e-08,
+      "loss": 2.9888,
+      "step": 7849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.674991420585144e-08,
+      "loss": 2.97,
+      "step": 7850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.637747429201759e-08,
+      "loss": 2.8826,
+      "step": 7851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.600582862359652e-08,
+      "loss": 3.0041,
+      "step": 7852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.56349772606635e-08,
+      "loss": 3.0238,
+      "step": 7853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.52649202631628e-08,
+      "loss": 2.9226,
+      "step": 7854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.489565769091046e-08,
+      "loss": 2.9858,
+      "step": 7855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.452718960359541e-08,
+      "loss": 2.7763,
+      "step": 7856
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.415951606077666e-08,
+      "loss": 2.868,
+      "step": 7857
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.379263712188667e-08,
+      "loss": 2.8884,
+      "step": 7858
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.342655284622636e-08,
+      "loss": 3.0033,
+      "step": 7859
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.306126329297282e-08,
+      "loss": 3.0241,
+      "step": 7860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.269676852116936e-08,
+      "loss": 2.7876,
+      "step": 7861
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.2333068589735e-08,
+      "loss": 3.0032,
+      "step": 7862
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.19701635574588e-08,
+      "loss": 3.0757,
+      "step": 7863
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.160805348300105e-08,
+      "loss": 3.027,
+      "step": 7864
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.124673842489327e-08,
+      "loss": 2.9397,
+      "step": 7865
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.088621844153932e-08,
+      "loss": 2.8005,
+      "step": 7866
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.052649359121423e-08,
+      "loss": 2.8932,
+      "step": 7867
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 8.016756393206427e-08,
+      "loss": 2.9659,
+      "step": 7868
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.980942952210746e-08,
+      "loss": 2.9961,
+      "step": 7869
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.945209041923363e-08,
+      "loss": 3.0347,
+      "step": 7870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.909554668120322e-08,
+      "loss": 2.9382,
+      "step": 7871
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.873979836564905e-08,
+      "loss": 2.9147,
+      "step": 7872
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.838484553007342e-08,
+      "loss": 3.0258,
+      "step": 7873
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.803068823185266e-08,
+      "loss": 3.0756,
+      "step": 7874
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.767732652823323e-08,
+      "loss": 2.9973,
+      "step": 7875
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.73247604763322e-08,
+      "loss": 2.8279,
+      "step": 7876
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.697299013314064e-08,
+      "loss": 2.9639,
+      "step": 7877
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.662201555551696e-08,
+      "loss": 3.1158,
+      "step": 7878
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.627183680019468e-08,
+      "loss": 2.9547,
+      "step": 7879
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.592245392377684e-08,
+      "loss": 3.0627,
+      "step": 7880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.557386698273827e-08,
+      "loss": 2.9778,
+      "step": 7881
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.522607603342557e-08,
+      "loss": 2.8865,
+      "step": 7882
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.487908113205489e-08,
+      "loss": 2.7552,
+      "step": 7883
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.453288233471689e-08,
+      "loss": 2.9093,
+      "step": 7884
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.418747969736962e-08,
+      "loss": 2.9368,
+      "step": 7885
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.384287327584616e-08,
+      "loss": 2.8524,
+      "step": 7886
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.349906312584864e-08,
+      "loss": 2.9046,
+      "step": 7887
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.315604930295039e-08,
+      "loss": 2.9349,
+      "step": 7888
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.28138318625976e-08,
+      "loss": 2.8068,
+      "step": 7889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.247241086010603e-08,
+      "loss": 2.9496,
+      "step": 7890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.213178635066487e-08,
+      "loss": 3.0118,
+      "step": 7891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.179195838933172e-08,
+      "loss": 2.9593,
+      "step": 7892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.145292703103712e-08,
+      "loss": 3.054,
+      "step": 7893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.111469233058332e-08,
+      "loss": 3.0324,
+      "step": 7894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.077725434264271e-08,
+      "loss": 2.9024,
+      "step": 7895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.044061312175943e-08,
+      "loss": 2.9394,
+      "step": 7896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.010476872234828e-08,
+      "loss": 2.921,
+      "step": 7897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.97697211986953e-08,
+      "loss": 2.7169,
+      "step": 7898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.943547060495937e-08,
+      "loss": 2.9172,
+      "step": 7899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.910201699516838e-08,
+      "loss": 2.9179,
+      "step": 7900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.876936042322257e-08,
+      "loss": 2.8416,
+      "step": 7901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.843750094289281e-08,
+      "loss": 3.0086,
+      "step": 7902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.810643860782174e-08,
+      "loss": 3.0983,
+      "step": 7903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.77761734715221e-08,
+      "loss": 3.0845,
+      "step": 7904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.744670558737898e-08,
+      "loss": 2.9839,
+      "step": 7905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.711803500864866e-08,
+      "loss": 2.7429,
+      "step": 7906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.679016178845643e-08,
+      "loss": 2.9357,
