diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,89616 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999706994051979,
+  "eval_steps": 500,
+  "global_step": 14931,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4642857142857145e-08,
+      "loss": 1.4396,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.928571428571429e-08,
+      "loss": 1.6583,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3392857142857142e-07,
+      "loss": 1.516,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7857142857142858e-07,
+      "loss": 1.5863,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2321428571428574e-07,
+      "loss": 1.6348,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6785714285714284e-07,
+      "loss": 1.513,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.125e-07,
+      "loss": 1.5881,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5714285714285716e-07,
+      "loss": 1.8098,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0178571428571434e-07,
+      "loss": 1.7353,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4642857142857147e-07,
+      "loss": 1.5284,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.910714285714286e-07,
+      "loss": 1.7193,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.357142857142857e-07,
+      "loss": 1.6774,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.80357142857143e-07,
+      "loss": 1.8178,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.25e-07,
+      "loss": 1.5249,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.696428571428571e-07,
+      "loss": 1.7163,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 1.4586,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.589285714285714e-07,
+      "loss": 1.4983,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.035714285714287e-07,
+      "loss": 1.3665,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.482142857142859e-07,
+      "loss": 1.4485,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.928571428571429e-07,
+      "loss": 1.3594,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.375000000000001e-07,
+      "loss": 1.4409,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.821428571428572e-07,
+      "loss": 1.3446,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0267857142857143e-06,
+      "loss": 1.362,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0714285714285714e-06,
+      "loss": 1.2474,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1160714285714287e-06,
+      "loss": 1.3399,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.160714285714286e-06,
+      "loss": 1.3458,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.205357142857143e-06,
+      "loss": 1.3134,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.25e-06,
+      "loss": 1.2769,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2946428571428574e-06,
+      "loss": 1.2461,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3392857142857143e-06,
+      "loss": 1.3378,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3839285714285715e-06,
+      "loss": 1.2484,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 1.3125,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.473214285714286e-06,
+      "loss": 1.236,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5178571428571428e-06,
+      "loss": 1.2928,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5625e-06,
+      "loss": 1.1524,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6071428571428574e-06,
+      "loss": 1.1102,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6517857142857144e-06,
+      "loss": 1.1946,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6964285714285717e-06,
+      "loss": 1.1618,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7410714285714286e-06,
+      "loss": 1.1809,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7857142857142859e-06,
+      "loss": 1.1964,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.830357142857143e-06,
+      "loss": 1.1829,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8750000000000003e-06,
+      "loss": 1.142,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9196428571428573e-06,
+      "loss": 1.2241,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9642857142857144e-06,
+      "loss": 1.1746,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0089285714285715e-06,
+      "loss": 1.165,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0535714285714286e-06,
+      "loss": 1.194,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.098214285714286e-06,
+      "loss": 1.1126,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1428571428571427e-06,
+      "loss": 1.1433,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1875000000000002e-06,
+      "loss": 1.1537,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2321428571428573e-06,
+      "loss": 1.0878,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2767857142857144e-06,
+      "loss": 1.1393,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.321428571428572e-06,
+      "loss": 1.0923,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3660714285714285e-06,
+      "loss": 1.1648,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.410714285714286e-06,
+      "loss": 1.1646,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.455357142857143e-06,
+      "loss": 1.1796,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-06,
+      "loss": 1.2425,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5446428571428573e-06,
+      "loss": 1.0603,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5892857142857148e-06,
+      "loss": 1.1337,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.633928571428572e-06,
+      "loss": 1.0913,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6785714285714285e-06,
+      "loss": 1.0274,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7232142857142856e-06,
+      "loss": 1.1785,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.767857142857143e-06,
+      "loss": 1.1238,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8125e-06,
+      "loss": 1.0879,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 1.1288,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9017857142857148e-06,
+      "loss": 1.0356,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.946428571428572e-06,
+      "loss": 1.1736,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.991071428571429e-06,
+      "loss": 1.0499,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0357142857142856e-06,
+      "loss": 1.0284,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.080357142857143e-06,
+      "loss": 1.07,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.125e-06,
+      "loss": 1.1434,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1696428571428572e-06,
+      "loss": 1.0735,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2142857142857147e-06,
+      "loss": 1.193,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.258928571428572e-06,
+      "loss": 1.1229,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.303571428571429e-06,
+      "loss": 1.0497,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3482142857142855e-06,
+      "loss": 1.1029,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3928571428571435e-06,
+      "loss": 1.0605,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4375e-06,
+      "loss": 1.0836,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.482142857142857e-06,
+      "loss": 1.0799,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5267857142857147e-06,
+      "loss": 1.2205,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5714285714285718e-06,
+      "loss": 1.1091,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.616071428571429e-06,
+      "loss": 1.0905,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.660714285714286e-06,
+      "loss": 1.0983,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7053571428571434e-06,
+      "loss": 1.1661,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 1.0436,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.794642857142857e-06,
+      "loss": 1.1115,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.839285714285715e-06,
+      "loss": 1.0058,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.883928571428572e-06,
+      "loss": 1.0636,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.928571428571429e-06,
+      "loss": 1.1198,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9732142857142855e-06,
+      "loss": 1.1358,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.017857142857143e-06,
+      "loss": 1.1564,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0625000000000005e-06,
+      "loss": 1.0657,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.107142857142857e-06,
+      "loss": 1.0363,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.151785714285715e-06,
+      "loss": 1.078,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.196428571428572e-06,
+      "loss": 1.017,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.241071428571429e-06,
+      "loss": 1.0098,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 1.039,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.330357142857143e-06,
+      "loss": 1.0832,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3750000000000005e-06,
+      "loss": 0.987,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.419642857142857e-06,
+      "loss": 0.9817,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.464285714285715e-06,
+      "loss": 1.1297,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.508928571428572e-06,
+      "loss": 1.0508,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.553571428571429e-06,
+      "loss": 1.1475,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5982142857142854e-06,
+      "loss": 1.1363,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.642857142857144e-06,
+      "loss": 1.0754,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6875000000000004e-06,
+      "loss": 1.1253,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.732142857142857e-06,
+      "loss": 1.0634,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.776785714285715e-06,
+      "loss": 1.0632,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.821428571428572e-06,
+      "loss": 1.0183,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.866071428571429e-06,
+      "loss": 1.0801,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.910714285714286e-06,
+      "loss": 1.1319,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.955357142857144e-06,
+      "loss": 1.085,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 1.0336,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.044642857142857e-06,
+      "loss": 1.156,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0892857142857146e-06,
+      "loss": 1.0921,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.133928571428571e-06,
+      "loss": 1.0403,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1785714285714296e-06,
+      "loss": 1.0776,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.223214285714286e-06,
+      "loss": 1.0391,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.267857142857144e-06,
+      "loss": 1.1214,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3125e-06,
+      "loss": 1.0208,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.357142857142857e-06,
+      "loss": 1.0684,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4017857142857145e-06,
+      "loss": 1.0624,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.446428571428571e-06,
+      "loss": 1.1484,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4910714285714295e-06,
+      "loss": 1.1094,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.535714285714286e-06,
+      "loss": 1.1361,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.580357142857144e-06,
+      "loss": 1.0591,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.625e-06,
+      "loss": 1.0761,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.669642857142858e-06,
+      "loss": 0.991,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 1.0793,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.758928571428571e-06,
+      "loss": 1.0694,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8035714285714295e-06,
+      "loss": 1.0124,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.848214285714286e-06,
+      "loss": 1.0139,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.892857142857144e-06,
+      "loss": 1.1553,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9375e-06,
+      "loss": 1.0538,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.982142857142858e-06,
+      "loss": 1.0198,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0267857142857145e-06,
+      "loss": 0.9774,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.071428571428571e-06,
+      "loss": 1.0277,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.1160714285714295e-06,
+      "loss": 1.1862,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.160714285714286e-06,
+      "loss": 1.1273,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.205357142857144e-06,
+      "loss": 1.0565,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.25e-06,
+      "loss": 1.0312,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.294642857142858e-06,
+      "loss": 1.0245,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3392857142857145e-06,
+      "loss": 1.0627,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.383928571428571e-06,
+      "loss": 0.9888,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4285714285714295e-06,
+      "loss": 1.0366,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.473214285714287e-06,
+      "loss": 1.1311,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.517857142857144e-06,
+      "loss": 1.1544,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5625e-06,
+      "loss": 1.1015,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.607142857142858e-06,
+      "loss": 1.0829,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.6517857142857144e-06,
+      "loss": 1.0857,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.696428571428571e-06,
+      "loss": 0.9963,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.7410714285714294e-06,
+      "loss": 1.0601,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.785714285714287e-06,
+      "loss": 1.0961,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.830357142857144e-06,
+      "loss": 1.0078,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.875e-06,
+      "loss": 1.0756,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.919642857142858e-06,
+      "loss": 1.0137,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.964285714285714e-06,
+      "loss": 1.0394,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.008928571428572e-06,
+      "loss": 1.0334,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.053571428571429e-06,
+      "loss": 1.1255,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.098214285714287e-06,
+      "loss": 1.0774,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 1.0764,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1875e-06,
+      "loss": 1.0885,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.232142857142858e-06,
+      "loss": 1.0694,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.276785714285714e-06,
+      "loss": 1.0156,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.321428571428572e-06,
+      "loss": 1.0621,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.366071428571429e-06,
+      "loss": 1.1363,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.410714285714287e-06,
+      "loss": 1.0934,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.4553571428571435e-06,
+      "loss": 1.0752,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 1.0461,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.544642857142858e-06,
+      "loss": 1.1153,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.589285714285714e-06,
+      "loss": 1.0728,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.633928571428572e-06,
+      "loss": 1.0262,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.67857142857143e-06,
+      "loss": 1.0377,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.723214285714287e-06,
+      "loss": 1.1136,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.767857142857144e-06,
+      "loss": 1.0745,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.8125e-06,
+      "loss": 1.0265,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.857142857142858e-06,
+      "loss": 0.9894,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.901785714285715e-06,
+      "loss": 1.124,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.946428571428571e-06,
+      "loss": 1.0629,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.99107142857143e-06,
+      "loss": 1.1672,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.035714285714286e-06,
+      "loss": 1.1036,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.080357142857143e-06,
+      "loss": 1.0638,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.125000000000001e-06,
+      "loss": 1.1284,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.169642857142858e-06,
+      "loss": 1.0868,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.214285714285714e-06,
+      "loss": 1.1294,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.258928571428572e-06,
+      "loss": 1.1176,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.30357142857143e-06,
+      "loss": 1.0144,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.348214285714287e-06,
+      "loss": 1.0027,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.392857142857144e-06,
+      "loss": 1.0158,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.4375e-06,
+      "loss": 1.0183,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.482142857142858e-06,
+      "loss": 1.0625,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.526785714285715e-06,
+      "loss": 1.1043,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 1.0469,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.61607142857143e-06,
+      "loss": 0.9871,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.660714285714286e-06,
+      "loss": 1.1147,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.705357142857143e-06,
+      "loss": 0.9994,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 1.1105,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.794642857142858e-06,
+      "loss": 1.0258,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.839285714285714e-06,
+      "loss": 1.1176,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.883928571428572e-06,
+      "loss": 1.0384,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.92857142857143e-06,
+      "loss": 1.0757,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.973214285714287e-06,
+      "loss": 1.0715,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.017857142857144e-06,
+      "loss": 0.9927,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.0625e-06,
+      "loss": 1.0711,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.107142857142858e-06,
+      "loss": 1.078,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.151785714285715e-06,
+      "loss": 0.9961,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.196428571428571e-06,
+      "loss": 1.0652,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.24107142857143e-06,
+      "loss": 1.0962,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 1.0272,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.330357142857143e-06,
+      "loss": 0.9187,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.375000000000001e-06,
+      "loss": 1.0528,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.419642857142858e-06,
+      "loss": 1.0846,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.464285714285714e-06,
+      "loss": 1.0312,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.508928571428572e-06,
+      "loss": 1.0902,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.55357142857143e-06,
+      "loss": 1.0432,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.598214285714287e-06,
+      "loss": 1.101,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.642857142857144e-06,
+      "loss": 1.1044,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.6875e-06,
+      "loss": 1.1098,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.732142857142858e-06,
+      "loss": 1.0603,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.776785714285715e-06,
+      "loss": 1.0076,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.821428571428573e-06,
+      "loss": 1.01,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.86607142857143e-06,
+      "loss": 1.1087,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.910714285714288e-06,
+      "loss": 1.0639,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.955357142857143e-06,
+      "loss": 1.04,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1e-05,
+      "loss": 1.0207,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0044642857142858e-05,
+      "loss": 1.0489,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0089285714285714e-05,
+      "loss": 1.0621,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0133928571428573e-05,
+      "loss": 1.0662,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0178571428571429e-05,
+      "loss": 1.0749,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0223214285714287e-05,
+      "loss": 1.1098,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0267857142857142e-05,
+      "loss": 1.0489,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0312500000000002e-05,
+      "loss": 1.0208,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0357142857142859e-05,
+      "loss": 1.0617,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0401785714285715e-05,
+      "loss": 1.0698,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0446428571428572e-05,
+      "loss": 1.0784,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0491071428571428e-05,
+      "loss": 1.1716,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0535714285714287e-05,
+      "loss": 1.0757,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0580357142857143e-05,
+      "loss": 1.101,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0625e-05,
+      "loss": 1.1199,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0669642857142858e-05,
+      "loss": 1.0654,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0714285714285714e-05,
+      "loss": 1.1243,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0758928571428573e-05,
+      "loss": 1.1148,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0803571428571429e-05,
+      "loss": 1.1014,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0848214285714287e-05,
+      "loss": 1.0552,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0892857142857142e-05,
+      "loss": 1.0385,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0937500000000002e-05,
+      "loss": 1.0841,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0982142857142859e-05,
+      "loss": 0.9458,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1026785714285715e-05,
+      "loss": 1.0996,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1071428571428572e-05,
+      "loss": 1.1026,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1116071428571428e-05,
+      "loss": 1.073,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1160714285714287e-05,
+      "loss": 1.0354,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1205357142857143e-05,
+      "loss": 1.0829,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.125e-05,
+      "loss": 0.9827,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1294642857142858e-05,
+      "loss": 0.9989,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1339285714285716e-05,
+      "loss": 1.0298,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1383928571428573e-05,
+      "loss": 1.0256,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 1.0828,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1473214285714287e-05,
+      "loss": 1.0385,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1517857142857142e-05,
+      "loss": 1.0938,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1562500000000002e-05,
+      "loss": 1.0865,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1607142857142859e-05,
+      "loss": 1.0832,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1651785714285715e-05,
+      "loss": 1.0392,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1696428571428572e-05,
+      "loss": 1.1124,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1741071428571428e-05,
+      "loss": 1.1312,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1785714285714287e-05,
+      "loss": 1.1613,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1830357142857143e-05,
+      "loss": 1.084,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1875e-05,
+      "loss": 0.9959,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.191964285714286e-05,
+      "loss": 1.042,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1964285714285716e-05,
+      "loss": 1.0929,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2008928571428573e-05,
+      "loss": 1.1348,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2053571428571429e-05,
+      "loss": 1.1212,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2098214285714286e-05,