+      "step": 7907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.646308597980156e-08,
+      "loss": 3.0006,
+      "step": 7908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.61368076355523e-08,
+      "loss": 2.9386,
+      "step": 7909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.581132680844926e-08,
+      "loss": 2.8327,
+      "step": 7910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.548664355110313e-08,
+      "loss": 2.9904,
+      "step": 7911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.51627579159969e-08,
+      "loss": 3.0019,
+      "step": 7912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.483966995548374e-08,
+      "loss": 2.8934,
+      "step": 7913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.451737972178739e-08,
+      "loss": 2.8889,
+      "step": 7914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.419588726700455e-08,
+      "loss": 2.8102,
+      "step": 7915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.387519264310027e-08,
+      "loss": 3.0325,
+      "step": 7916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.355529590191311e-08,
+      "loss": 2.9811,
+      "step": 7917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.323619709515117e-08,
+      "loss": 3.0529,
+      "step": 7918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.291789627439482e-08,
+      "loss": 2.8888,
+      "step": 7919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.260039349109403e-08,
+      "loss": 3.011,
+      "step": 7920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.228368879657055e-08,
+      "loss": 3.1564,
+      "step": 7921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.196778224201783e-08,
+      "loss": 2.9423,
+      "step": 7922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.165267387849782e-08,
+      "loss": 2.9774,
+      "step": 7923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.133836375694591e-08,
+      "loss": 2.8705,
+      "step": 7924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.102485192816865e-08,
+      "loss": 2.8672,
+      "step": 7925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.071213844284107e-08,
+      "loss": 2.8298,
+      "step": 7926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.040022335151163e-08,
+      "loss": 2.8765,
+      "step": 7927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.008910670459778e-08,
+      "loss": 3.0398,
+      "step": 7928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.977878855239039e-08,
+      "loss": 2.9264,
+      "step": 7929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.946926894504823e-08,
+      "loss": 2.9041,
+      "step": 7930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.916054793260295e-08,
+      "loss": 3.037,
+      "step": 7931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8852625564957966e-08,
+      "loss": 3.04,
+      "step": 7932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.854550189188457e-08,
+      "loss": 2.9459,
+      "step": 7933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.82391769630275e-08,
+      "loss": 2.827,
+      "step": 7934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.793365082790159e-08,
+      "loss": 2.9604,
+      "step": 7935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.762892353589178e-08,
+      "loss": 2.9953,
+      "step": 7936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.732499513625645e-08,
+      "loss": 2.9035,
+      "step": 7937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.702186567812073e-08,
+      "loss": 3.0407,
+      "step": 7938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6719535210484896e-08,
+      "loss": 2.9418,
+      "step": 7939
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.6418003782217067e-08,
+      "loss": 2.9406,
+      "step": 7940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.6117271442057696e-08,
+      "loss": 2.7824,
+      "step": 7941
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.581733823861679e-08,
+      "loss": 3.0058,
+      "step": 7942
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.5518204220377234e-08,
+      "loss": 2.9666,
+      "step": 7943
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.521986943569091e-08,
+      "loss": 2.8623,
+      "step": 7944
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.4922333932780346e-08,
+      "loss": 2.9669,
+      "step": 7945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.4625597759740964e-08,
+      "loss": 3.0314,
+      "step": 7946
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.432966096453662e-08,
+      "loss": 3.0337,
+      "step": 7947
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.403452359500405e-08,
+      "loss": 2.9961,
+      "step": 7948
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.374018569884898e-08,
+      "loss": 2.7684,
+      "step": 7949
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.344664732364835e-08,
+      "loss": 2.9515,
+      "step": 7950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.315390851685087e-08,
+      "loss": 2.8208,
+      "step": 7951
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.286196932577481e-08,
+      "loss": 2.7741,
+      "step": 7952
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.257082979761019e-08,
+      "loss": 2.9317,
+      "step": 7953
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2280489979416596e-08,
+      "loss": 2.8848,
+      "step": 7954
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.199094991812481e-08,
+      "loss": 2.9498,
+      "step": 7955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1702209660537964e-08,
+      "loss": 2.8539,
+      "step": 7956
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.141426925332649e-08,
+      "loss": 3.0674,
+      "step": 7957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1127128743034824e-08,
+      "loss": 2.9254,
+      "step": 7958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.084078817607641e-08,
+      "loss": 2.9417,
+      "step": 7959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.055524759873642e-08,
+      "loss": 2.9661,
+      "step": 7960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.027050705716907e-08,
+      "loss": 2.895,
+      "step": 7961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9986566597400867e-08,
+      "loss": 3.0588,
+      "step": 7962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.970342626532787e-08,
+      "loss": 2.8841,
+      "step": 7963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9421086106717917e-08,
+      "loss": 2.8619,
+      "step": 7964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.913954616720895e-08,
+      "loss": 3.058,
+      "step": 7965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.885880649230845e-08,
+      "loss": 3.0407,
+      "step": 7966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.857886712739679e-08,
+      "loss": 3.0107,
+      "step": 7967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8299728117722765e-08,
+      "loss": 2.9336,
+      "step": 7968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8021389508408066e-08,
+      "loss": 3.0379,
+      "step": 7969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7743851344442814e-08,
+      "loss": 2.8261,
+      "step": 7970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7467113670688904e-08,
+      "loss": 2.9865,
+      "step": 7971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7191176531879435e-08,
+      "loss": 3.0043,
+      "step": 7972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6916039972615955e-08,
+      "loss": 2.8282,
+      "step": 7973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6641704037372894e-08,
+      "loss": 2.9477,
+      "step": 7974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.636816877049366e-08,
+      "loss": 3.0099,
+      "step": 7975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6095434216192894e-08,
+      "loss": 2.9881,
+      "step": 7976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.582350041855699e-08,
+      "loss": 3.0371,
+      "step": 7977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.555236742154079e-08,
+      "loss": 3.0236,
+      "step": 7978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.52820352689709e-08,
+      "loss": 2.949,
+      "step": 7979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5012504004543486e-08,
+      "loss": 3.0396,
+      "step": 7980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.474377367182758e-08,
+      "loss": 2.893,
+      "step": 7981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.447584431425955e-08,
+      "loss": 2.9557,
+      "step": 7982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.420871597514864e-08,
+      "loss": 2.9988,
+      "step": 7983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.39423886976742e-08,
+      "loss": 2.982,
+      "step": 7984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.367686252488512e-08,
+      "loss": 2.9601,
+      "step": 7985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.341213749970152e-08,
+      "loss": 3.0176,
+      "step": 7986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.314821366491473e-08,
+      "loss": 3.0608,
+      "step": 7987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.288509106318506e-08,
+      "loss": 2.8845,
+      "step": 7988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.262276973704349e-08,
+      "loss": 2.9259,
+      "step": 7989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2361249728893326e-08,
+      "loss": 3.0277,
+      "step": 7990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.210053108100687e-08,
+      "loss": 2.99,
+      "step": 7991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.184061383552651e-08,
+      "loss": 2.8868,
+      "step": 7992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.158149803446587e-08,
+      "loss": 2.9367,
+      "step": 7993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1323183719708114e-08,
+      "loss": 2.9429,
+      "step": 7994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.106567093300873e-08,
+      "loss": 3.07,
+      "step": 7995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0808959715992194e-08,
+      "loss": 2.9111,
+      "step": 7996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.055305011015309e-08,
+      "loss": 2.9667,
+      "step": 7997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.029794215685834e-08,
+      "loss": 2.8273,
+      "step": 7998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.004363589734217e-08,
+      "loss": 2.9377,
+      "step": 7999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9790131372712814e-08,
+      "loss": 3.0563,
+      "step": 8000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.953742862394527e-08,
+      "loss": 2.8164,
+      "step": 8001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9285527691888536e-08,
+      "loss": 2.9184,
+      "step": 8002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.903442861725948e-08,
+      "loss": 2.9885,
+      "step": 8003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8784131440645634e-08,
+      "loss": 3.0145,
+      "step": 8004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.853463620250686e-08,
+      "loss": 2.9052,
+      "step": 8005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.828594294317034e-08,
+      "loss": 3.0347,
+      "step": 8006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.803805170283614e-08,
+      "loss": 2.9458,
+      "step": 8007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.779096252157333e-08,
+      "loss": 2.9462,
+      "step": 8008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7544675439322163e-08,
+      "loss": 2.8187,
+      "step": 8009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.729919049589303e-08,
+      "loss": 3.0473,
+      "step": 8010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7054507730965864e-08,
+      "loss": 3.0776,
+      "step": 8011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.681062718409234e-08,
+      "loss": 2.8599,
+      "step": 8012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.656754889469261e-08,
+      "loss": 2.9717,
+      "step": 8013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6325272902059116e-08,
+      "loss": 3.0444,
+      "step": 8014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.608379924535332e-08,
+      "loss": 2.9045,
+      "step": 8015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5843127963607335e-08,
+      "loss": 3.1745,
+      "step": 8016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.560325909572449e-08,
+      "loss": 3.0643,
+      "step": 8017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5364192680475975e-08,
+      "loss": 2.9448,
+      "step": 8018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.512592875650644e-08,
+      "loss": 3.0319,
+      "step": 8019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.488846736232843e-08,
+      "loss": 2.7398,
+      "step": 8020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4651808536325106e-08,
+      "loss": 2.9913,
+      "step": 8021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.441595231675143e-08,
+      "loss": 2.8991,
+      "step": 8022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4180898741730805e-08,
+      "loss": 2.8519,
+      "step": 8023
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3946647849258384e-08,
+      "loss": 3.1309,
+      "step": 8024
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.371319967719777e-08,