+      "loss": 0.9997,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2142857142857142e-05,
+      "loss": 1.1294,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2187500000000001e-05,
+      "loss": 1.1223,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2232142857142859e-05,
+      "loss": 1.1042,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2276785714285715e-05,
+      "loss": 1.0421,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2321428571428572e-05,
+      "loss": 1.0305,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.236607142857143e-05,
+      "loss": 1.1442,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2410714285714287e-05,
+      "loss": 1.0875,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2455357142857143e-05,
+      "loss": 1.073,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.25e-05,
+      "loss": 1.1101,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.254464285714286e-05,
+      "loss": 1.0641,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2589285714285716e-05,
+      "loss": 1.1352,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2633928571428573e-05,
+      "loss": 1.0828,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2678571428571429e-05,
+      "loss": 1.1203,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2723214285714286e-05,
+      "loss": 1.0849,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2767857142857142e-05,
+      "loss": 1.0965,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2812500000000001e-05,
+      "loss": 1.0512,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 1.0987,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2901785714285715e-05,
+      "loss": 1.0971,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2946428571428574e-05,
+      "loss": 1.1125,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.299107142857143e-05,
+      "loss": 1.0876,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3035714285714287e-05,
+      "loss": 1.0235,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3080357142857143e-05,
+      "loss": 1.0625,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3125e-05,
+      "loss": 1.1059,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.316964285714286e-05,
+      "loss": 1.1311,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3214285714285716e-05,
+      "loss": 1.1174,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3258928571428573e-05,
+      "loss": 1.1449,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3303571428571429e-05,
+      "loss": 1.0025,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3348214285714286e-05,
+      "loss": 1.0949,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3392857142857142e-05,
+      "loss": 1.0977,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3437500000000001e-05,
+      "loss": 1.0989,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3482142857142859e-05,
+      "loss": 1.084,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3526785714285715e-05,
+      "loss": 1.1076,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3571428571428574e-05,
+      "loss": 1.107,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.361607142857143e-05,
+      "loss": 1.1089,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3660714285714287e-05,
+      "loss": 1.1109,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3705357142857143e-05,
+      "loss": 1.1263,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.375e-05,
+      "loss": 1.1217,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.379464285714286e-05,
+      "loss": 1.0592,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3839285714285715e-05,
+      "loss": 1.0692,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3883928571428573e-05,
+      "loss": 1.1016,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3928571428571429e-05,
+      "loss": 1.0232,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3973214285714288e-05,
+      "loss": 0.9832,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4017857142857144e-05,
+      "loss": 1.0691,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4062500000000001e-05,
+      "loss": 1.0307,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4107142857142859e-05,
+      "loss": 1.0094,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4151785714285715e-05,
+      "loss": 1.0275,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4196428571428574e-05,
+      "loss": 1.0445,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.424107142857143e-05,
+      "loss": 1.0753,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 1.1227,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4330357142857143e-05,
+      "loss": 1.112,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4375e-05,
+      "loss": 1.0749,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.441964285714286e-05,
+      "loss": 1.0258,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4464285714285715e-05,
+      "loss": 1.0493,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4508928571428573e-05,
+      "loss": 1.1066,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4553571428571429e-05,
+      "loss": 1.107,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4598214285714288e-05,
+      "loss": 1.121,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4642857142857144e-05,
+      "loss": 1.0884,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4687500000000001e-05,
+      "loss": 1.121,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4732142857142859e-05,
+      "loss": 1.094,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4776785714285715e-05,
+      "loss": 1.0506,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4821428571428574e-05,
+      "loss": 1.0448,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.486607142857143e-05,
+      "loss": 1.0564,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4910714285714287e-05,
+      "loss": 1.0871,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4955357142857143e-05,
+      "loss": 1.1322,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.0772,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.504464285714286e-05,
+      "loss": 1.0816,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5089285714285715e-05,
+      "loss": 1.1859,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5133928571428573e-05,
+      "loss": 1.0859,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5178571428571429e-05,
+      "loss": 1.0542,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5223214285714288e-05,
+      "loss": 1.0371,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5267857142857144e-05,
+      "loss": 1.0832,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5312500000000003e-05,
+      "loss": 1.0807,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.535714285714286e-05,
+      "loss": 1.0637,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5401785714285715e-05,
+      "loss": 1.0091,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5446428571428574e-05,
+      "loss": 1.1546,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.549107142857143e-05,
+      "loss": 1.069,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.553571428571429e-05,
+      "loss": 1.0671,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5580357142857145e-05,
+      "loss": 1.1115,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5625e-05,
+      "loss": 1.0665,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.566964285714286e-05,
+      "loss": 1.0465,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 1.0906,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5758928571428575e-05,
+      "loss": 1.0552,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.580357142857143e-05,
+      "loss": 1.0367,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5848214285714286e-05,
+      "loss": 1.0474,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5892857142857142e-05,
+      "loss": 1.0838,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.59375e-05,
+      "loss": 1.1379,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.598214285714286e-05,
+      "loss": 1.2116,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6026785714285716e-05,
+      "loss": 1.0804,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6071428571428572e-05,
+      "loss": 1.1315,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6116071428571428e-05,
+      "loss": 1.0724,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6160714285714287e-05,
+      "loss": 1.0817,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6205357142857143e-05,
+      "loss": 1.183,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 1.0674,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6294642857142858e-05,
+      "loss": 1.1917,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6339285714285717e-05,
+      "loss": 1.0731,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6383928571428573e-05,
+      "loss": 1.0415,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.642857142857143e-05,
+      "loss": 1.1134,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6473214285714288e-05,
+      "loss": 1.0823,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6517857142857144e-05,
+      "loss": 1.1268,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6562500000000003e-05,
+      "loss": 0.9884,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.660714285714286e-05,
+      "loss": 1.0718,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6651785714285714e-05,
+      "loss": 1.1679,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6696428571428574e-05,
+      "loss": 1.0945,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.674107142857143e-05,
+      "loss": 1.0767,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.678571428571429e-05,
+      "loss": 1.0417,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6830357142857144e-05,
+      "loss": 1.1791,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6875e-05,
+      "loss": 1.105,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.691964285714286e-05,
+      "loss": 1.1033,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6964285714285715e-05,
+      "loss": 1.0734,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7008928571428574e-05,
+      "loss": 1.1065,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.705357142857143e-05,
+      "loss": 1.0973,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7098214285714286e-05,
+      "loss": 1.0486,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 1.1369,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.71875e-05,
+      "loss": 1.134,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.723214285714286e-05,
+      "loss": 1.09,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7276785714285716e-05,
+      "loss": 1.0251,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7321428571428572e-05,
+      "loss": 1.0277,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.736607142857143e-05,
+      "loss": 1.0118,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7410714285714287e-05,
+      "loss": 1.1403,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7455357142857143e-05,
+      "loss": 1.1321,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 1.0434,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.754464285714286e-05,
+      "loss": 1.074,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7589285714285717e-05,
+      "loss": 1.1488,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7633928571428573e-05,
+      "loss": 1.0915,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.767857142857143e-05,
+      "loss": 1.1665,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7723214285714288e-05,
+      "loss": 1.0678,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7767857142857143e-05,
+      "loss": 1.1017,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7812500000000003e-05,
+      "loss": 1.1471,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.785714285714286e-05,
+      "loss": 1.1114,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7901785714285714e-05,
+      "loss": 1.1154,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7946428571428573e-05,
+      "loss": 1.1117,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.799107142857143e-05,
+      "loss": 1.1016,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.803571428571429e-05,
+      "loss": 1.1462,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8080357142857144e-05,
+      "loss": 1.1434,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8125e-05,
+      "loss": 1.156,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.816964285714286e-05,
+      "loss": 1.0906,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8214285714285715e-05,
+      "loss": 1.0997,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8258928571428574e-05,
+      "loss": 1.1675,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.830357142857143e-05,
+      "loss": 1.0897,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8348214285714286e-05,
+      "loss": 1.1024,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8392857142857142e-05,
+      "loss": 1.0782,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.84375e-05,
+      "loss": 1.1393,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.848214285714286e-05,
+      "loss": 1.1703,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8526785714285716e-05,
+      "loss": 1.1555,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 1.0451,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.861607142857143e-05,
+      "loss": 1.0914,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8660714285714287e-05,
+      "loss": 1.0772,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8705357142857143e-05,
+      "loss": 1.099,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 1.0816,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.879464285714286e-05,
+      "loss": 1.0312,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8839285714285717e-05,
+      "loss": 1.0561,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8883928571428573e-05,
+      "loss": 1.012,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.892857142857143e-05,
+      "loss": 1.0519,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8973214285714288e-05,
+      "loss": 1.1413,
+      "step": 425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9017857142857143e-05,
+      "loss": 1.1697,
+      "step": 426
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9062500000000003e-05,
+      "loss": 1.0571,
+      "step": 427
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.910714285714286e-05,
+      "loss": 1.0945,
+      "step": 428
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9151785714285714e-05,
+      "loss": 1.0829,
+      "step": 429
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9196428571428573e-05,
+      "loss": 1.0869,
+      "step": 430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.924107142857143e-05,
+      "loss": 1.0652,
+      "step": 431
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.928571428571429e-05,
+      "loss": 1.1155,
+      "step": 432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9330357142857144e-05,
+      "loss": 1.1343,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9375e-05,
+      "loss": 1.0834,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.941964285714286e-05,
+      "loss": 1.1288,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9464285714285715e-05,
+      "loss": 1.163,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9508928571428574e-05,
+      "loss": 1.151,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.955357142857143e-05,
+      "loss": 1.0773,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.959821428571429e-05,
+      "loss": 1.091,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9642857142857145e-05,
+      "loss": 1.0971,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.96875e-05,
+      "loss": 1.1185,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.973214285714286e-05,
+      "loss": 1.0861,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9776785714285716e-05,
+      "loss": 1.15,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9821428571428575e-05,
+      "loss": 1.0626,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.986607142857143e-05,
+      "loss": 0.9715,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9910714285714287e-05,
+      "loss": 1.0252,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9955357142857142e-05,
+      "loss": 1.1248,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.0934,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999764737512e-05,
+      "loss": 1.0971,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999058950056e-05,
+      "loss": 1.1552,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999997882637672e-05,
+      "loss": 1.1166,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996235800404e-05,
+      "loss": 1.1927,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994118438342e-05,
+      "loss": 1.1285,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999991530551575e-05,
+      "loss": 1.0976,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998847214023e-05,
+      "loss": 1.1115,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998494320445e-05,
+      "loss": 1.1323,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99999809437444e-05,
+      "loss": 1.0744,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999976473760275e-05,
+      "loss": 1.1519,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999971533252274e-05,
+      "loss": 1.1273,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996612222064e-05,
+      "loss": 1.1053,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996024066562e-05,
+      "loss": 1.097,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999953888587492e-05,
+      "loss": 1.1354,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999947065986557e-05,
+      "loss": 1.0882,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999993977286314e-05,
+      "loss": 1.1332,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999932009217574e-05,
+      "loss": 1.0869,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999923775050235e-05,
+      "loss": 1.1162,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99999150703615e-05,
+      "loss": 1.0998,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999990589515179e-05,
+      "loss": 1.1713,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999896249421527e-05,
+      "loss": 1.1151,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988613317117e-05,
+      "loss": 1.0995,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999987554640119e-05,
+      "loss": 1.1132,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999986448911209e-05,
+      "loss": 1.0739,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999985296130439e-05,
+      "loss": 1.1471,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999840962978633e-05,
+      "loss": 1.0709,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982849413538e-05,
+      "loss": 1.0693,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999981555477522e-05,
+      "loss": 1.1126,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999802144898762e-05,
+      "loss": 1.0472,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999788264506636e-05,
+      "loss": 1.1266,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999773913599495e-05,
+      "loss": 1.0503,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999759092178014e-05,
+      "loss": 1.0236,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999743800242893e-05,
+      "loss": 1.0865,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999728037794848e-05,
+      "loss": 1.0694,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999711804834624e-05,
+      "loss": 1.1134,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999969510136298e-05,
+      "loss": 1.0927,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999967792738071e-05,
+      "loss": 1.0665,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999660282888613e-05,
+      "loss": 1.0698,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999642167887523e-05,
+      "loss": 1.152,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999623582378293e-05,
+      "loss": 1.0526,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99996045263618e-05,
+      "loss": 1.076,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999584999838935e-05,
+      "loss": 1.0743,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999565002810623e-05,
+      "loss": 1.1522,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999544535277796e-05,
+      "loss": 1.1117,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999523597241427e-05,
+      "loss": 1.1135,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999502188702497e-05,
+      "loss": 1.0677,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999948030966201e-05,
+      "loss": 1.0912,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999457960121003e-05,
+      "loss": 1.147,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999943514008052e-05,
+      "loss": 1.0533,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999941184954164e-05,
+      "loss": 1.1019,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999938808850545e-05,
+      "loss": 1.1084,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999936385697308e-05,
+      "loss": 1.1034,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999339154945667e-05,
+      "loss": 1.0576,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999931398242437e-05,
+      "loss": 1.0652,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999928833941037e-05,
+      "loss": 1.069,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999262225904883e-05,
+      "loss": 1.1222,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999235641909132e-05,
+      "loss": 1.1563,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999208587424368e-05,
+      "loss": 1.1141,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999181062451866e-05,
+      "loss": 1.0945,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999915306699292e-05,
+      "loss": 1.1518,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999124601048845e-05,
+      "loss": 1.2343,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999095664620983e-05,
+      "loss": 1.1262,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999066257710696e-05,
+      "loss": 1.0256,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999036380319367e-05,
+      "loss": 1.0474,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99990060324484e-05,
+      "loss": 1.0751,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998975214099227e-05,
+      "loss": 1.0508,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998943925273293e-05,
+      "loss": 1.1478,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998912165972073e-05,
+      "loss": 1.0729,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999887993619706e-05,
+      "loss": 1.1111,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998847235949777e-05,
+      "loss": 1.1341,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999881406523175e-05,
+      "loss": 1.1231,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998780424044552e-05,
+      "loss": 1.0416,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999874631238976e-05,
+      "loss": 1.0707,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999871173026898e-05,
+      "loss": 1.1493,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999867667768384e-05,
+      "loss": 1.1675,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998641154635986e-05,
+      "loss": 1.1041,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998605161127092e-05,
+      "loss": 1.0966,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999856869715885e-05,
+      "loss": 1.0831,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998531762732982e-05,
+      "loss": 1.1533,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998494357851215e-05,
+      "loss": 1.086,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999845648251532e-05,
+      "loss": 1.0488,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998418136727077e-05,
+      "loss": 1.1392,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999837932048828e-05,
+      "loss": 1.128,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998340033800766e-05,
+      "loss": 1.1303,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999830027666638e-05,
+      "loss": 1.1392,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998260049086994e-05,
+      "loss": 1.1481,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99982193510645e-05,
+      "loss": 1.0252,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998178182600812e-05,
+      "loss": 1.1581,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998136543697865e-05,
+      "loss": 1.1656,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998094434357626e-05,
+      "loss": 1.0878,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999805185458207e-05,
+      "loss": 1.1565,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998008804373203e-05,
+      "loss": 1.1124,
+      "step": 540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997965283733044e-05,
+      "loss": 1.0606,
+      "step": 541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999792129266365e-05,
+      "loss": 1.1081,
+      "step": 542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997876831167086e-05,
+      "loss": 1.1491,
+      "step": 543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999783189924545e-05,
+      "loss": 1.1342,
+      "step": 544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997786496900845e-05,
+      "loss": 1.1517,
+      "step": 545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997740624135417e-05,
+      "loss": 1.1473,
+      "step": 546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997694280951322e-05,
+      "loss": 1.0569,
+      "step": 547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999764746735074e-05,
+      "loss": 1.0578,
+      "step": 548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997600183335866e-05,
+      "loss": 1.1651,
+      "step": 549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997552428908938e-05,
+      "loss": 1.1535,
+      "step": 550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99975042040722e-05,
+      "loss": 0.9723,
+      "step": 551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997455508827914e-05,
+      "loss": 1.1027,
+      "step": 552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997406343178376e-05,
+      "loss": 1.1604,
+      "step": 553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99973567071259e-05,
+      "loss": 1.0583,
+      "step": 554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999730660067282e-05,
+      "loss": 1.1229,
+      "step": 555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997256023821493e-05,
+      "loss": 1.1312,
+      "step": 556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99972049765743e-05,
+      "loss": 1.1289,
+      "step": 557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997153458933642e-05,
+      "loss": 1.0383,
+      "step": 558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997101470901944e-05,
+      "loss": 1.1246,
+      "step": 559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997049012481656e-05,
+      "loss": 1.0705,
+      "step": 560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996996083675238e-05,
+      "loss": 1.1165,
+      "step": 561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996942684485183e-05,
+      "loss": 1.1095,
+      "step": 562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999688881491401e-05,
+      "loss": 1.0772,
+      "step": 563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996834474964243e-05,
+      "loss": 1.1242,
+      "step": 564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999677966463845e-05,
+      "loss": 1.1297,
+      "step": 565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99967243839392e-05,
+      "loss": 1.0528,
+      "step": 566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99966686328691e-05,
+      "loss": 1.0253,
+      "step": 567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996612411430774e-05,
+      "loss": 1.0879,
+      "step": 568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996555719626864e-05,
+      "loss": 1.0469,
+      "step": 569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996498557460042e-05,
+      "loss": 1.1126,
+      "step": 570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999644092493299e-05,
+      "loss": 1.1245,
+      "step": 571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996382822048424e-05,
+      "loss": 1.111,
+      "step": 572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999632424880908e-05,
+      "loss": 1.1647,
+      "step": 573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996265205217714e-05,
+      "loss": 1.079,
+      "step": 574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99962056912771e-05,
+      "loss": 1.1245,
+      "step": 575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999614570699004e-05,
+      "loss": 1.1287,
+      "step": 576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996085252359357e-05,
+      "loss": 1.1385,
+      "step": 577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996024327387896e-05,
+      "loss": 1.1611,
+      "step": 578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995962932078524e-05,
+      "loss": 1.0946,
+      "step": 579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995901066434128e-05,
+      "loss": 1.1792,
+      "step": 580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999583873045762e-05,
+      "loss": 1.0944,
+      "step": 581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995775924151934e-05,
+      "loss": 1.1502,
+      "step": 582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995712647520024e-05,
+      "loss": 1.0984,
+      "step": 583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995648900564864e-05,
+      "loss": 1.0472,
+      "step": 584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999558468328946e-05,
+      "loss": 1.1538,
+      "step": 585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995519995696832e-05,
+      "loss": 1.0943,
+      "step": 586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995454837790018e-05,
+      "loss": 1.069,
+      "step": 587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999538920957209e-05,
+      "loss": 1.1812,
+      "step": 588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995323111046132e-05,
+      "loss": 1.1248,
+      "step": 589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999525654221526e-05,
+      "loss": 1.1411,
+      "step": 590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99951895030826e-05,
+      "loss": 1.1022,
+      "step": 591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999512199365131e-05,
+      "loss": 1.1338,
+      "step": 592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995054013924563e-05,
+      "loss": 1.1324,
+      "step": 593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999498556390556e-05,
+      "loss": 1.1174,
+      "step": 594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994916643597525e-05,
+      "loss": 1.1567,
+      "step": 595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994847253003695e-05,
+      "loss": 1.2129,
+      "step": 596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994777392127334e-05,
+      "loss": 1.0787,
+      "step": 597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994707060971738e-05,
+      "loss": 1.1201,
+      "step": 598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999463625954021e-05,
+      "loss": 1.0714,
+      "step": 599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994564987836083e-05,
+      "loss": 1.1432,
+      "step": 600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994493245862708e-05,
+      "loss": 1.1618,
+      "step": 601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999442103362346e-05,
+      "loss": 1.1532,
+      "step": 602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999434835112174e-05,
+      "loss": 1.2176,
+      "step": 603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999427519836097e-05,
+      "loss": 1.093,
+      "step": 604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994201575344587e-05,
+      "loss": 1.0327,
+      "step": 605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999412748207606e-05,
+      "loss": 1.1338,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994052918558867e-05,
+      "loss": 1.1007,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993977884796525e-05,
+      "loss": 1.1229,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999390238079256e-05,
+      "loss": 1.0602,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999382640655053e-05,
+      "loss": 1.0741,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993749962074e-05,
+      "loss": 1.1691,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993673047366578e-05,
+      "loss": 1.0974,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999359566243187e-05,
+      "loss": 1.075,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993517807273533e-05,
+      "loss": 1.0792,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999343948189522e-05,
+      "loss": 1.0762,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993360686300617e-05,
+      "loss": 1.1387,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993281420493432e-05,
+      "loss": 1.0784,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99932016844774e-05,
+      "loss": 1.1258,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993121478256267e-05,
+      "loss": 1.0357,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993040801833808e-05,
+      "loss": 1.0512,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992959655213816e-05,
+      "loss": 1.1683,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999287803840012e-05,
+      "loss": 1.1171,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992795951396547e-05,
+      "loss": 1.0964,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992713394206967e-05,
+      "loss": 1.1416,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992630366835263e-05,
+      "loss": 1.0763,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992546869285338e-05,
+      "loss": 1.1018,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999246290156113e-05,
+      "loss": 1.0946,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992378463666582e-05,
+      "loss": 1.1304,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992293555605667e-05,
+      "loss": 1.1192,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992208177382384e-05,
+      "loss": 1.1184,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999212232900075e-05,
+      "loss": 1.1105,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99920360104648e-05,
+      "loss": 1.0493,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99919492217786e-05,
+      "loss": 1.0309,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991861962946232e-05,
+      "loss": 1.147,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99917742339718e-05,
+      "loss": 1.0503,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991686034859436e-05,
+      "loss": 1.1132,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991597365613286e-05,
+      "loss": 1.0626,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991508226237525e-05,
+      "loss": 1.0963,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991418616736346e-05,
+      "loss": 1.0847,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991328537113962e-05,
+      "loss": 1.1716,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999123798737462e-05,
+      "loss": 1.0698,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991146967522576e-05,
+      "loss": 1.0701,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991055477562105e-05,
+      "loss": 1.1217,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990963517497525e-05,
+      "loss": 1.085,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990871087333152e-05,
+      "loss": 1.167,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990778187073344e-05,
+      "loss": 1.0635,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990684816722467e-05,
+      "loss": 1.0475,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990590976284917e-05,
+      "loss": 1.1404,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990496665765105e-05,
+      "loss": 1.0317,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990401885167472e-05,
+      "loss": 1.0646,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990306634496477e-05,
+      "loss": 1.0706,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990210913756603e-05,
+      "loss": 1.1272,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990114722952353e-05,
+      "loss": 1.0628,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990018062088253e-05,
+      "loss": 1.0865,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998992093116885e-05,
+      "loss": 1.1263,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989823330198714e-05,
+      "loss": 1.0625,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998972525918244e-05,
+      "loss": 1.0377,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989626718124643e-05,
+      "loss": 1.1055,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989527707029954e-05,
+      "loss": 1.0885,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989428225903035e-05,
+      "loss": 1.109,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998932827474857e-05,
+      "loss": 1.0985,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989227853571258e-05,
+      "loss": 1.1372,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989126962375825e-05,
+      "loss": 1.0343,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989025601167018e-05,
+      "loss": 1.1164,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998892376994961e-05,
+      "loss": 1.0237,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988821468728387e-05,
+      "loss": 1.0441,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988718697508162e-05,
+      "loss": 1.0835,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988615456293774e-05,
+      "loss": 1.0991,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998851174509008e-05,
+      "loss": 1.0936,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988407563901965e-05,
+      "loss": 1.0592,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998830291273432e-05,
+      "loss": 1.0518,
+      "step": 671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988197791592077e-05,
+      "loss": 1.13,
+      "step": 672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998809220048018e-05,
+      "loss": 1.0772,
+      "step": 673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99879861394036e-05,
+      "loss": 1.111,
+      "step": 674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998787960836732e-05,
+      "loss": 1.078,
+      "step": 675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998777260737636e-05,
+      "loss": 1.191,
+      "step": 676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987665136435753e-05,
+      "loss": 1.1069,
+      "step": 677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987557195550553e-05,
+      "loss": 1.0902,
+      "step": 678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987448784725843e-05,
+      "loss": 1.0953,
+      "step": 679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987339903966722e-05,
+      "loss": 1.1288,
+      "step": 680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998723055327831e-05,
+      "loss": 1.0877,
+      "step": 681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987120732665758e-05,
+      "loss": 1.1551,
+      "step": 682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987010442134228e-05,
+      "loss": 1.078,
+      "step": 683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986899681688915e-05,
+      "loss": 1.0898,
+      "step": 684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986788451335025e-05,
+      "loss": 1.128,
+      "step": 685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986676751077796e-05,
+      "loss": 1.028,
+      "step": 686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998656458092248e-05,
+      "loss": 1.0773,
+      "step": 687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986451940874357e-05,
+      "loss": 1.0486,
+      "step": 688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998633883093873e-05,
+      "loss": 1.1399,
+      "step": 689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986225251120915e-05,
+      "loss": 1.1022,
+      "step": 690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986111201426263e-05,
+      "loss": 1.0419,
+      "step": 691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998599668186013e-05,
+      "loss": 1.018,
+      "step": 692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985881692427912e-05,
+      "loss": 1.1087,
+      "step": 693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985766233135022e-05,
+      "loss": 1.0881,
+      "step": 694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985650303986888e-05,
+      "loss": 1.1131,
+      "step": 695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985533904988968e-05,
+      "loss": 1.0519,
+      "step": 696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998541703614673e-05,
+      "loss": 1.0773,
+      "step": 697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985299697465686e-05,
+      "loss": 1.057,
+      "step": 698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998518188895135e-05,
+      "loss": 1.1033,
+      "step": 699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985063610609262e-05,
+      "loss": 1.1155,
+      "step": 700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984944862444996e-05,
+      "loss": 1.0579,
+      "step": 701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998482564446413e-05,
+      "loss": 1.1225,
+      "step": 702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984705956672284e-05,
+      "loss": 1.104,
+      "step": 703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984585799075078e-05,
+      "loss": 1.0722,
+      "step": 704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984465171678175e-05,
+      "loss": 1.0968,
+      "step": 705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984344074487245e-05,
+      "loss": 1.1148,
+      "step": 706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998422250750799e-05,
+      "loss": 1.1337,
+      "step": 707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984100470746127e-05,
+      "loss": 1.1284,
+      "step": 708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99839779642074e-05,
+      "loss": 1.0691,
+      "step": 709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998385498789757e-05,
+      "loss": 1.0848,
+      "step": 710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998373154182243e-05,
+      "loss": 1.158,
+      "step": 711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983607625987786e-05,
+      "loss": 1.1518,
+      "step": 712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983483240399463e-05,
+      "loss": 1.0302,
+      "step": 713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983358385063317e-05,
+      "loss": 1.0453,
+      "step": 714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983233059985222e-05,
+      "loss": 1.0769,
+      "step": 715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998310726517108e-05,
+      "loss": 1.1285,
+      "step": 716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982981000626807e-05,
+      "loss": 0.9904,
+      "step": 717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998285426635834e-05,
+      "loss": 1.0891,
+      "step": 718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982727062371645e-05,
+      "loss": 1.074,
+      "step": 719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998259938867271e-05,
+      "loss": 1.0941,
+      "step": 720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998247124526754e-05,
+      "loss": 1.0353,
+      "step": 721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982342632162164e-05,
+      "loss": 1.1021,
+      "step": 722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982213549362634e-05,
+      "loss": 1.0171,
+      "step": 723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982083996875024e-05,
+      "loss": 1.0761,
+      "step": 724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981953974705428e-05,
+      "loss": 1.0987,
+      "step": 725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981823482859966e-05,
+      "loss": 1.1326,
+      "step": 726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981692521344778e-05,
+      "loss": 1.1046,
+      "step": 727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981561090166026e-05,
+      "loss": 1.1201,
+      "step": 728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981429189329893e-05,
+      "loss": 1.0771,
+      "step": 729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981296818842587e-05,
+      "loss": 1.1884,
+      "step": 730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998116397871033e-05,
+      "loss": 1.1076,
+      "step": 731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981030668939384e-05,
+      "loss": 1.1073,
+      "step": 732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980896889536013e-05,
+      "loss": 1.0686,
+      "step": 733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980762640506514e-05,
+      "loss": 1.1148,
+      "step": 734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980627921857202e-05,
+      "loss": 1.0149,
+      "step": 735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998049273359442e-05,
+      "loss": 1.1193,
+      "step": 736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980357075724527e-05,
+      "loss": 1.131,
+      "step": 737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99802209482539e-05,
+      "loss": 1.163,
+      "step": 738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980084351188952e-05,
+      "loss": 1.1179,
+      "step": 739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997994728453611e-05,
+      "loss": 1.0962,
+      "step": 740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997980974830182e-05,
+      "loss": 1.0763,
+      "step": 741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979671742492556e-05,
+      "loss": 1.1513,
+      "step": 742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997953326711481e-05,
+      "loss": 1.1314,
+      "step": 743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979394322175094e-05,
+      "loss": 1.0579,
+      "step": 744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979254907679952e-05,
+      "loss": 1.0451,
+      "step": 745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979115023635944e-05,
+      "loss": 1.1184,
+      "step": 746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978974670049643e-05,
+      "loss": 1.1671,
+      "step": 747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997883384692766e-05,
+      "loss": 1.0758,
+      "step": 748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978692554276623e-05,
+      "loss": 1.1488,
+      "step": 749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997855079210318e-05,
+      "loss": 0.983,
+      "step": 750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978408560413993e-05,
+      "loss": 1.1104,
+      "step": 751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978265859215764e-05,
+      "loss": 1.0526,
+      "step": 752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99781226885152e-05,
+      "loss": 1.0588,
+      "step": 753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977979048319045e-05,
+      "loss": 1.0761,
+      "step": 754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997783493863405e-05,
+      "loss": 1.0555,
+      "step": 755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977690359467e-05,
+      "loss": 1.1055,
+      "step": 756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977545310824696e-05,
+      "loss": 1.0695,
+      "step": 757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977399792713968e-05,
+      "loss": 1.0883,
+      "step": 758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977253805141656e-05,
+      "loss": 1.1754,
+      "step": 759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977107348114635e-05,
+      "loss": 1.0979,
+      "step": 760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997696042163979e-05,
+      "loss": 1.0859,
+      "step": 761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997681302572404e-05,
+      "loss": 1.0921,
+      "step": 762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976665160374314e-05,
+      "loss": 1.1529,
+      "step": 763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976516825597577e-05,
+      "loss": 1.1051,
+      "step": 764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976368021400802e-05,
+      "loss": 1.0809,
+      "step": 765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976218747790996e-05,
+      "loss": 1.0646,
+      "step": 766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976069004775177e-05,
+      "loss": 1.1162,
+      "step": 767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975918792360395e-05,
+      "loss": 1.0658,
+      "step": 768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997576811055372e-05,
+      "loss": 1.1151,
+      "step": 769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975616959362238e-05,
+      "loss": 1.0967,
+      "step": 770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997546533879306e-05,
+      "loss": 1.1288,
+      "step": 771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997531324885332e-05,
+      "loss": 1.1037,
+      "step": 772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975160689550178e-05,
+      "loss": 1.139,
+      "step": 773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975007660890813e-05,
+      "loss": 1.0926,