+      "loss": 2.898,
+      "step": 8025
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.348055426328434e-08,
+      "loss": 2.7679,
+      "step": 8026
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.324871164512355e-08,
+      "loss": 2.8478,
+      "step": 8027
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3017671860191004e-08,
+      "loss": 2.9778,
+      "step": 8028
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.278743494583181e-08,
+      "loss": 2.9099,
+      "step": 8029
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.255800093926176e-08,
+      "loss": 2.8895,
+      "step": 8030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.232936987756674e-08,
+      "loss": 2.9219,
+      "step": 8031
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.210154179770386e-08,
+      "loss": 3.0778,
+      "step": 8032
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.187451673649866e-08,
+      "loss": 2.972,
+      "step": 8033
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1648294730649007e-08,
+      "loss": 2.9489,
+      "step": 8034
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.14228758167201e-08,
+      "loss": 2.9832,
+      "step": 8035
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.119826003115056e-08,
+      "loss": 2.9552,
+      "step": 8036
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.097444741024691e-08,
+      "loss": 2.9803,
+      "step": 8037
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.075143799018632e-08,
+      "loss": 2.928,
+      "step": 8038
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0529231807016615e-08,
+      "loss": 2.9974,
+      "step": 8039
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.030782889665629e-08,
+      "loss": 2.8931,
+      "step": 8040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.008722929489227e-08,
+      "loss": 2.9706,
+      "step": 8041
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.986743303738271e-08,
+      "loss": 2.79,
+      "step": 8042
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9648440159655844e-08,
+      "loss": 2.9473,
+      "step": 8043
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9430250697110584e-08,
+      "loss": 2.9448,
+      "step": 8044
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.921286468501483e-08,
+      "loss": 2.9403,
+      "step": 8045
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.899628215850714e-08,
+      "loss": 2.9991,
+      "step": 8046
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.878050315259673e-08,
+      "loss": 2.8871,
+      "step": 8047
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.856552770216181e-08,
+      "loss": 2.9059,
+      "step": 8048
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8351355841952365e-08,
+      "loss": 2.9875,
+      "step": 8049
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.813798760658626e-08,
+      "loss": 3.088,
+      "step": 8050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.792542303055368e-08,
+      "loss": 2.878,
+      "step": 8051
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7713662148212694e-08,
+      "loss": 2.8275,
+      "step": 8052
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.750270499379426e-08,
+      "loss": 2.8673,
+      "step": 8053
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.72925516013961e-08,
+      "loss": 2.9373,
+      "step": 8054
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.708320200498882e-08,
+      "loss": 3.0479,
+      "step": 8055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.687465623841201e-08,
+      "loss": 3.0418,
+      "step": 8056
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6666914335374826e-08,
+      "loss": 2.9243,
+      "step": 8057
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.645997632945707e-08,
+      "loss": 2.8562,
+      "step": 8058
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6253842254109207e-08,
+      "loss": 2.953,
+      "step": 8059
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.604851214265014e-08,
+      "loss": 3.011,
+      "step": 8060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5843986028269985e-08,
+      "loss": 2.9459,
+      "step": 8061
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5640263944028964e-08,
+      "loss": 3.0348,
+      "step": 8062
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5437345922856847e-08,
+      "loss": 2.8127,
+      "step": 8063
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5235231997553513e-08,
+      "loss": 2.9371,
+      "step": 8064
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5033922200788928e-08,
+      "loss": 2.8945,
+      "step": 8065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4833416565103184e-08,
+      "loss": 2.8903,
+      "step": 8066
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4633715122906465e-08,
+      "loss": 2.9323,
+      "step": 8067
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4434817906479058e-08,
+      "loss": 2.9185,
+      "step": 8068
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.42367249479708e-08,
+      "loss": 2.8828,
+      "step": 8069
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4039436279401085e-08,
+      "loss": 3.0338,
+      "step": 8070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.384295193266051e-08,
+      "loss": 3.1021,
+      "step": 8071
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.364727193950922e-08,
+      "loss": 2.9527,
+      "step": 8072
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.345239633157692e-08,
+      "loss": 2.9432,
+      "step": 8073
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.325832514036397e-08,
+      "loss": 3.0055,
+      "step": 8074
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.306505839724027e-08,
+      "loss": 2.9325,
+      "step": 8075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.287259613344528e-08,
+      "loss": 3.0065,
+      "step": 8076
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2680938380089114e-08,
+      "loss": 2.95,
+      "step": 8077
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2490085168151987e-08,
+      "loss": 2.9569,
+      "step": 8078
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.230003652848367e-08,
+      "loss": 2.9048,
+      "step": 8079
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2110792491803478e-08,
+      "loss": 2.9911,
+      "step": 8080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1922353088701386e-08,
+      "loss": 2.8836,
+      "step": 8081
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.173471834963692e-08,
+      "loss": 2.8384,
+      "step": 8082
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1547888304939146e-08,
+      "loss": 2.9745,
+      "step": 8083