+      "step": 774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997485416288242e-05,
+      "loss": 1.0592,
+      "step": 775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974700195532223e-05,
+      "loss": 1.0822,
+      "step": 776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997454575884747e-05,
+      "loss": 1.0977,
+      "step": 777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974390852835427e-05,
+      "loss": 1.0493,
+      "step": 778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974235477503377e-05,
+      "loss": 1.0606,
+      "step": 779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974079632858635e-05,
+      "loss": 1.0766,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973923318908536e-05,
+      "loss": 1.0667,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973766535660433e-05,
+      "loss": 1.087,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973609283121704e-05,
+      "loss": 1.1414,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973451561299746e-05,
+      "loss": 1.1474,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997329337020198e-05,
+      "loss": 1.0747,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997313470983585e-05,
+      "loss": 1.1345,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972975580208822e-05,
+      "loss": 1.0711,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972815981328385e-05,
+      "loss": 1.0764,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972655913202046e-05,
+      "loss": 1.054,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972495375837337e-05,
+      "loss": 1.1108,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997233436924181e-05,
+      "loss": 1.0845,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972172893423046e-05,
+      "loss": 1.17,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972010948388637e-05,
+      "loss": 1.0783,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997184853414621e-05,
+      "loss": 1.0385,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971685650703398e-05,
+      "loss": 1.0632,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971522298067872e-05,
+      "loss": 1.1651,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971358476247315e-05,
+      "loss": 1.1514,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997119418524944e-05,
+      "loss": 1.1196,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997102942508197e-05,
+      "loss": 1.1061,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997086419575266e-05,
+      "loss": 1.0516,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970698497269282e-05,
+      "loss": 1.0719,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997053232963964e-05,
+      "loss": 1.0677,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970365692871547e-05,
+      "loss": 1.1246,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970198586972845e-05,
+      "loss": 1.135,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99700310119514e-05,
+      "loss": 1.066,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996986296781509e-05,
+      "loss": 1.0493,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969694454571825e-05,
+      "loss": 1.0316,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969525472229536e-05,
+      "loss": 1.1533,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969356020796172e-05,
+      "loss": 1.0924,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969186100279706e-05,
+      "loss": 1.0934,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969015710688137e-05,
+      "loss": 1.1207,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968844852029474e-05,
+      "loss": 1.065,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968673524311765e-05,
+      "loss": 1.1103,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968501727543063e-05,
+      "loss": 1.0708,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968329461731462e-05,
+      "loss": 1.1302,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968156726885057e-05,
+      "loss": 1.0829,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967983523011985e-05,
+      "loss": 1.105,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996780985012039e-05,
+      "loss": 1.1027,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996763570821844e-05,
+      "loss": 1.1327,
+      "step": 819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996746109731434e-05,
+      "loss": 1.0812,
+      "step": 820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967286017416298e-05,
+      "loss": 1.0685,
+      "step": 821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996711046853255e-05,
+      "loss": 1.0377,
+      "step": 822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996693445067136e-05,
+      "loss": 1.0612,
+      "step": 823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966757963841012e-05,
+      "loss": 1.1713,
+      "step": 824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966581008049806e-05,
+      "loss": 1.172,
+      "step": 825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996640358330607e-05,
+      "loss": 1.1239,
+      "step": 826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996622568961815e-05,
+      "loss": 1.0838,
+      "step": 827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966047326994423e-05,
+      "loss": 1.1191,
+      "step": 828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965868495443273e-05,
+      "loss": 1.1996,
+      "step": 829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996568919497312e-05,
+      "loss": 1.1367,
+      "step": 830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99655094255924e-05,
+      "loss": 1.1217,
+      "step": 831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996532918730957e-05,
+      "loss": 1.1408,
+      "step": 832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965148480133112e-05,
+      "loss": 1.1528,
+      "step": 833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964967304071522e-05,
+      "loss": 1.1207,
+      "step": 834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996478565913334e-05,
+      "loss": 1.0194,
+      "step": 835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964603545327096e-05,
+      "loss": 1.099,
+      "step": 836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996442096266137e-05,
+      "loss": 1.1511,
+      "step": 837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964237911144745e-05,
+      "loss": 1.1189,
+      "step": 838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964054390785845e-05,
+      "loss": 1.041,
+      "step": 839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963870401593294e-05,
+      "loss": 1.106,
+      "step": 840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963685943575756e-05,
+      "loss": 1.0703,
+      "step": 841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963501016741907e-05,
+      "loss": 0.9752,
+      "step": 842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963315621100448e-05,
+      "loss": 1.1247,
+      "step": 843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963129756660103e-05,
+      "loss": 1.0631,
+      "step": 844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962943423429618e-05,
+      "loss": 1.0728,
+      "step": 845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962756621417764e-05,
+      "loss": 1.1514,
+      "step": 846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962569350633327e-05,
+      "loss": 1.117,
+      "step": 847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996238161108511e-05,
+      "loss": 1.04,
+      "step": 848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962193402781965e-05,
+      "loss": 1.0498,
+      "step": 849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996200472573273e-05,
+      "loss": 1.1327,
+      "step": 850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996181557994629e-05,
+      "loss": 1.1057,
+      "step": 851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961625965431555e-05,
+      "loss": 1.0913,
+      "step": 852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996143588219743e-05,
+      "loss": 1.1031,
+      "step": 853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961245330252865e-05,
+      "loss": 1.0676,
+      "step": 854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961054309606827e-05,
+      "loss": 1.0576,
+      "step": 855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960862820268305e-05,
+      "loss": 1.0675,
+      "step": 856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996067086224631e-05,
+      "loss": 1.152,
+      "step": 857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960478435549867e-05,
+      "loss": 1.1569,
+      "step": 858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996028554018804e-05,
+      "loss": 1.1146,
+      "step": 859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960092176169897e-05,
+      "loss": 1.0058,
+      "step": 860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959898343504545e-05,
+      "loss": 1.0165,
+      "step": 861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959704042201095e-05,
+      "loss": 1.0931,
+      "step": 862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995950927226869e-05,
+      "loss": 1.1194,
+      "step": 863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99593140337165e-05,
+      "loss": 1.0951,
+      "step": 864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995911832655371e-05,
+      "loss": 1.1633,
+      "step": 865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995892215078953e-05,
+      "loss": 1.0428,
+      "step": 866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958725506433184e-05,
+      "loss": 1.0748,
+      "step": 867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995852839349393e-05,
+      "loss": 1.0584,
+      "step": 868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958330811981044e-05,
+      "loss": 1.0556,
+      "step": 869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958132761903817e-05,
+      "loss": 1.048,
+      "step": 870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957934243271576e-05,
+      "loss": 1.0949,
+      "step": 871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995773525609365e-05,
+      "loss": 1.1616,
+      "step": 872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995753580037941e-05,
+      "loss": 1.1068,
+      "step": 873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957335876138245e-05,
+      "loss": 1.101,
+      "step": 874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995713548337955e-05,
+      "loss": 1.0278,
+      "step": 875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956934622112762e-05,
+      "loss": 1.0517,
+      "step": 876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956733292347332e-05,
+      "loss": 1.1084,
+      "step": 877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956531494092732e-05,
+      "loss": 1.105,
+      "step": 878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956329227358453e-05,
+      "loss": 1.1216,
+      "step": 879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995612649215402e-05,
+      "loss": 1.1301,
+      "step": 880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955923288488965e-05,
+      "loss": 1.0537,
+      "step": 881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955719616372853e-05,
+      "loss": 1.1511,
+      "step": 882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955515475815266e-05,
+      "loss": 1.0929,
+      "step": 883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995531086682581e-05,
+      "loss": 1.0601,
+      "step": 884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955105789414112e-05,
+      "loss": 1.0825,
+      "step": 885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954900243589824e-05,
+      "loss": 1.0495,
+      "step": 886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995469422936261e-05,
+      "loss": 1.0661,
+      "step": 887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954487746742174e-05,
+      "loss": 1.0555,
+      "step": 888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954280795738223e-05,
+      "loss": 1.0758,
+      "step": 889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954073376360498e-05,
+      "loss": 1.1219,
+      "step": 890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953865488618755e-05,
+      "loss": 1.094,
+      "step": 891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953657132522784e-05,
+      "loss": 1.0617,
+      "step": 892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953448308082382e-05,
+      "loss": 1.158,
+      "step": 893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953239015307377e-05,
+      "loss": 1.0431,
+      "step": 894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995302925420761e-05,
+      "loss": 1.1393,
+      "step": 895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995281902479296e-05,
+      "loss": 1.1053,
+      "step": 896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995260832707332e-05,
+      "loss": 1.1076,
+      "step": 897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952397161058596e-05,
+      "loss": 1.0952,
+      "step": 898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995218552675873e-05,
+      "loss": 1.1003,
+      "step": 899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951973424183672e-05,
+      "loss": 1.1182,
+      "step": 900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995176085334341e-05,
+      "loss": 1.1597,
+      "step": 901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951547814247943e-05,
+      "loss": 1.0979,
+      "step": 902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951334306907294e-05,
+      "loss": 1.1606,
+      "step": 903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951120331331514e-05,
+      "loss": 1.0921,
+      "step": 904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950905887530662e-05,
+      "loss": 1.0471,
+      "step": 905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950690975514836e-05,
+      "loss": 1.1134,
+      "step": 906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950475595294145e-05,
+      "loss": 1.1378,
+      "step": 907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950259746878724e-05,
+      "loss": 1.0373,
+      "step": 908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995004343027873e-05,
+      "loss": 1.0355,
+      "step": 909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994982664550434e-05,
+      "loss": 1.0811,
+      "step": 910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994960939256575e-05,
+      "loss": 1.0067,
+      "step": 911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949391671473193e-05,
+      "loss": 1.1027,
+      "step": 912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949173482236905e-05,
+      "loss": 1.1516,
+      "step": 913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994895482486715e-05,
+      "loss": 1.0267,
+      "step": 914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994873569937423e-05,
+      "loss": 1.0788,
+      "step": 915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994851610576844e-05,
+      "loss": 1.1206,
+      "step": 916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994829604406012e-05,
+      "loss": 1.093,
+      "step": 917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948075514259623e-05,
+      "loss": 1.0581,
+      "step": 918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947854516377326e-05,
+      "loss": 1.0729,
+      "step": 919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947633050423626e-05,
+      "loss": 1.0275,
+      "step": 920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947411116408942e-05,
+      "loss": 1.1422,
+      "step": 921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947188714343723e-05,
+      "loss": 1.0525,
+      "step": 922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994696584423843e-05,
+      "loss": 1.0881,
+      "step": 923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946742506103546e-05,
+      "loss": 1.1078,
+      "step": 924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946518699949588e-05,
+      "loss": 1.1159,
+      "step": 925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946294425787077e-05,
+      "loss": 1.022,
+      "step": 926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994606968362657e-05,
+      "loss": 1.0098,
+      "step": 927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945844473478643e-05,
+      "loss": 1.1141,
+      "step": 928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945618795353895e-05,
+      "loss": 0.9854,
+      "step": 929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945392649262936e-05,
+      "loss": 1.1477,
+      "step": 930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994516603521642e-05,
+      "loss": 1.1457,
+      "step": 931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944938953224994e-05,
+      "loss": 1.1167,
+      "step": 932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994471140329935e-05,
+      "loss": 1.1213,
+      "step": 933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99444833854502e-05,
+      "loss": 1.0921,
+      "step": 934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944254899688268e-05,
+      "loss": 1.1019,
+      "step": 935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99440259460243e-05,
+      "loss": 1.0771,
+      "step": 936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994379652446908e-05,
+      "loss": 1.1289,
+      "step": 937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943566635033394e-05,
+      "loss": 1.0651,
+      "step": 938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994333627772806e-05,
+      "loss": 1.0464,
+      "step": 939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994310545256392e-05,
+      "loss": 1.1519,
+      "step": 940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942874159551833e-05,
+      "loss": 1.0192,
+      "step": 941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942642398702683e-05,
+      "loss": 1.1012,
+      "step": 942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942410170027375e-05,
+      "loss": 1.114,
+      "step": 943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942177473536833e-05,
+      "loss": 1.1015,
+      "step": 944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994194430924201e-05,
+      "loss": 1.0547,
+      "step": 945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941710677153873e-05,
+      "loss": 1.0566,
+      "step": 946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994147657728342e-05,
+      "loss": 1.1703,
+      "step": 947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994124200964166e-05,
+      "loss": 1.0764,
+      "step": 948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941006974239634e-05,
+      "loss": 1.1533,
+      "step": 949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99407714710884e-05,
+      "loss": 1.0641,
+      "step": 950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994053550019904e-05,
+      "loss": 1.0935,
+      "step": 951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940299061582653e-05,
+      "loss": 1.1067,
+      "step": 952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940062155250374e-05,
+      "loss": 1.0952,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939824781213338e-05,
+      "loss": 1.1313,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939586939482717e-05,
+      "loss": 1.1661,
+      "step": 955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993934863006971e-05,
+      "loss": 1.1094,
+      "step": 956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993910985298552e-05,
+      "loss": 1.0393,
+      "step": 957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938870608241384e-05,
+      "loss": 1.1184,
+      "step": 958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993863089584857e-05,
+      "loss": 1.1295,
+      "step": 959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993839071581834e-05,
+      "loss": 1.0656,
+      "step": 960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938150068162007e-05,
+      "loss": 1.054,
+      "step": 961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993790895289089e-05,
+      "loss": 1.155,
+      "step": 962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937667370016335e-05,
+      "loss": 1.0138,
+      "step": 963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937425319549707e-05,
+      "loss": 1.0871,
+      "step": 964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99371828015024e-05,
+      "loss": 1.0718,
+      "step": 965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993693981588582e-05,
+      "loss": 1.085,
+      "step": 966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99366963627114e-05,
+      "loss": 1.0119,
+      "step": 967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99364524419906e-05,
+      "loss": 1.0855,
+      "step": 968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936208053734892e-05,
+      "loss": 1.083,
+      "step": 969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9935963197955783e-05,
+      "loss": 1.0595,
+      "step": 970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993571787466478e-05,
+      "loss": 1.1037,
+      "step": 971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993547208387344e-05,
+      "loss": 1.0961,
+      "step": 972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993522582559332e-05,
+      "loss": 1.1145,
+      "step": 973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993497909983601e-05,
+      "loss": 1.0867,
+      "step": 974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993473190661312e-05,
+      "loss": 1.1326,
+      "step": 975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934484245936277e-05,
+      "loss": 1.0945,
+      "step": 976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993423611781714e-05,
+      "loss": 1.1076,
+      "step": 977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993398752226738e-05,
+      "loss": 1.0681,
+      "step": 978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933738459298692e-05,
+      "loss": 1.0752,
+      "step": 979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99334889289228e-05,
+      "loss": 1.1493,
+      "step": 980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933238931151445e-05,
+      "loss": 1.0269,
+      "step": 981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932988465996383e-05,
+      "loss": 1.1403,
+      "step": 982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932737533469405e-05,
+      "loss": 1.1105,
+      "step": 983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932486133582318e-05,
+      "loss": 1.0811,
+      "step": 984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993223426634695e-05,
+      "loss": 1.1245,
+      "step": 985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993198193177515e-05,
+      "loss": 1.0697,
+      "step": 986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931729129878798e-05,
+      "loss": 1.0261,
+      "step": 987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931475860669777e-05,
+      "loss": 1.1008,
+      "step": 988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931222124160014e-05,
+      "loss": 1.0679,
+      "step": 989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930967920361442e-05,
+      "loss": 1.1342,
+      "step": 990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993071324928603e-05,
+      "loss": 1.1226,
+      "step": 991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993045811094575e-05,
+      "loss": 1.1135,
+      "step": 992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930202505352616e-05,
+      "loss": 1.0961,
+      "step": 993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992994643251865e-05,
+      "loss": 1.0458,
+      "step": 994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929689892455902e-05,
+      "loss": 1.144,
+      "step": 995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929432885176444e-05,
+      "loss": 1.0546,
+      "step": 996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929175410692368e-05,
+      "loss": 1.0873,
+      "step": 997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992891746901579e-05,
+      "loss": 1.1004,
+      "step": 998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992865906015884e-05,
+      "loss": 1.1045,
+      "step": 999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928400184133684e-05,
+      "loss": 1.0887,
+      "step": 1000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928140840952504e-05,
+      "loss": 1.0867,
+      "step": 1001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927881030627496e-05,
+      "loss": 1.0942,
+      "step": 1002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927620753170892e-05,
+      "loss": 1.0799,
+      "step": 1003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927360008594934e-05,
+      "loss": 1.0565,
+      "step": 1004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992709879691189e-05,
+      "loss": 0.9897,
+      "step": 1005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926837118134054e-05,
+      "loss": 1.1258,
+      "step": 1006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926574972273737e-05,
+      "loss": 1.1465,
+      "step": 1007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926312359343274e-05,
+      "loss": 1.0929,
+      "step": 1008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992604927935502e-05,
+      "loss": 1.0962,
+      "step": 1009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925785732321356e-05,
+      "loss": 1.1031,
+      "step": 1010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925521718254684e-05,
+      "loss": 1.0942,
+      "step": 1011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992525723716742e-05,
+      "loss": 1.1167,
+      "step": 1012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924992289072014e-05,
+      "loss": 1.1026,
+      "step": 1013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992472687398093e-05,
+      "loss": 1.059,
+      "step": 1014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924460991906656e-05,
+      "loss": 1.112,
+      "step": 1015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924194642861708e-05,
+      "loss": 1.0836,
+      "step": 1016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923927826858613e-05,
+      "loss": 1.1327,
+      "step": 1017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923660543909926e-05,
+      "loss": 1.1247,
+      "step": 1018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923392794028225e-05,
+      "loss": 1.1027,
+      "step": 1019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923124577226104e-05,
+      "loss": 1.1263,
+      "step": 1020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992285589351619e-05,
+      "loss": 1.1209,
+      "step": 1021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992258674291112e-05,
+      "loss": 1.0976,
+      "step": 1022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922317125423557e-05,
+      "loss": 0.9933,
+      "step": 1023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922047041066195e-05,
+      "loss": 1.0457,
+      "step": 1024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921776489851738e-05,
+      "loss": 1.0053,
+      "step": 1025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992150547179291e-05,
+      "loss": 1.0294,
+      "step": 1026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992123398690247e-05,
+      "loss": 1.1188,
+      "step": 1027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920962035193193e-05,
+      "loss": 1.0776,
+      "step": 1028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920689616677872e-05,
+      "loss": 1.0946,
+      "step": 1029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920416731369325e-05,
+      "loss": 1.0808,
+      "step": 1030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992014337928039e-05,
+      "loss": 1.1096,
+      "step": 1031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991986956042393e-05,
+      "loss": 1.0111,
+      "step": 1032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919595274812838e-05,
+      "loss": 1.053,
+      "step": 1033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919320522460004e-05,
+      "loss": 1.0908,
+      "step": 1034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991904530337837e-05,
+      "loss": 1.1203,
+      "step": 1035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918769617580875e-05,
+      "loss": 1.1196,
+      "step": 1036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918493465080496e-05,
+      "loss": 1.0515,
+      "step": 1037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918216845890228e-05,
+      "loss": 1.0312,
+      "step": 1038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917939760023078e-05,
+      "loss": 1.0716,
+      "step": 1039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917662207492098e-05,
+      "loss": 1.1261,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917384188310335e-05,
+      "loss": 1.0826,
+      "step": 1041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917105702490878e-05,
+      "loss": 1.1326,
+      "step": 1042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916826750046824e-05,
+      "loss": 1.1037,
+      "step": 1043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916547330991304e-05,
+      "loss": 1.1682,
+      "step": 1044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916267445337466e-05,
+      "loss": 1.0924,
+      "step": 1045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915987093098474e-05,
+      "loss": 1.0752,
+      "step": 1046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991570627428752e-05,
+      "loss": 1.0967,
+      "step": 1047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991542498891782e-05,
+      "loss": 1.0543,
+      "step": 1048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991514323700261e-05,
+      "loss": 1.0782,
+      "step": 1049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914861018555146e-05,
+      "loss": 1.0741,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914578333588704e-05,
+      "loss": 1.0459,
+      "step": 1051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914295182116593e-05,
+      "loss": 1.0178,
+      "step": 1052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991401156415212e-05,
+      "loss": 1.1298,
+      "step": 1053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913727479708652e-05,
+      "loss": 1.0611,
+      "step": 1054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913442928799538e-05,
+      "loss": 1.1552,
+      "step": 1055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913157911438175e-05,
+      "loss": 1.0535,
+      "step": 1056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991287242763797e-05,
+      "loss": 1.0554,
+      "step": 1057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991258647741236e-05,
+      "loss": 1.1229,
+      "step": 1058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912300060774796e-05,
+      "loss": 1.1206,
+      "step": 1059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912013177738755e-05,
+      "loss": 1.0819,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991172582831774e-05,
+      "loss": 1.0537,
+      "step": 1061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911438012525267e-05,
+      "loss": 1.0813,
+      "step": 1062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911149730374878e-05,
+      "loss": 1.0869,
+      "step": 1063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910860981880142e-05,
+      "loss": 1.1316,
+      "step": 1064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991057176705464e-05,
+      "loss": 1.0888,
+      "step": 1065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991028208591198e-05,
+      "loss": 1.1266,
+      "step": 1066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99099919384658e-05,
+      "loss": 1.0902,
+      "step": 1067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990970132472974e-05,
+      "loss": 1.0989,
+      "step": 1068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909410244717485e-05,
+      "loss": 1.0701,
+      "step": 1069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990911869844273e-05,
+      "loss": 1.0405,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908826685919186e-05,
+      "loss": 1.0985,
+      "step": 1071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99085342071606e-05,
+      "loss": 1.0917,
+      "step": 1072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908241262180725e-05,
+      "loss": 1.0836,
+      "step": 1073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907947850993356e-05,
+      "loss": 1.0956,
+      "step": 1074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990765397361229e-05,
+      "loss": 1.0987,
+      "step": 1075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907359630051363e-05,
+      "loss": 1.0353,
+      "step": 1076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907064820324416e-05,
+      "loss": 1.0761,
+      "step": 1077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906769544445324e-05,
+      "loss": 1.1505,
+      "step": 1078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990647380242798e-05,
+      "loss": 1.0822,
+      "step": 1079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99061775942863e-05,
+      "loss": 1.0355,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905880920034225e-05,
+      "loss": 1.1353,
+      "step": 1081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905583779685708e-05,
+      "loss": 1.107,
+      "step": 1082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905286173254734e-05,
+      "loss": 1.1991,
+      "step": 1083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99049881007553e-05,
+      "loss": 1.0962,
+      "step": 1084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990468956220144e-05,
+      "loss": 1.1106,
+      "step": 1085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99043905576072e-05,
+      "loss": 1.1156,
+      "step": 1086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990409108698664e-05,
+      "loss": 1.1136,
+      "step": 1087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990379115035386e-05,
+      "loss": 1.0712,
+      "step": 1088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903490747722972e-05,
+      "loss": 1.0994,
+      "step": 1089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903189879108102e-05,
+      "loss": 1.1347,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902888544523417e-05,
+      "loss": 1.1102,
+      "step": 1091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902586743983088e-05,
+      "loss": 1.1389,
+      "step": 1092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902284477501323e-05,
+      "loss": 1.0468,
+      "step": 1093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901981745092337e-05,
+      "loss": 1.1136,
+      "step": 1094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990167854677038e-05,
+      "loss": 1.0492,
+      "step": 1095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901374882549714e-05,
+      "loss": 1.0993,
+      "step": 1096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990107075244463e-05,
+      "loss": 1.0253,
+      "step": 1097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900766156469436e-05,
+      "loss": 1.1548,
+      "step": 1098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900461094638466e-05,
+      "loss": 1.063,
+      "step": 1099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990015556696607e-05,
+      "loss": 1.0405,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989984957346663e-05,
+      "loss": 1.0648,
+      "step": 1101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989954311415454e-05,
+      "loss": 1.0238,
+      "step": 1102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899236189044218e-05,
+      "loss": 1.0684,
+      "step": 1103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989892879815011e-05,
+      "loss": 1.0912,
+      "step": 1104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898620941486676e-05,
+      "loss": 1.0825,
+      "step": 1105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898312619068404e-05,
+      "loss": 1.0468,
+      "step": 1106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98980038309098e-05,
+      "loss": 1.1923,
+      "step": 1107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897694577025394e-05,
+      "loss": 1.1798,
+      "step": 1108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897384857429734e-05,
+      "loss": 1.0572,
+      "step": 1109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897074672137396e-05,
+      "loss": 1.0016,
+      "step": 1110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896764021162976e-05,
+      "loss": 1.1054,
+      "step": 1111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989645290452109e-05,
+      "loss": 1.0393,
+      "step": 1112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896141322226376e-05,
+      "loss": 1.1039,
+      "step": 1113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895829274293493e-05,
+      "loss": 1.0977,
+      "step": 1114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989551676073713e-05,
+      "loss": 1.1181,
+      "step": 1115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989520378157198e-05,
+      "loss": 1.0286,
+      "step": 1116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894890336812783e-05,
+      "loss": 1.0815,
+      "step": 1117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894576426474277e-05,
+      "loss": 1.0645,
+      "step": 1118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894262050571238e-05,
+      "loss": 1.04,
+      "step": 1119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893947209118457e-05,
+      "loss": 0.9925,
+      "step": 1120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893631902130747e-05,
+      "loss": 1.1506,
+      "step": 1121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893316129622945e-05,
+      "loss": 1.1506,
+      "step": 1122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892999891609904e-05,
+      "loss": 1.0224,
+      "step": 1123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892683188106512e-05,
+      "loss": 1.0536,
+      "step": 1124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989236601912767e-05,
+      "loss": 1.1519,
+      "step": 1125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892048384688294e-05,
+      "loss": 1.1046,
+      "step": 1126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891730284803336e-05,
+      "loss": 1.0085,
+      "step": 1127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989141171948776e-05,
+      "loss": 1.1457,
+      "step": 1128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989109268875656e-05,
+      "loss": 1.1129,
+      "step": 1129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890773192624737e-05,
+      "loss": 1.1651,
+      "step": 1130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890453231107332e-05,
+      "loss": 1.0272,
+      "step": 1131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890132804219405e-05,
+      "loss": 1.1542,
+      "step": 1132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988981191197602e-05,
+      "loss": 1.0909,
+      "step": 1133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988949055439229e-05,
+      "loss": 1.075,
+      "step": 1134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889168731483323e-05,
+      "loss": 1.1118,
+      "step": 1135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988884644326427e-05,
+      "loss": 1.039,
+      "step": 1136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988852368975029e-05,
+      "loss": 1.0895,
+      "step": 1137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988820047095657e-05,
+      "loss": 1.1138,
+      "step": 1138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887876786898325e-05,
+      "loss": 1.0736,
+      "step": 1139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887552637590774e-05,
+      "loss": 1.1705,
+      "step": 1140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887228023049178e-05,
+      "loss": 1.1295,
+      "step": 1141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988690294328881e-05,
+      "loss": 1.0232,
+      "step": 1142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988657739832496e-05,
+      "loss": 1.1763,
+      "step": 1143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988625138817295e-05,
+      "loss": 1.0582,
+      "step": 1144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885924912848123e-05,
+      "loss": 1.1325,
+      "step": 1145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988559797236583e-05,
+      "loss": 1.0371,
+      "step": 1146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885270566741467e-05,
+      "loss": 1.1428,
+      "step": 1147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988494269599043e-05,
+      "loss": 1.0682,
+      "step": 1148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988461436012815e-05,
+      "loss": 1.0458,
+      "step": 1149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988428555917008e-05,
+      "loss": 1.0564,
+      "step": 1150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988395629313168e-05,
+      "loss": 1.0741,
+      "step": 1151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988362656202845e-05,
+      "loss": 1.1164,
+      "step": 1152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883296365875908e-05,
+      "loss": 1.0509,
+      "step": 1153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882965704689582e-05,
+      "loss": 1.1246,
+      "step": 1154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882634578485036e-05,
+      "loss": 1.0883,
+      "step": 1155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882302987277846e-05,
+      "loss": 1.1223,
+      "step": 1156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881970931083622e-05,
+      "loss": 1.1269,
+      "step": 1157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988163840991798e-05,
+      "loss": 1.1349,
+      "step": 1158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988130542379657e-05,
+      "loss": 1.0086,
+      "step": 1159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880971972735058e-05,
+      "loss": 1.0634,
+      "step": 1160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988063805674914e-05,
+      "loss": 1.1241,
+      "step": 1161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988030367585452e-05,
+      "loss": 1.1112,
+      "step": 1162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987996883006693e-05,
+      "loss": 1.0635,
+      "step": 1163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879633519402133e-05,
+      "loss": 1.053,
+      "step": 1164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98792977438759e-05,
+      "loss": 1.1011,
+      "step": 1165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878961503504035e-05,
+      "loss": 1.0464,
+      "step": 1166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878624798302354e-05,
+      "loss": 1.1862,
+      "step": 1167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878287628286702e-05,
+      "loss": 1.1764,
+      "step": 1168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877949993472945e-05,
+      "loss": 1.1517,
+      "step": 1169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987761189387697e-05,
+      "loss": 1.1252,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877273329514685e-05,
+      "loss": 1.0867,
+      "step": 1171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876934300402013e-05,
+      "loss": 1.0987,
+      "step": 1172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876594806554915e-05,
+      "loss": 1.0699,
+      "step": 1173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876254847989365e-05,
+      "loss": 1.1095,
+      "step": 1174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875914424721353e-05,
+      "loss": 1.1092,
+      "step": 1175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875573536766902e-05,
+      "loss": 1.0298,
+      "step": 1176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987523218414205e-05,
+      "loss": 1.1105,
+      "step": 1177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987489036686286e-05,
+      "loss": 1.101,
+      "step": 1178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987454808494541e-05,
+      "loss": 0.9858,
+      "step": 1179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987420533840581e-05,
+      "loss": 1.183,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873862127260187e-05,
+      "loss": 1.0695,
+      "step": 1181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873518451524683e-05,
+      "loss": 1.0648,
+      "step": 1182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987317431121548e-05,
+      "loss": 1.1153,
+      "step": 1183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872829706348767e-05,
+      "loss": 1.1427,
+      "step": 1184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872484636940753e-05,
+      "loss": 1.1475,
+      "step": 1185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987213910300768e-05,
+      "loss": 1.0214,
+      "step": 1186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871793104565803e-05,
+      "loss": 1.1248,
+      "step": 1187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98714466416314e-05,
+      "loss": 1.1453,
+      "step": 1188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871099714220784e-05,
+      "loss": 1.0742,
+      "step": 1189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870752322350265e-05,
+      "loss": 1.0997,
+      "step": 1190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98704044660362e-05,
+      "loss": 1.1065,
+      "step": 1191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870056145294947e-05,
+      "loss": 1.104,
+      "step": 1192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98697073601429e-05,
+      "loss": 1.1193,
+      "step": 1193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869358110596474e-05,
+      "loss": 1.1326,
+      "step": 1194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869008396672096e-05,
+      "loss": 1.1341,
+      "step": 1195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868658218386222e-05,
+      "loss": 1.0761,
+      "step": 1196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986830757575533e-05,
+      "loss": 1.0698,
+      "step": 1197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867956468795917e-05,
+      "loss": 1.0967,
+      "step": 1198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867604897524507e-05,
+      "loss": 1.1357,
+      "step": 1199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986725286195764e-05,
+      "loss": 1.1103,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986690036211188e-05,
+      "loss": 1.1055,
+      "step": 1201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986654739800381e-05,
+      "loss": 1.055,
+      "step": 1202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866193969650044e-05,
+      "loss": 1.1792,
+      "step": 1203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986584007706721e-05,
+      "loss": 1.0506,
+      "step": 1204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865485720271957e-05,
+      "loss": 1.076,
+      "step": 1205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865130899280958e-05,
+      "loss": 1.0789,
+      "step": 1206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986477561411091e-05,
+      "loss": 1.1467,
+      "step": 1207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986441986477853e-05,
+      "loss": 1.1113,
+      "step": 1208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864063651300557e-05,
+      "loss": 1.1349,
+      "step": 1209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863706973693752e-05,
+      "loss": 1.1034,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98633498319749e-05,
+      "loss": 1.1009,
+      "step": 1211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98629922261608e-05,
+      "loss": 1.109,
+      "step": 1212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986263415626828e-05,
+      "loss": 1.0582,
+      "step": 1213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862275622314188e-05,
+      "loss": 1.1107,
+      "step": 1214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98619166243154e-05,
+      "loss": 1.13,
+      "step": 1215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861557162288798e-05,
+      "loss": 1.0494,
+      "step": 1216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861197236251303e-05,
+      "loss": 1.0566,
+      "step": 1217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860836846219843e-05,
+      "loss": 1.117,
+      "step": 1218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860475992211385e-05,
+      "loss": 1.0784,
+      "step": 1219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860114674242896e-05,
+      "loss": 1.0916,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985975289233139e-05,
+      "loss": 1.073,
+      "step": 1221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985939064649388e-05,
+      "loss": 1.0602,
+      "step": 1222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859027936747416e-05,
+      "loss": 1.1125,
+      "step": 1223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985866476310906e-05,
+      "loss": 1.0147,
+      "step": 1224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858301125595906e-05,
+      "loss": 1.0855,
+      "step": 1225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985793702422506e-05,
+      "loss": 1.106,
+      "step": 1226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857572459013654e-05,
+      "loss": 1.1221,
+      "step": 1227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857207429978843e-05,