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1361862984808913e-08,
+      "loss": 2.8714,
+      "step": 8084
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1176642419314386e-08,
+      "loss": 2.8996,
+      "step": 8085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0992226638395507e-08,
+      "loss": 3.0449,
+      "step": 8086
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.080861567186121e-08,
+      "loss": 2.931,
+      "step": 8087
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.062580954939053e-08,
+      "loss": 2.9842,
+      "step": 8088
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0443808300533162e-08,
+      "loss": 2.8343,
+      "step": 8089
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0262611954707245e-08,
+      "loss": 2.8218,
+      "step": 8090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.008222054120268e-08,
+      "loss": 2.9642,
+      "step": 8091
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9902634089177252e-08,
+      "loss": 2.803,
+      "step": 8092
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9723852627659968e-08,
+      "loss": 2.8111,
+      "step": 8093
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9545876185549372e-08,
+      "loss": 2.9614,
+      "step": 8094
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9368704791613568e-08,
+      "loss": 2.8232,
+      "step": 8095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9192338474491313e-08,
+      "loss": 3.1288,
+      "step": 8096
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9016777262690357e-08,
+      "loss": 2.9045,
+      "step": 8097
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8842021184588554e-08,
+      "loss": 2.9673,
+      "step": 8098
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8668070268433868e-08,
+      "loss": 2.9842,
+      "step": 8099
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.849492454234436e-08,
+      "loss": 2.9678,
+      "step": 8100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.832258403430709e-08,
+      "loss": 2.854,
+      "step": 8101
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.815104877217977e-08,
+      "loss": 2.9367,
+      "step": 8102
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.798031878368911e-08,
+      "loss": 2.9642,
+      "step": 8103
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7810394096433035e-08,
+      "loss": 3.0622,
+      "step": 8104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7641274737877357e-08,
+      "loss": 3.0111,
+      "step": 8105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7472960735359645e-08,
+      "loss": 2.9124,
+      "step": 8106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7305452116086474e-08,
+      "loss": 2.9012,
+      "step": 8107
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7138748907133407e-08,
+      "loss": 2.9434,
+      "step": 8108
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.697285113544722e-08,
+      "loss": 2.9682,
+      "step": 8109
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.680775882784369e-08,
+      "loss": 3.1579,
+      "step": 8110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.664347201100869e-08,
+      "loss": 2.8537,
+      "step": 8111
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.647999071149764e-08,
+      "loss": 2.863,
+      "step": 8112
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6317314955735518e-08,
+      "loss": 2.7859,
+      "step": 8113
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6155444770018512e-08,
+      "loss": 3.005,
+      "step": 8114
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5994380180510695e-08,
+      "loss": 2.9691,
+      "step": 8115
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5834121213247344e-08,
+      "loss": 2.8895,
+      "step": 8116
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5674667894132743e-08,
+      "loss": 2.9518,
+      "step": 8117
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.551602024894072e-08,
+      "loss": 2.9201,
+      "step": 8118
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.535817830331632e-08,
+      "loss": 2.9031,
+      "step": 8119
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5201142082772457e-08,
+      "loss": 2.9583,
+      "step": 8120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.504491161269328e-08,
+      "loss": 2.9561,
+      "step": 8121
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4889486918331365e-08,
+      "loss": 2.8536,
+      "step": 8122
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.47348680248105e-08,
+      "loss": 2.9122,
+      "step": 8123
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.458105495712292e-08,
+      "loss": 2.8138,
+      "step": 8124
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4428047740132068e-08,
+      "loss": 3.0191,
+      "step": 8125
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4275846398569825e-08,
+      "loss": 2.8633,
+      "step": 8126
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4124450957038182e-08,
+      "loss": 2.9206,
+      "step": 8127
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3973861440008674e-08,
+      "loss": 2.9016,
+      "step": 8128
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3824077871823494e-08,
+      "loss": 2.8808,
+      "step": 8129
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3675100276693277e-08,
+      "loss": 2.883,
+      "step": 8130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3526928678698759e-08,
+      "loss": 2.862,
+      "step": 8131
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3379563101791893e-08,
+      "loss": 3.029,
+      "step": 8132
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3233003569791403e-08,
+      "loss": 2.8424,
+      "step": 8133
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3087250106388894e-08,
+      "loss": 2.8611,
+      "step": 8134
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2942302735143298e-08,
+      "loss": 2.932,
+      "step": 8135
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2798161479484205e-08,
+      "loss": 3.0753,
+      "step": 8136
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2654826362711314e-08,
+      "loss": 3.0229,
+      "step": 8137
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2512297407992757e-08,
+      "loss": 2.8398,
+      "step": 8138
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2370574638367883e-08,
+      "loss": 2.8484,
+      "step": 8139
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2229658076744477e-08,
+      "loss": 2.8509,
+      "step": 8140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2089547745900988e-08,
+      "loss": 2.9519,
+      "step": 8141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1950243668484296e-08,
+      "loss": 2.8456,
+      "step": 8142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1811745867012502e-08,
+      "loss": 2.9487,
+      "step": 8143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1674054363872146e-08,
+      "loss": 2.8539,
+      "step": 8144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1537169181320418e-08,
+      "loss": 2.8298,
+      "step": 8145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1401090341482401e-08,
+      "loss": 2.896,
+      "step": 8146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1265817866356054e-08,
+      "loss": 3.0511,
+      "step": 8147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1131351777804999e-08,
+      "loss": 3.0239,
+      "step": 8148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0997692097565738e-08,
+      "loss": 3.1087,
+      "step": 8149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0864838847243208e-08,
+      "loss": 2.8132,
+      "step": 8150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0732792048311347e-08,
+      "loss": 3.034,
+      "step": 8151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.06015517221153e-08,
+      "loss": 2.9996,
+      "step": 8152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0471117889868099e-08,
+      "loss": 2.9648,
+      "step": 8153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0341490572653434e-08,
+      "loss": 2.8852,
+      "step": 8154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.02126697914251e-08,
+      "loss": 2.8802,
+      "step": 8155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0084655567004775e-08,
+      "loss": 3.0417,
+      "step": 8156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.95744792008646e-09,
+      "loss": 3.0129,
+      "step": 8157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.831046871230376e-09,
+      "loss": 2.9352,
+      "step": 8158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.705452440869623e-09,
+      "loss": 3.0991,
+      "step": 8159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.580664649304627e-09,
+      "loss": 2.9245,
+      "step": 8160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.456683516707033e-09,
+      "loss": 2.9111,
+      "step": 8161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.333509063116364e-09,
+      "loss": 3.1066,
+      "step": 8162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.211141308443361e-09,
+      "loss": 2.9588,
+      "step": 8163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.089580272467757e-09,
+      "loss": 2.8921,
+      "step": 8164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.968825974838835e-09,
+      "loss": 2.7653,
+      "step": 8165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.84887843507487e-09,
+      "loss": 3.0085,
+      "step": 8166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.729737672565352e-09,
+      "loss": 2.8967,
+      "step": 8167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.611403706567657e-09,
+      "loss": 2.9637,
+      "step": 8168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.493876556209813e-09,
+      "loss": 2.9574,
+      "step": 8169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.377156240488848e-09,
+      "loss": 3.0604,
+      "step": 8170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.261242778271894e-09,
+      "loss": 2.8213,
+      "step": 8171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.146136188295073e-09,
+      "loss": 3.0985,
+      "step": 8172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.031836489164613e-09,
+      "loss": 2.9266,
+      "step": 8173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.91834369935629e-09,
+      "loss": 3.0281,
+      "step": 8174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.805657837214875e-09,
+      "loss": 2.9194,
+      "step": 8175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.693778920955241e-09,
+      "loss": 2.888,
+      "step": 8176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.582706968661258e-09,
+      "loss": 2.7524,
+      "step": 8177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.47244199828745e-09,
+      "loss": 2.9896,
+      "step": 8178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.362984027656783e-09,
+      "loss": 2.7829,
+      "step": 8179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.2543330744623275e-09,
+      "loss": 2.9848,
+      "step": 8180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.146489156266701e-09,
+      "loss": 3.0305,
+      "step": 8181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.039452290501514e-09,
+      "loss": 2.9624,
+      "step": 8182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.933222494469038e-09,
+      "loss": 2.9328,
+      "step": 8183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.827799785339984e-09,
+      "loss": 2.7855,
+      "step": 8184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.723184180155162e-09,
+      "loss": 2.9649,
+      "step": 8185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.619375695824382e-09,
+      "loss": 2.9123,
+      "step": 8186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.5163743491281074e-09,
+      "loss": 2.803,
+      "step": 8187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.414180156715799e-09,
+      "loss": 2.9277,
+      "step": 8188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.3127931351048e-09,
+      "loss": 2.9531,
+      "step": 8189
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.212213300685332e-09,
+      "loss": 2.87,
+      "step": 8190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.112440669713838e-09,
+      "loss": 2.9297,
+      "step": 8191
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.013475258318524e-09,
+      "loss": 2.8542,
+      "step": 8192
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9153170824960415e-09,
+      "loss": 2.9887,
+      "step": 8193
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.817966158112587e-09,
+      "loss": 3.0433,
+      "step": 8194
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.721422500905016e-09,
+      "loss": 3.0123,
+      "step": 8195
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.625686126477514e-09,
+      "loss": 2.911,
+      "step": 8196
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.530757050306035e-09,
+      "loss": 2.9796,
+      "step": 8197
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.436635287734415e-09,
+      "loss": 2.9216,
+      "step": 8198
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.343320853977152e-09,
+      "loss": 2.9305,
+      "step": 8199
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.25081376411718e-09,
+      "loss": 2.9681,
+      "step": 8200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.159114033108093e-09,