+      "loss": 1.1564,
+      "step": 1228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856841937137803e-05,
+      "loss": 1.0748,
+      "step": 1229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985647598050773e-05,
+      "loss": 1.148,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985610956010584e-05,
+      "loss": 1.1001,
+      "step": 1231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985574267594938e-05,
+      "loss": 1.083,
+      "step": 1232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855375328055608e-05,
+      "loss": 1.0693,
+      "step": 1233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855007516441813e-05,
+      "loss": 1.1053,
+      "step": 1234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98546392411253e-05,
+      "loss": 1.1096,
+      "step": 1235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854270502123393e-05,
+      "loss": 1.0324,
+      "step": 1236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853901299453448e-05,
+      "loss": 1.0275,
+      "step": 1237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853531633132836e-05,
+      "loss": 1.0983,
+      "step": 1238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853161503178948e-05,
+      "loss": 1.0748,
+      "step": 1239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98527909096092e-05,
+      "loss": 1.076,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985241985244103e-05,
+      "loss": 1.083,
+      "step": 1241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98520483316919e-05,
+      "loss": 1.1041,
+      "step": 1242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851676347379287e-05,
+      "loss": 1.114,
+      "step": 1243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851303899520695e-05,
+      "loss": 1.1351,
+      "step": 1244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985093098813365e-05,
+      "loss": 1.0124,
+      "step": 1245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850557613235694e-05,
+      "loss": 1.1289,
+      "step": 1246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850183774844397e-05,
+      "loss": 1.0329,
+      "step": 1247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849809472977355e-05,
+      "loss": 1.0983,
+      "step": 1248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849434707652175e-05,
+      "loss": 1.1054,
+      "step": 1249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849059478886486e-05,
+      "loss": 1.077,
+      "step": 1250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848683786697952e-05,
+      "loss": 1.1019,
+      "step": 1251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984830763110425e-05,
+      "loss": 1.052,
+      "step": 1252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984793101212307e-05,
+      "loss": 1.0975,
+      "step": 1253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984755392977214e-05,
+      "loss": 1.1234,
+      "step": 1254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847176384069204e-05,
+      "loss": 1.0737,
+      "step": 1255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984679837503202e-05,
+      "loss": 1.0179,
+      "step": 1256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984641990267838e-05,
+      "loss": 1.1127,
+      "step": 1257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846040967026086e-05,
+      "loss": 1.1413,
+      "step": 1258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984566156809298e-05,
+      "loss": 1.1307,
+      "step": 1259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845281705896898e-05,
+      "loss": 1.1044,
+      "step": 1260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844901380455724e-05,
+      "loss": 1.0488,
+      "step": 1261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844520591787352e-05,
+      "loss": 1.0437,
+      "step": 1262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844139339909692e-05,
+      "loss": 1.0119,
+      "step": 1263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984375762484069e-05,
+      "loss": 1.1426,
+      "step": 1264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843375446598306e-05,
+      "loss": 1.1449,
+      "step": 1265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842992805200522e-05,
+      "loss": 1.0786,
+      "step": 1266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842609700665342e-05,
+      "loss": 1.1324,
+      "step": 1267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984222613301079e-05,
+      "loss": 1.1089,
+      "step": 1268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841842102254917e-05,
+      "loss": 1.0313,
+      "step": 1269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984145760841579e-05,
+      "loss": 1.0429,
+      "step": 1270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98410726515115e-05,
+      "loss": 1.0911,
+      "step": 1271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840687231560162e-05,
+      "loss": 1.2113,
+      "step": 1272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840301348579912e-05,
+      "loss": 1.1539,
+      "step": 1273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839915002588903e-05,
+      "loss": 1.0832,
+      "step": 1274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839528193605316e-05,
+      "loss": 1.1063,
+      "step": 1275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839140921647352e-05,
+      "loss": 1.0782,
+      "step": 1276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838753186733233e-05,
+      "loss": 1.1467,
+      "step": 1277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98383649888812e-05,
+      "loss": 1.0741,
+      "step": 1278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983797632810952e-05,
+      "loss": 1.0052,
+      "step": 1279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837587204436486e-05,
+      "loss": 1.0721,
+      "step": 1280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837197617880403e-05,
+      "loss": 1.0825,
+      "step": 1281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836807568459595e-05,
+      "loss": 1.0872,
+      "step": 1282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983641705619243e-05,
+      "loss": 1.0655,
+      "step": 1283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836026081097264e-05,
+      "loss": 0.9857,
+      "step": 1284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983563464319251e-05,
+      "loss": 1.0374,
+      "step": 1285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835242742496574e-05,
+      "loss": 1.0083,
+      "step": 1286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834850379027906e-05,
+      "loss": 1.0475,
+      "step": 1287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983445755280496e-05,
+      "loss": 1.0896,
+      "step": 1288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834064263846225e-05,
+      "loss": 1.1153,
+      "step": 1289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833670512170205e-05,
+      "loss": 1.028,
+      "step": 1290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833276297795422e-05,
+      "loss": 1.0825,
+      "step": 1291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832881620740427e-05,
+      "loss": 0.9953,
+      "step": 1292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832486481023796e-05,
+      "loss": 1.0174,
+      "step": 1293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832090878664117e-05,
+      "loss": 1.0292,
+      "step": 1294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831694813680004e-05,
+      "loss": 1.1424,
+      "step": 1295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983129828609009e-05,
+      "loss": 0.9975,
+      "step": 1296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983090129591304e-05,
+      "loss": 1.0436,
+      "step": 1297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830503843167527e-05,
+      "loss": 1.1184,
+      "step": 1298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830105927872253e-05,
+      "loss": 1.0923,
+      "step": 1299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829707550045944e-05,
+      "loss": 1.1162,
+      "step": 1300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829308709707343e-05,
+      "loss": 1.0897,
+      "step": 1301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828909406875218e-05,
+      "loss": 1.115,
+      "step": 1302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828509641568355e-05,
+      "loss": 1.0329,
+      "step": 1303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982810941380556e-05,
+      "loss": 1.0362,
+      "step": 1304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827708723605673e-05,
+      "loss": 1.05,
+      "step": 1305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827307570987544e-05,
+      "loss": 1.0338,
+      "step": 1306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826905955970046e-05,
+      "loss": 1.0416,
+      "step": 1307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982650387857208e-05,
+      "loss": 1.1002,
+      "step": 1308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982610133881256e-05,
+      "loss": 1.0162,
+      "step": 1309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825698336710433e-05,
+      "loss": 1.0842,
+      "step": 1310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825294872284655e-05,
+      "loss": 1.1228,
+      "step": 1311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824890945554212e-05,
+      "loss": 1.1392,
+      "step": 1312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982448655653811e-05,
+      "loss": 1.0636,
+      "step": 1313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824081705255377e-05,
+      "loss": 1.1113,
+      "step": 1314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823676391725063e-05,
+      "loss": 1.054,
+      "step": 1315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982327061596624e-05,
+      "loss": 1.1199,
+      "step": 1316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822864377997995e-05,
+      "loss": 1.1476,
+      "step": 1317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982245767783945e-05,
+      "loss": 1.1467,
+      "step": 1318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982205051550973e-05,
+      "loss": 1.1031,
+      "step": 1319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821642891028006e-05,
+      "loss": 1.0618,
+      "step": 1320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821234804413453e-05,
+      "loss": 1.0986,
+      "step": 1321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982082625568527e-05,
+      "loss": 1.1476,
+      "step": 1322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820417244862684e-05,
+      "loss": 1.1318,
+      "step": 1323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820007771964936e-05,
+      "loss": 1.1146,
+      "step": 1324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819597837011295e-05,
+      "loss": 1.1129,
+      "step": 1325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981918744002105e-05,
+      "loss": 1.0374,
+      "step": 1326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818776581013507e-05,
+      "loss": 1.1149,
+      "step": 1327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981836526000801e-05,
+      "loss": 1.1306,
+      "step": 1328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817953477023894e-05,
+      "loss": 1.068,
+      "step": 1329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981754123208055e-05,
+      "loss": 1.0125,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817128525197365e-05,
+      "loss": 1.1151,
+      "step": 1331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981671535639377e-05,
+      "loss": 1.0523,
+      "step": 1332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981630172568919e-05,
+      "loss": 1.0863,
+      "step": 1333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98158876331031e-05,
+      "loss": 1.0,
+      "step": 1334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815473078654978e-05,
+      "loss": 1.1287,
+      "step": 1335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815058062364333e-05,
+      "loss": 1.0526,
+      "step": 1336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981464258425069e-05,
+      "loss": 0.9749,
+      "step": 1337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814226644333598e-05,
+      "loss": 1.0875,
+      "step": 1338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981381024263263e-05,
+      "loss": 1.0967,
+      "step": 1339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813393379167377e-05,
+      "loss": 1.0745,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812976053957458e-05,
+      "loss": 1.0548,
+      "step": 1341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812558267022504e-05,
+      "loss": 0.9865,
+      "step": 1342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812140018382172e-05,
+      "loss": 1.1507,
+      "step": 1343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811721308056145e-05,
+      "loss": 1.0226,
+      "step": 1344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811302136064127e-05,
+      "loss": 1.057,
+      "step": 1345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810882502425834e-05,
+      "loss": 1.1003,
+      "step": 1346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810462407161015e-05,
+      "loss": 1.1402,
+      "step": 1347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810041850289434e-05,
+      "loss": 1.0888,
+      "step": 1348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809620831830883e-05,
+      "loss": 1.1134,
+      "step": 1349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809199351805174e-05,
+      "loss": 1.0844,
+      "step": 1350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980877741023213e-05,
+      "loss": 1.135,
+      "step": 1351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808355007131613e-05,
+      "loss": 1.0504,
+      "step": 1352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980793214252349e-05,
+      "loss": 1.0797,
+      "step": 1353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807508816427666e-05,
+      "loss": 1.0891,
+      "step": 1354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980708502886405e-05,
+      "loss": 1.1774,
+      "step": 1355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806660779852594e-05,
+      "loss": 1.0059,
+      "step": 1356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806236069413254e-05,
+      "loss": 1.0742,
+      "step": 1357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980581089756601e-05,
+      "loss": 1.1392,
+      "step": 1358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805385264330876e-05,
+      "loss": 0.9721,
+      "step": 1359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804959169727868e-05,
+      "loss": 1.2093,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804532613777045e-05,
+      "loss": 1.1981,
+      "step": 1361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804105596498473e-05,
+      "loss": 1.0244,
+      "step": 1362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803678117912243e-05,
+      "loss": 1.012,
+      "step": 1363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803250178038476e-05,
+      "loss": 1.088,
+      "step": 1364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802821776897295e-05,
+      "loss": 1.0356,
+      "step": 1365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980239291450887e-05,
+      "loss": 1.1171,
+      "step": 1366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801963590893372e-05,
+      "loss": 1.1292,
+      "step": 1367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801533806071007e-05,
+      "loss": 1.082,
+      "step": 1368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801103560061994e-05,
+      "loss": 1.1321,
+      "step": 1369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800672852886577e-05,
+      "loss": 1.176,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800241684565024e-05,
+      "loss": 1.1223,
+      "step": 1371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979981005511762e-05,
+      "loss": 1.1056,
+      "step": 1372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799377964564677e-05,
+      "loss": 1.0434,
+      "step": 1373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798945412926523e-05,
+      "loss": 1.1646,
+      "step": 1374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798512400223515e-05,
+      "loss": 1.0537,
+      "step": 1375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798078926476024e-05,
+      "loss": 1.0823,
+      "step": 1376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797644991704445e-05,
+      "loss": 1.1167,
+      "step": 1377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797210595929198e-05,
+      "loss": 1.0266,
+      "step": 1378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796775739170723e-05,
+      "loss": 0.9982,
+      "step": 1379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979634042144948e-05,
+      "loss": 1.0579,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979590464278595e-05,
+      "loss": 1.1181,
+      "step": 1381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979546840320064e-05,
+      "loss": 1.0724,
+      "step": 1382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795031702714075e-05,
+      "loss": 1.0379,
+      "step": 1383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794594541346802e-05,
+      "loss": 1.0843,
+      "step": 1384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794156919119395e-05,
+      "loss": 1.0621,
+      "step": 1385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793718836052438e-05,
+      "loss": 1.1196,
+      "step": 1386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793280292166554e-05,
+      "loss": 1.0739,
+      "step": 1387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792841287482366e-05,
+      "loss": 1.089,
+      "step": 1388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792401822020535e-05,
+      "loss": 1.0747,
+      "step": 1389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791961895801743e-05,
+      "loss": 1.1062,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791521508846684e-05,
+      "loss": 1.127,
+      "step": 1391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979108066117608e-05,
+      "loss": 1.0505,
+      "step": 1392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979063935281068e-05,
+      "loss": 1.1152,
+      "step": 1393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979019758377124e-05,
+      "loss": 1.1028,
+      "step": 1394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789755354078554e-05,
+      "loss": 1.047,
+      "step": 1395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789312663753422e-05,
+      "loss": 1.0765,
+      "step": 1396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788869512816682e-05,
+      "loss": 1.101,
+      "step": 1397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788425901289182e-05,
+      "loss": 1.0641,
+      "step": 1398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787981829191793e-05,
+      "loss": 1.0782,
+      "step": 1399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978753729654541e-05,
+      "loss": 1.0859,
+      "step": 1400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787092303370953e-05,
+      "loss": 1.0486,
+      "step": 1401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786646849689358e-05,
+      "loss": 1.0357,
+      "step": 1402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978620093552158e-05,
+      "loss": 1.0762,
+      "step": 1403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978575456088861e-05,
+      "loss": 1.1249,
+      "step": 1404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978530772581144e-05,
+      "loss": 1.1012,
+      "step": 1405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784860430311106e-05,
+      "loss": 1.1169,
+      "step": 1406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978441267440865e-05,
+      "loss": 1.0831,
+      "step": 1407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783964458125132e-05,
+      "loss": 1.0551,
+      "step": 1408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783515781481656e-05,
+      "loss": 1.1009,
+      "step": 1409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783066644499323e-05,
+      "loss": 1.1816,
+      "step": 1410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782617047199267e-05,
+      "loss": 1.0558,
+      "step": 1411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978216698960265e-05,
+      "loss": 1.0266,
+      "step": 1412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781716471730637e-05,
+      "loss": 1.0151,
+      "step": 1413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781265493604436e-05,
+      "loss": 1.1822,
+      "step": 1414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978081405524526e-05,
+      "loss": 1.0901,
+      "step": 1415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780362156674357e-05,
+      "loss": 1.0717,
+      "step": 1416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779909797912983e-05,
+      "loss": 1.0374,
+      "step": 1417
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779456978982424e-05,
+      "loss": 1.0663,
+      "step": 1418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779003699903988e-05,
+      "loss": 1.114,
+      "step": 1419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778549960699007e-05,
+      "loss": 1.1543,
+      "step": 1420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778095761388822e-05,
+      "loss": 1.0629,
+      "step": 1421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977764110199481e-05,
+      "loss": 1.008,
+      "step": 1422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777185982538364e-05,
+      "loss": 1.1294,
+      "step": 1423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776730403040894e-05,
+      "loss": 1.0814,
+      "step": 1424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776274363523842e-05,
+      "loss": 1.0227,
+      "step": 1425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977581786400866e-05,
+      "loss": 1.0529,
+      "step": 1426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977536090451683e-05,
+      "loss": 1.059,
+      "step": 1427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774903485069857e-05,
+      "loss": 1.0843,
+      "step": 1428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774445605689254e-05,
+      "loss": 1.0579,
+      "step": 1429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773987266396577e-05,
+      "loss": 1.09,
+      "step": 1430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977352846721338e-05,
+      "loss": 1.0636,
+      "step": 1431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773069208161265e-05,
+      "loss": 0.9934,
+      "step": 1432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772609489261828e-05,
+      "loss": 1.0562,
+      "step": 1433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977214931053671e-05,
+      "loss": 1.0793,
+      "step": 1434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771688672007558e-05,
+      "loss": 1.0424,
+      "step": 1435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771227573696044e-05,
+      "loss": 1.0748,
+      "step": 1436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770766015623866e-05,
+      "loss": 1.0807,
+      "step": 1437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770303997812744e-05,
+      "loss": 1.072,
+      "step": 1438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769841520284418e-05,
+      "loss": 1.1002,
+      "step": 1439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769378583060642e-05,
+      "loss": 1.0316,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768915186163206e-05,
+      "loss": 1.1459,
+      "step": 1441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768451329613908e-05,
+      "loss": 1.1195,
+      "step": 1442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976798701343458e-05,
+      "loss": 0.9978,
+      "step": 1443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976752223764706e-05,
+      "loss": 1.1569,
+      "step": 1444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767057002273225e-05,
+      "loss": 1.0482,
+      "step": 1445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976659130733496e-05,
+      "loss": 1.0368,
+      "step": 1446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766125152854182e-05,
+      "loss": 1.0802,
+      "step": 1447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765658538852825e-05,
+      "loss": 1.1559,
+      "step": 1448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765191465352837e-05,
+      "loss": 1.0113,
+      "step": 1449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764723932376203e-05,
+      "loss": 1.023,
+      "step": 1450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976425593994492e-05,
+      "loss": 1.1449,
+      "step": 1451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763787488081e-05,
+      "loss": 1.0581,
+      "step": 1452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97633185768065e-05,
+      "loss": 1.1044,
+      "step": 1453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976284920614347e-05,
+      "loss": 1.0712,
+      "step": 1454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762379376114e-05,
+      "loss": 1.138,
+      "step": 1455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761909086740195e-05,
+      "loss": 1.1363,
+      "step": 1456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976143833804419e-05,
+      "loss": 1.0658,
+      "step": 1457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760967130048125e-05,
+      "loss": 1.0212,
+      "step": 1458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976049546277418e-05,
+      "loss": 1.0688,
+      "step": 1459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976002333624454e-05,
+      "loss": 1.1288,
+      "step": 1460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975955075048143e-05,
+      "loss": 1.0501,
+      "step": 1461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759077705507076e-05,
+      "loss": 1.1042,
+      "step": 1462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975860420134374e-05,
+      "loss": 1.0338,
+      "step": 1463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975813023801371e-05,
+      "loss": 1.1019,
+      "step": 1464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975765581553927e-05,
+      "loss": 1.1555,
+      "step": 1465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757180933942756e-05,
+      "loss": 1.1039,
+      "step": 1466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756705593246508e-05,
+      "loss": 1.0852,
+      "step": 1467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975622979347289e-05,
+      "loss": 1.0861,
+      "step": 1468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755753534644296e-05,
+      "loss": 0.9948,
+      "step": 1469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755276816783132e-05,
+      "loss": 1.0441,
+      "step": 1470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754799639911823e-05,
+      "loss": 1.0041,
+      "step": 1471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754322004052827e-05,
+      "loss": 1.0768,
+      "step": 1472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975384390922862e-05,
+      "loss": 1.1452,
+      "step": 1473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753365355461693e-05,
+      "loss": 1.0431,
+      "step": 1474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752886342774566e-05,
+      "loss": 1.0609,
+      "step": 1475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752406871189772e-05,
+      "loss": 1.1512,
+      "step": 1476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975192694072988e-05,
+      "loss": 1.065,
+      "step": 1477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751446551417467e-05,
+      "loss": 1.1383,
+      "step": 1478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975096570327514e-05,
+      "loss": 1.1077,
+      "step": 1479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975048439632552e-05,
+      "loss": 1.0281,
+      "step": 1480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750002630591253e-05,
+      "loss": 1.0994,
+      "step": 1481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974952040609501e-05,
+      "loss": 1.1614,
+      "step": 1482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974903772285948e-05,
+      "loss": 1.1644,
+      "step": 1483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974855458090738e-05,
+      "loss": 1.0551,
+      "step": 1484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974807098026143e-05,
+      "loss": 1.1364,
+      "step": 1485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747586920944397e-05,
+      "loss": 1.0537,
+      "step": 1486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747102402979053e-05,
+      "loss": 1.0817,
+      "step": 1487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974661742638819e-05,
+      "loss": 1.0555,
+      "step": 1488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974613199119464e-05,
+      "loss": 1.0466,
+      "step": 1489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745646097421232e-05,
+      "loss": 1.1246,
+      "step": 1490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745159745090838e-05,
+      "loss": 1.1313,
+      "step": 1491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744672934226333e-05,
+      "loss": 1.0236,
+      "step": 1492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744185664850626e-05,
+      "loss": 1.104,
+      "step": 1493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974369793698665e-05,
+      "loss": 1.0228,
+      "step": 1494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743209750657348e-05,
+      "loss": 1.1246,
+      "step": 1495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974272110588569e-05,
+      "loss": 1.0345,
+      "step": 1496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742232002694666e-05,
+      "loss": 1.1455,
+      "step": 1497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741742441107297e-05,
+      "loss": 1.0864,
+      "step": 1498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741252421146613e-05,
+      "loss": 1.1232,
+      "step": 1499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740761942835673e-05,
+      "loss": 1.1114,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974027100619755e-05,
+      "loss": 1.0627,
+      "step": 1501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973977961125535e-05,
+      "loss": 1.0388,
+      "step": 1502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739287758032194e-05,
+      "loss": 1.0829,
+      "step": 1503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973879544655122e-05,
+      "loss": 1.1062,
+      "step": 1504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738302676835598e-05,
+      "loss": 1.1013,
+      "step": 1505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737809448908513e-05,
+      "loss": 1.0573,
+      "step": 1506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737315762793165e-05,
+      "loss": 1.0687,
+      "step": 1507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736821618512794e-05,
+      "loss": 1.0382,
+      "step": 1508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736327016090644e-05,
+      "loss": 1.1186,
+      "step": 1509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973583195554999e-05,
+      "loss": 1.0559,
+      "step": 1510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735336436914125e-05,
+      "loss": 1.0316,
+      "step": 1511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734840460206365e-05,
+      "loss": 1.1096,
+      "step": 1512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734344025450046e-05,
+      "loss": 1.0736,
+      "step": 1513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733847132668527e-05,
+      "loss": 1.1009,
+      "step": 1514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733349781885187e-05,
+      "loss": 1.1757,
+      "step": 1515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732851973123432e-05,
+      "loss": 1.0251,
+      "step": 1516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732353706406676e-05,
+      "loss": 1.1766,
+      "step": 1517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731854981758373e-05,
+      "loss": 1.0715,
+      "step": 1518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731355799201985e-05,
+      "loss": 1.0541,
+      "step": 1519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730856158761002e-05,
+      "loss": 1.1439,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973035606045893e-05,
+      "loss": 1.0998,
+      "step": 1521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729855504319303e-05,
+      "loss": 1.0222,
+      "step": 1522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972935449036567e-05,
+      "loss": 1.0305,
+      "step": 1523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972885301862161e-05,
+      "loss": 1.0888,
+      "step": 1524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728351089110714e-05,
+      "loss": 1.1274,
+      "step": 1525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97278487018566e-05,
+      "loss": 1.1056,
+      "step": 1526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727345856882907e-05,
+      "loss": 1.0554,
+      "step": 1527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726842554213295e-05,
+      "loss": 1.0421,
+      "step": 1528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726338793871447e-05,
+      "loss": 1.0133,
+      "step": 1529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725834575881065e-05,
+      "loss": 1.0843,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725329900265873e-05,
+      "loss": 1.0998,
+      "step": 1531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972482476704962e-05,
+      "loss": 1.1715,
+      "step": 1532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972431917625607e-05,
+      "loss": 1.0331,
+      "step": 1533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723813127909014e-05,
+      "loss": 1.1025,
+      "step": 1534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723306622032266e-05,
+      "loss": 1.0784,
+      "step": 1535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972279965864965e-05,
+      "loss": 1.1496,
+      "step": 1536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972229223778503e-05,
+      "loss": 1.065,
+      "step": 1537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972178435946227e-05,
+      "loss": 1.0575,
+      "step": 1538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721276023705282e-05,
+      "loss": 1.0447,
+      "step": 1539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720767230537973e-05,
+      "loss": 1.1493,
+      "step": 1540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720257979984286e-05,
+      "loss": 0.9972,
+      "step": 1541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719748272068184e-05,
+      "loss": 1.0475,
+      "step": 1542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719238106813643e-05,
+      "loss": 1.067,
+      "step": 1543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971872748424468e-05,
+      "loss": 1.0266,
+      "step": 1544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718216404385307e-05,
+      "loss": 1.0998,
+      "step": 1545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717704867259583e-05,
+      "loss": 1.0938,
+      "step": 1546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717192872891572e-05,
+      "loss": 1.0594,
+      "step": 1547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716680421305368e-05,
+      "loss": 1.002,
+      "step": 1548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971616751252508e-05,
+      "loss": 1.1033,
+      "step": 1549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971565414657484e-05,
+      "loss": 1.1014,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715140323478806e-05,
+      "loss": 1.1498,
+      "step": 1551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714626043261154e-05,
+      "loss": 1.0313,
+      "step": 1552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714111305946083e-05,
+      "loss": 1.0424,
+      "step": 1553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713596111557813e-05,
+      "loss": 1.0191,
+      "step": 1554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971308046012058e-05,
+      "loss": 1.1379,
+      "step": 1555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712564351658657e-05,
+      "loss": 1.1464,
+      "step": 1556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971204778619632e-05,
+      "loss": 1.0887,
+      "step": 1557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711530763757873e-05,
+      "loss": 1.084,
+      "step": 1558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971101328436765e-05,
+      "loss": 0.9706,
+      "step": 1559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710495348049995e-05,
+      "loss": 1.1151,
+      "step": 1560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970997695482928e-05,
+      "loss": 1.097,
+      "step": 1561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709458104729898e-05,
+      "loss": 1.0811,
+      "step": 1562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708938797776258e-05,
+      "loss": 1.1022,
+      "step": 1563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708419033992802e-05,
+      "loss": 1.0027,
+      "step": 1564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707898813403977e-05,
+      "loss": 1.0333,
+      "step": 1565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707378136034267e-05,
+      "loss": 1.147,
+      "step": 1566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706857001908167e-05,
+      "loss": 1.0382,
+      "step": 1567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706335411050202e-05,
+      "loss": 1.0832,
+      "step": 1568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705813363484913e-05,
+      "loss": 1.0599,
+      "step": 1569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705290859236864e-05,
+      "loss": 1.0705,
+      "step": 1570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704767898330636e-05,
+      "loss": 1.085,
+      "step": 1571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704244480790843e-05,
+      "loss": 1.1017,
+      "step": 1572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703720606642104e-05,
+      "loss": 1.0813,
+      "step": 1573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703196275909074e-05,
+      "loss": 1.0888,
+      "step": 1574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702671488616425e-05,
+      "loss": 1.0391,
+      "step": 1575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702146244788847e-05,
+      "loss": 1.1059,
+      "step": 1576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701620544451053e-05,
+      "loss": 1.0534,
+      "step": 1577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701094387627783e-05,
+      "loss": 0.995,
+      "step": 1578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700567774343794e-05,
+      "loss": 1.1212,
+      "step": 1579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700040704623856e-05,
+      "loss": 1.1186,
+      "step": 1580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969951317849278e-05,
+      "loss": 1.0044,
+      "step": 1581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969898519597538e-05,
+      "loss": 1.0796,
+      "step": 1582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96984567570965e-05,
+      "loss": 1.0263,
+      "step": 1583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697927861881005e-05,
+      "loss": 1.1876,
+      "step": 1584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697398510353783e-05,
+      "loss": 1.06,
+      "step": 1585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969686870253974e-05,
+      "loss": 1.045,
+      "step": 1586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696338438463804e-05,
+      "loss": 1.0436,
+      "step": 1587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969580771815093e-05,
+      "loss": 1.0693,
+      "step": 1588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969527654162608e-05,
+      "loss": 1.0351,
+      "step": 1589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694744908914254e-05,
+      "loss": 1.0569,
+      "step": 1590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694212820040465e-05,
+      "loss": 1.1445,
+      "step": 1591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969368027502975e-05,
+      "loss": 1.0541,
+      "step": 1592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693147273907167e-05,
+      "loss": 1.1446,
+      "step": 1593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969261381669779e-05,
+      "loss": 1.0566,
+      "step": 1594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692079903426727e-05,
+      "loss": 1.1447,
+      "step": 1595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691545534119094e-05,
+      "loss": 1.0411,
+      "step": 1596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969101070880004e-05,
+      "loss": 1.1172,
+      "step": 1597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690475427494725e-05,
+      "loss": 1.1379,
+      "step": 1598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968993969022834e-05,
+      "loss": 1.0779,
+      "step": 1599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689403497026085e-05,
+      "loss": 1.093,
+      "step": 1600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688866847913198e-05,
+      "loss": 1.0811,
+      "step": 1601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968832974291492e-05,
+      "loss": 1.0938,
+      "step": 1602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687792182056535e-05,
+      "loss": 1.1191,
+      "step": 1603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687254165363326e-05,
+      "loss": 1.0456,
+      "step": 1604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686715692860615e-05,
+      "loss": 1.0426,
+      "step": 1605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686176764573737e-05,
+      "loss": 1.0194,
+      "step": 1606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968563738052805e-05,
+      "loss": 1.1561,
+      "step": 1607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685097540748928e-05,
+      "loss": 1.0841,
+      "step": 1608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684557245261778e-05,
+      "loss": 1.0431,
+      "step": 1609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684016494092018e-05,
+      "loss": 1.1523,
+      "step": 1610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683475287265095e-05,
+      "loss": 1.1053,
+      "step": 1611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682933624806475e-05,
+      "loss": 1.1785,
+      "step": 1612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682391506741644e-05,
+      "loss": 1.0169,
+      "step": 1613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681848933096106e-05,
+      "loss": 1.0236,
+      "step": 1614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681305903895395e-05,
+      "loss": 1.1271,
+      "step": 1615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680762419165054e-05,
+      "loss": 1.0469,
+      "step": 1616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680218478930668e-05,
+      "loss": 1.0446,
+      "step": 1617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679674083217822e-05,
+      "loss": 1.0631,
+      "step": 1618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679129232052135e-05,
+      "loss": 1.0951,
+      "step": 1619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678583925459242e-05,
+      "loss": 1.0027,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96780381634648e-05,
+      "loss": 1.1071,
+      "step": 1621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967749194609449e-05,
+      "loss": 1.0761,
+      "step": 1622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967694527337401e-05,
+      "loss": 1.1285,
+      "step": 1623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676398145329083e-05,
+      "loss": 1.1163,
+      "step": 1624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675850561985458e-05,
+      "loss": 1.0421,
+      "step": 1625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967530252336889e-05,
+      "loss": 1.0753,
+      "step": 1626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674754029505184e-05,
+      "loss": 1.1373,
+      "step": 1627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967420508042013e-05,
+      "loss": 1.0783,
+      "step": 1628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967365567613956e-05,
+      "loss": 1.1229,
+      "step": 1629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967310581668933e-05,
+      "loss": 1.0917,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672555502095312e-05,
+      "loss": 1.1162,
+      "step": 1631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96720047323834e-05,
+      "loss": 1.0529,
+      "step": 1632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671453507579504e-05,
+      "loss": 1.0702,
+      "step": 1633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967090182770957e-05,
+      "loss": 1.03,
+      "step": 1634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967034969279954e-05,
+      "loss": 1.0447,
+      "step": 1635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669797102875412e-05,
+      "loss": 1.0682,
+      "step": 1636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966924405796317e-05,
+      "loss": 1.1024,
+      "step": 1637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668690558088853e-05,
+      "loss": 1.1151,
+      "step": 1638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668136603278492e-05,
+      "loss": 1.0951,
+      "step": 1639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667582193558154e-05,
+      "loss": 1.0877,
+      "step": 1640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667027328953926e-05,
+      "loss": 1.07,
+      "step": 1641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666472009491918e-05,
+      "loss": 1.1009,
+      "step": 1642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966591623519826e-05,
+      "loss": 1.1222,
+      "step": 1643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96653600060991e-05,
+      "loss": 1.0653,
+      "step": 1644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664803322220603e-05,
+      "loss": 1.0963,
+      "step": 1645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664246183588977e-05,
+      "loss": 1.0888,
+      "step": 1646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663688590230427e-05,
+      "loss": 1.0544,
+      "step": 1647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663130542171194e-05,
+      "loss": 0.9969,
+      "step": 1648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966257203943753e-05,
+      "loss": 1.1414,
+      "step": 1649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966201308205572e-05,
+      "loss": 1.0558,
+      "step": 1650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966145367005206e-05,
+      "loss": 1.0349,
+      "step": 1651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660893803452873e-05,
+      "loss": 1.0553,
+      "step": 1652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660333482284503e-05,
+      "loss": 1.0904,
+      "step": 1653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659772706573314e-05,
+      "loss": 1.0523,
+      "step": 1654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965921147634569e-05,
+      "loss": 1.0597,
+      "step": 1655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965864979162804e-05,
+      "loss": 1.0474,
+      "step": 1656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658087652446798e-05,
+      "loss": 1.0263,
+      "step": 1657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657525058828404e-05,
+      "loss": 1.0667,
+      "step": 1658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656962010799336e-05,
+      "loss": 1.0113,
+      "step": 1659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965639850838608e-05,
+      "loss": 1.0383,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655834551615162e-05,
+      "loss": 1.1313,
+      "step": 1661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965527014051311e-05,
+      "loss": 1.0671,
+      "step": 1662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654705275106476e-05,
+      "loss": 1.0442,
+      "step": 1663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965413995542185e-05,
+      "loss": 1.0269,
+      "step": 1664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965357418148582e-05,
+      "loss": 0.9978,
+      "step": 1665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653007953325017e-05,
+      "loss": 1.1647,
+      "step": 1666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652441270966084e-05,
+      "loss": 1.0634,
+      "step": 1667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651874134435675e-05,
+      "loss": 1.0782,
+      "step": 1668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651306543760477e-05,
+      "loss": 1.0797,
+      "step": 1669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650738498967203e-05,
+      "loss": 1.0288,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650170000082577e-05,
+      "loss": 1.0513,
+      "step": 1671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964960104713335e-05,
+      "loss": 1.0858,
+      "step": 1672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964903164014629e-05,
+      "loss": 1.0692,
+      "step": 1673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648461779148196e-05,