+      "loss": 2.9305,
+      "step": 8201
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.068221675772478e-09,
+      "loss": 2.9866,
+      "step": 8202
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.978136706801362e-09,
+      "loss": 2.9939,
+      "step": 8203
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.888859140756985e-09,
+      "loss": 3.0852,
+      "step": 8204
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.800388992070581e-09,
+      "loss": 2.8547,
+      "step": 8205
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.712726275041269e-09,
+      "loss": 2.9918,
+      "step": 8206
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.625871003840488e-09,
+      "loss": 2.905,
+      "step": 8207
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.539823192507009e-09,
+      "loss": 3.2163,
+      "step": 8208
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4545828549497076e-09,
+      "loss": 2.9459,
+      "step": 8209
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.370150004946449e-09,
+      "loss": 2.9939,
+      "step": 8210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.286524656146318e-09,
+      "loss": 2.9306,
+      "step": 8211
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.203706822065723e-09,
+      "loss": 3.0523,
+      "step": 8212
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.121696516091178e-09,
+      "loss": 2.95,
+      "step": 8213
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.040493751480412e-09,
+      "loss": 2.949,
+      "step": 8214
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.960098541357371e-09,
+      "loss": 2.8935,
+      "step": 8215
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8805108987188814e-09,
+      "loss": 3.0716,
+      "step": 8216
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.801730836428541e-09,
+      "loss": 2.9231,
+      "step": 8217
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.723758367220609e-09,
+      "loss": 2.9342,
+      "step": 8218
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.646593503698892e-09,
+      "loss": 2.9909,
+      "step": 8219
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5702362583367455e-09,
+      "loss": 2.8531,
+      "step": 8220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4946866434759643e-09,
+      "loss": 2.8375,
+      "step": 8221
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.419944671328446e-09,
+      "loss": 2.999,
+      "step": 8222
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3460103539767475e-09,
+      "loss": 2.8826,
+      "step": 8223
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.272883703370755e-09,
+      "loss": 2.8663,
+      "step": 8224
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2005647313310127e-09,
+      "loss": 2.9802,
+      "step": 8225
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1290534495470594e-09,
+      "loss": 3.0247,
+      "step": 8226
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.058349869578536e-09,
+      "loss": 3.0784,
+      "step": 8227
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.98845400285408e-09,
+      "loss": 2.9974,
+      "step": 8228
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9193658606707643e-09,
+      "loss": 3.1036,
+      "step": 8229
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.851085454197433e-09,
+      "loss": 2.8837,
+      "step": 8230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7836127944702584e-09,
+      "loss": 2.9276,
+      "step": 8231
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.716947892395516e-09,
+      "loss": 2.8943,
+      "step": 8232
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.651090758750141e-09,
+      "loss": 3.0166,
+      "step": 8233
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5860414041778415e-09,
+      "loss": 2.8642,
+      "step": 8234
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5217998391940947e-09,
+      "loss": 3.0888,
+      "step": 8235
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4583660741833714e-09,
+      "loss": 2.9363,
+      "step": 8236
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3957401193980267e-09,
+      "loss": 2.9534,
+      "step": 8237
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.333921984962184e-09,
+      "loss": 2.9994,
+      "step": 8238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.272911680867851e-09,
+      "loss": 3.0771,
+      "step": 8239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.212709216976583e-09,
+      "loss": 3.0289,
+      "step": 8240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1533146030200402e-09,
+      "loss": 3.0861,
+      "step": 8241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0947278485983214e-09,
+      "loss": 2.7885,
+      "step": 8242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0369489631816286e-09,
+      "loss": 2.9563,
+      "step": 8243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9799779561097134e-09,
+      "loss": 2.7353,
+      "step": 8244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.92381483659132e-09,
+      "loss": 3.0647,
+      "step": 8245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8684596137047427e-09,
+      "loss": 3.025,
+      "step": 8246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.813912296397269e-09,
+      "loss": 3.1402,
+      "step": 8247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7601728934868444e-09,
+      "loss": 3.0095,
+      "step": 8248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7072414136592997e-09,
+      "loss": 2.9799,
+      "step": 8249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.655117865471123e-09,
+      "loss": 2.9494,
+      "step": 8250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6038022573466872e-09,
+      "loss": 2.9633,
+      "step": 8251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5532945975821334e-09,
+      "loss": 2.9076,
+      "step": 8252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5035948943403766e-09,
+      "loss": 2.8907,
+      "step": 8253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4547031556555459e-09,
+      "loss": 2.8276,
+      "step": 8254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4066193894302082e-09,
+      "loss": 3.0688,
+      "step": 8255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3593436034370356e-09,
+      "loss": 2.8854,
+      "step": 8256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3128758053171375e-09,
+      "loss": 2.8863,
+      "step": 8257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2672160025828384e-09,
+      "loss": 2.9833,
+      "step": 8258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2223642026137904e-09,
+      "loss": 2.9319,
+      "step": 8259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1783204126597503e-09,
+      "loss": 2.8799,
+      "step": 8260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1350846398405779e-09,
+      "loss": 2.9436,
+      "step": 8261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0926568911445723e-09,
+      "loss": 2.9666,
+      "step": 8262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0510371734301362e-09,
+      "loss": 2.8507,
+      "step": 8263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0102254934246658e-09,
+      "loss": 2.7729,
+      "step": 8264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.702218577251065e-10,
+      "loss": 2.8875,
+      "step": 8265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.310262727973973e-10,
+      "loss": 2.9427,
+      "step": 8266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.926387449770257e-10,
+      "loss": 2.9271,
+      "step": 8267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.550592804695834e-10,
+      "loss": 2.9934,
+      "step": 8268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.182878853491006e-10,
+      "loss": 2.9309,
+      "step": 8269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.82324565559156e-10,
+      "loss": 2.9411,
+      "step": 8270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.471693269139879e-10,
+      "loss": 2.7824,
+      "step": 8271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.128221750946074e-10,
+      "loss": 3.0008,
+      "step": 8272
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.7928311565435e-10,
+      "loss": 2.886,
+      "step": 8273
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.465521540133246e-10,
+      "loss": 2.9826,
+      "step": 8274
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.146292954628541e-10,
+      "loss": 3.0155,
+      "step": 8275
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.835145451632551e-10,
+      "loss": 2.8988,
+      "step": 8276
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.532079081432829e-10,
+      "loss": 3.0733,
+      "step": 8277
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.237093893017964e-10,
+      "loss": 2.861,
+      "step": 8278
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.950189934077587e-10,
+      "loss": 2.8234,
+      "step": 8279
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.671367250980163e-10,
+      "loss": 2.8406,
+      "step": 8280
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.400625888800747e-10,
+      "loss": 2.8311,
+      "step": 8281
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1379658912932273e-10,
+      "loss": 2.9311,
+      "step": 8282
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8833873009236356e-10,
+      "loss": 2.8851,
+      "step": 8283
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.636890158836837e-10,
+      "loss": 2.9762,
+      "step": 8284
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.398474504884286e-10,
+      "loss": 2.9629,
+      "step": 8285
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1681403775907225e-10,
+      "loss": 2.9776,
+      "step": 8286
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.945887814198578e-10,
+      "loss": 3.0005,
+      "step": 8287
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.731716850629118e-10,
+      "loss": 2.9197,
+      "step": 8288
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5256275215046476e-10,
+      "loss": 2.954,
+      "step": 8289
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.327619860137409e-10,
+      "loss": 2.9435,
+      "step": 8290
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1376938985240293e-10,
+      "loss": 3.1021,
+      "step": 8291
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9558496673788284e-10,
+      "loss": 3.0988,
+      "step": 8292
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7820871960838593e-10,
+      "loss": 3.0752,
+      "step": 8293
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.616406512733315e-10,
+      "loss": 2.9333,
+      "step": 8294
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4588076441057753e-10,
+      "loss": 3.0172,
+      "step": 8295
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3092906156753071e-10,
+      "loss": 3.0592,
+      "step": 8296
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.167855451611466e-10,
+      "loss": 3.0631,
+      "step": 8297
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0345021747737438e-10,
+      "loss": 2.8598,
+      "step": 8298
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.092308067171207e-11,
+      "loss": 2.9935,
+      "step": 8299
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.920413676920646e-11,
+      "loss": 2.9622,
+      "step": 8300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.829338766445315e-11,
+      "loss": 3.0433,
+      "step": 8301
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.8190835120486287e-11,
+      "loss": 3.1155,
+      "step": 8302
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.8896480770999065e-11,
+      "loss": 2.8327,
+      "step": 8303
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0410326117568124e-11,
+      "loss": 2.8924,
+      "step": 8304
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2732372532429114e-11,
+      "loss": 2.8192,
+      "step": 8305
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.586262125681138e-11,
+      "loss": 2.8857,
+      "step": 8306
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.980107340038284e-11,
+      "loss": 2.8697,
+      "step": 8307
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4547729944025535e-11,
+      "loss": 3.0051,
+      "step": 8308
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0102591735949852e-11,
+      "loss": 3.1264,
+      "step": 8309
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.465659494470089e-12,
+      "loss": 2.9515,
+      "step": 8310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.6369338085595617e-12,
+      "loss": 3.0255,
+      "step": 8311
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6164151350750446e-12,
+      "loss": 2.9742,
+      "step": 8312
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.04103799866995e-13,
+      "loss": 2.9769,
+      "step": 8313
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 2.9509,
+      "step": 8314
+    },
+    {
+      "epoch": 1.0,
+      "step": 8314,
+      "total_flos": 1.0719039452925133e+18,
+      "train_loss": 2.989931782105808,
+      "train_runtime": 9263.8611,
+      "train_samples_per_second": 57.433,
+      "train_steps_per_second": 0.897
+    }
+  ],
+  "max_steps": 8314,
+  "num_train_epochs": 1,
+  "total_flos": 1.0719039452925133e+18,
+  "trial_name": null,
+  "trial_params": null
+}