+      "loss": 1.0697,
+      "step": 1674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964789146416587e-05,
+      "loss": 1.0836,
+      "step": 1675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647320695226156e-05,
+      "loss": 1.0985,
+      "step": 1676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646749472355906e-05,
+      "loss": 1.0922,
+      "step": 1677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646177795582e-05,
+      "loss": 1.0189,
+      "step": 1678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645605664931334e-05,
+      "loss": 1.0992,
+      "step": 1679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964503308043083e-05,
+      "loss": 1.0485,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644460042107426e-05,
+      "loss": 1.0727,
+      "step": 1681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643886549988093e-05,
+      "loss": 1.0957,
+      "step": 1682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643312604099807e-05,
+      "loss": 1.0475,
+      "step": 1683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642738204469577e-05,
+      "loss": 1.1656,
+      "step": 1684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964216335112443e-05,
+      "loss": 1.0714,
+      "step": 1685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641588044091415e-05,
+      "loss": 1.0495,
+      "step": 1686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641012283397597e-05,
+      "loss": 1.072,
+      "step": 1687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640436069070073e-05,
+      "loss": 1.0263,
+      "step": 1688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963985940113595e-05,
+      "loss": 1.0921,
+      "step": 1689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963928227962237e-05,
+      "loss": 1.0871,
+      "step": 1690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963870470455648e-05,
+      "loss": 1.1227,
+      "step": 1691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638126675965457e-05,
+      "loss": 1.026,
+      "step": 1692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96375481938765e-05,
+      "loss": 1.0358,
+      "step": 1693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636969258316833e-05,
+      "loss": 1.0901,
+      "step": 1694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963638986931369e-05,
+      "loss": 0.9871,
+      "step": 1695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635810026894333e-05,
+      "loss": 1.0479,
+      "step": 1696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635229731086047e-05,
+      "loss": 1.1565,
+      "step": 1697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963464898191614e-05,
+      "loss": 1.0667,
+      "step": 1698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963406777941193e-05,
+      "loss": 1.0037,
+      "step": 1699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633486123600768e-05,
+      "loss": 1.0191,
+      "step": 1700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963290401451002e-05,
+      "loss": 1.1415,
+      "step": 1701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963232145216708e-05,
+      "loss": 1.0797,
+      "step": 1702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631738436599363e-05,
+      "loss": 1.1166,
+      "step": 1703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631154967834286e-05,
+      "loss": 1.0832,
+      "step": 1704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630571045899315e-05,
+      "loss": 1.1083,
+      "step": 1705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962998667082192e-05,
+      "loss": 0.9734,
+      "step": 1706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629401842629604e-05,
+      "loss": 1.0861,
+      "step": 1707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628816561349877e-05,
+      "loss": 1.0603,
+      "step": 1708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628230827010283e-05,
+      "loss": 1.0205,
+      "step": 1709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627644639638375e-05,
+      "loss": 1.0921,
+      "step": 1710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627057999261742e-05,
+      "loss": 1.0818,
+      "step": 1711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626470905907985e-05,
+      "loss": 1.0766,
+      "step": 1712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625883359604726e-05,
+      "loss": 1.0702,
+      "step": 1713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625295360379613e-05,
+      "loss": 1.0459,
+      "step": 1714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624706908260312e-05,
+      "loss": 1.081,
+      "step": 1715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624118003274514e-05,
+      "loss": 1.1335,
+      "step": 1716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9623528645449923e-05,
+      "loss": 0.9964,
+      "step": 1717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622938834814272e-05,
+      "loss": 1.1133,
+      "step": 1718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622348571395312e-05,
+      "loss": 1.0467,
+      "step": 1719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962175785522082e-05,
+      "loss": 1.0267,
+      "step": 1720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621166686318588e-05,
+      "loss": 1.0566,
+      "step": 1721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620575064716432e-05,
+      "loss": 1.1464,
+      "step": 1722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961998299044219e-05,
+      "loss": 1.1281,
+      "step": 1723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619390463523723e-05,
+      "loss": 1.1216,
+      "step": 1724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618797483988908e-05,
+      "loss": 1.1429,
+      "step": 1725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618204051865642e-05,
+      "loss": 1.0299,
+      "step": 1726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617610167181855e-05,
+      "loss": 1.135,
+      "step": 1727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617015829965487e-05,
+      "loss": 1.0742,
+      "step": 1728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96164210402445e-05,
+      "loss": 1.0214,
+      "step": 1729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961582579804689e-05,
+      "loss": 1.1548,
+      "step": 1730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615230103400653e-05,
+      "loss": 1.1387,
+      "step": 1731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614633956333825e-05,
+      "loss": 0.9947,
+      "step": 1732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614037356874456e-05,
+      "loss": 1.0577,
+      "step": 1733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961344030505062e-05,
+      "loss": 0.9715,
+      "step": 1734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96128428008904e-05,
+      "loss": 1.0701,
+      "step": 1735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612244844421916e-05,
+      "loss": 1.0683,
+      "step": 1736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611646435673303e-05,
+      "loss": 1.0254,
+      "step": 1737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611047574672723e-05,
+      "loss": 1.0692,
+      "step": 1738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610448261448346e-05,
+      "loss": 1.0608,
+      "step": 1739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609848496028373e-05,
+      "loss": 1.0777,
+      "step": 1740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609248278441027e-05,
+      "loss": 1.077,
+      "step": 1741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608647608714548e-05,
+      "loss": 1.0417,
+      "step": 1742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608046486877203e-05,
+      "loss": 1.0744,
+      "step": 1743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607444912957266e-05,
+      "loss": 1.0129,
+      "step": 1744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960684288698305e-05,
+      "loss": 1.2067,
+      "step": 1745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606240408982887e-05,
+      "loss": 1.1455,
+      "step": 1746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605637478985113e-05,
+      "loss": 1.0898,
+      "step": 1747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605034097018106e-05,
+      "loss": 1.1236,
+      "step": 1748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604430263110253e-05,
+      "loss": 1.0698,
+      "step": 1749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603825977289967e-05,
+      "loss": 1.0912,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960322123958568e-05,
+      "loss": 1.1146,
+      "step": 1751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602616050025852e-05,
+      "loss": 1.0837,
+      "step": 1752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960201040863895e-05,
+      "loss": 1.0405,
+      "step": 1753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601404315453477e-05,
+      "loss": 1.1536,
+      "step": 1754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960079777049795e-05,
+      "loss": 1.0464,
+      "step": 1755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600190773800903e-05,
+      "loss": 1.0722,
+      "step": 1756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599583325390906e-05,
+      "loss": 0.9364,
+      "step": 1757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598975425296535e-05,
+      "loss": 1.1243,
+      "step": 1758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598367073546395e-05,
+      "loss": 1.0164,
+      "step": 1759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597758270169113e-05,
+      "loss": 1.1548,
+      "step": 1760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597149015193328e-05,
+      "loss": 1.0463,
+      "step": 1761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596539308647714e-05,
+      "loss": 1.0688,
+      "step": 1762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595929150560953e-05,
+      "loss": 1.0316,
+      "step": 1763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595318540961757e-05,
+      "loss": 1.0605,
+      "step": 1764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594707479878862e-05,
+      "loss": 1.0917,
+      "step": 1765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959409596734101e-05,
+      "loss": 1.0583,
+      "step": 1766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593484003376982e-05,
+      "loss": 1.0642,
+      "step": 1767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959287158801557e-05,
+      "loss": 0.9954,
+      "step": 1768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592258721285587e-05,
+      "loss": 1.0156,
+      "step": 1769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591645403215876e-05,
+      "loss": 1.1237,
+      "step": 1770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959103163383529e-05,
+      "loss": 1.0485,
+      "step": 1771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959041741317271e-05,
+      "loss": 1.114,
+      "step": 1772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589802741257037e-05,
+      "loss": 1.1049,
+      "step": 1773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958918761811719e-05,
+      "loss": 1.0751,
+      "step": 1774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588572043782114e-05,
+      "loss": 1.0095,
+      "step": 1775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958795601828078e-05,
+      "loss": 1.1135,
+      "step": 1776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587339541642162e-05,
+      "loss": 0.935,
+      "step": 1777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586722613895274e-05,
+      "loss": 1.1125,
+      "step": 1778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586105235069138e-05,
+      "loss": 1.0377,
+      "step": 1779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958548740519281e-05,
+      "loss": 1.1061,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958486912429536e-05,
+      "loss": 1.0679,
+      "step": 1781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584250392405873e-05,
+      "loss": 1.0937,
+      "step": 1782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583631209553473e-05,
+      "loss": 1.0622,
+      "step": 1783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583011575767284e-05,
+      "loss": 1.0988,
+      "step": 1784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958239149107646e-05,
+      "loss": 1.0952,
+      "step": 1785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581770955510187e-05,
+      "loss": 1.0778,
+      "step": 1786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581149969097658e-05,
+      "loss": 1.1156,
+      "step": 1787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958052853186809e-05,
+      "loss": 1.1403,
+      "step": 1788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579906643850728e-05,
+      "loss": 1.1001,
+      "step": 1789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579284305074827e-05,
+      "loss": 1.0739,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578661515569676e-05,
+      "loss": 1.0974,
+      "step": 1791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578038275364575e-05,
+      "loss": 1.074,
+      "step": 1792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577414584488846e-05,
+      "loss": 1.0386,
+      "step": 1793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576790442971846e-05,
+      "loss": 1.135,
+      "step": 1794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576165850842934e-05,
+      "loss": 1.0777,
+      "step": 1795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575540808131497e-05,
+      "loss": 1.0639,
+      "step": 1796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957491531486695e-05,
+      "loss": 1.0511,
+      "step": 1797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957428937107872e-05,
+      "loss": 1.1024,
+      "step": 1798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957366297679626e-05,
+      "loss": 1.031,
+      "step": 1799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573036132049047e-05,
+      "loss": 1.0899,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572408836866574e-05,
+      "loss": 1.1058,
+      "step": 1801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571781091278357e-05,
+      "loss": 1.0642,
+      "step": 1802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957115289531393e-05,
+      "loss": 1.1131,
+      "step": 1803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570524249002852e-05,
+      "loss": 0.9957,
+      "step": 1804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956989515237471e-05,
+      "loss": 1.1089,
+      "step": 1805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956926560545909e-05,
+      "loss": 1.0455,
+      "step": 1806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568635608285625e-05,
+      "loss": 1.057,
+      "step": 1807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568005160883954e-05,
+      "loss": 1.0746,
+      "step": 1808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567374263283746e-05,
+      "loss": 1.0727,
+      "step": 1809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566742915514675e-05,
+      "loss": 1.1199,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956611111760646e-05,
+      "loss": 1.0888,
+      "step": 1811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565478869588822e-05,
+      "loss": 1.0445,
+      "step": 1812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564846171491512e-05,
+      "loss": 0.9699,
+      "step": 1813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95642130233443e-05,
+      "loss": 1.0842,
+      "step": 1814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563579425176976e-05,
+      "loss": 1.0493,
+      "step": 1815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562945377019354e-05,
+      "loss": 1.0579,
+      "step": 1816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562310878901262e-05,
+      "loss": 1.1846,
+      "step": 1817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561675930852564e-05,
+      "loss": 1.0093,
+      "step": 1818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956104053290313e-05,
+      "loss": 1.0932,
+      "step": 1819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560404685082858e-05,
+      "loss": 1.1166,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559768387421663e-05,
+      "loss": 1.1005,
+      "step": 1821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955913163994949e-05,
+      "loss": 1.0573,
+      "step": 1822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558494442696297e-05,
+      "loss": 1.0043,
+      "step": 1823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955785679569207e-05,
+      "loss": 1.0354,
+      "step": 1824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557218698966803e-05,
+      "loss": 1.1108,
+      "step": 1825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955658015255053e-05,
+      "loss": 1.1387,
+      "step": 1826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555941156473286e-05,
+      "loss": 1.0999,
+      "step": 1827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555301710765145e-05,
+      "loss": 1.1251,
+      "step": 1828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955466181545619e-05,
+      "loss": 1.0965,
+      "step": 1829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554021470576533e-05,
+      "loss": 1.1141,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553380676156305e-05,
+      "loss": 1.1024,
+      "step": 1831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552739432225653e-05,
+      "loss": 1.0542,
+      "step": 1832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552097738814748e-05,
+      "loss": 1.0571,
+      "step": 1833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955145559595379e-05,
+      "loss": 1.1122,
+      "step": 1834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550813003672987e-05,
+      "loss": 0.9607,
+      "step": 1835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550169962002576e-05,
+      "loss": 1.0629,
+      "step": 1836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549526470972813e-05,
+      "loss": 0.9988,
+      "step": 1837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954888253061398e-05,
+      "loss": 1.0335,
+      "step": 1838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548238140956373e-05,
+      "loss": 1.1064,
+      "step": 1839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547593302030316e-05,
+      "loss": 1.0941,
+      "step": 1840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546948013866145e-05,
+      "loss": 1.0749,
+      "step": 1841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546302276494226e-05,
+      "loss": 1.1306,
+      "step": 1842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545656089944936e-05,
+      "loss": 1.1053,
+      "step": 1843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545009454248687e-05,
+      "loss": 1.0458,
+      "step": 1844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95443623694359e-05,
+      "loss": 1.0724,
+      "step": 1845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954371483553703e-05,
+      "loss": 1.0337,
+      "step": 1846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543066852582534e-05,
+      "loss": 1.1027,
+      "step": 1847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954241842060291e-05,
+      "loss": 1.1133,
+      "step": 1848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541769539628667e-05,
+      "loss": 1.1263,
+      "step": 1849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954112020969033e-05,
+      "loss": 1.1228,
+      "step": 1850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954047043081846e-05,
+      "loss": 1.1036,
+      "step": 1851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539820203043623e-05,
+      "loss": 1.0169,
+      "step": 1852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539169526396417e-05,
+      "loss": 1.0661,
+      "step": 1853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538518400907462e-05,
+      "loss": 1.1075,
+      "step": 1854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953786682660739e-05,
+      "loss": 1.1602,
+      "step": 1855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953721480352686e-05,
+      "loss": 1.084,
+      "step": 1856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536562331696555e-05,
+      "loss": 1.1094,
+      "step": 1857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535909411147172e-05,
+      "loss": 1.0189,
+      "step": 1858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953525604190943e-05,
+      "loss": 1.097,
+      "step": 1859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534602224014078e-05,
+      "loss": 1.0517,
+      "step": 1860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533947957491876e-05,
+      "loss": 1.1176,
+      "step": 1861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533293242373608e-05,
+      "loss": 1.0989,
+      "step": 1862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532638078690086e-05,
+      "loss": 1.1256,
+      "step": 1863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531982466472126e-05,
+      "loss": 1.1205,
+      "step": 1864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531326405750586e-05,
+      "loss": 1.0799,
+      "step": 1865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9530669896556332e-05,
+      "loss": 1.0316,
+      "step": 1866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530012938920254e-05,
+      "loss": 1.0424,
+      "step": 1867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529355532873265e-05,
+      "loss": 1.0848,
+      "step": 1868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528697678446296e-05,
+      "loss": 1.0829,
+      "step": 1869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528039375670303e-05,
+      "loss": 1.1543,
+      "step": 1870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527380624576255e-05,
+      "loss": 1.0612,
+      "step": 1871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526721425195155e-05,
+      "loss": 1.0684,
+      "step": 1872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526061777558017e-05,
+      "loss": 1.0657,
+      "step": 1873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525401681695877e-05,
+      "loss": 1.0953,
+      "step": 1874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524741137639796e-05,
+      "loss": 1.049,
+      "step": 1875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524080145420856e-05,
+      "loss": 1.1758,
+      "step": 1876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523418705070157e-05,
+      "loss": 1.0746,
+      "step": 1877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522756816618818e-05,
+      "loss": 1.0625,
+      "step": 1878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522094480097987e-05,
+      "loss": 1.1101,
+      "step": 1879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952143169553883e-05,
+      "loss": 1.0545,
+      "step": 1880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520768462972526e-05,
+      "loss": 1.1126,
+      "step": 1881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520104782430287e-05,
+      "loss": 1.0851,
+      "step": 1882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519440653943338e-05,
+      "loss": 1.0889,
+      "step": 1883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518776077542933e-05,
+      "loss": 1.0832,
+      "step": 1884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518111053260336e-05,
+      "loss": 1.0688,
+      "step": 1885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951744558112684e-05,
+      "loss": 1.0469,
+      "step": 1886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516779661173758e-05,
+      "loss": 1.0723,
+      "step": 1887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516113293432423e-05,
+      "loss": 1.1651,
+      "step": 1888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951544647793419e-05,
+      "loss": 1.0497,
+      "step": 1889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514779214710432e-05,
+      "loss": 1.0037,
+      "step": 1890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514111503792544e-05,
+      "loss": 1.0253,
+      "step": 1891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951344334521195e-05,
+      "loss": 1.1066,
+      "step": 1892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512774739000084e-05,
+      "loss": 1.0724,
+      "step": 1893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512105685188406e-05,
+      "loss": 1.0178,
+      "step": 1894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511436183808395e-05,
+      "loss": 1.0488,
+      "step": 1895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510766234891557e-05,
+      "loss": 1.0854,
+      "step": 1896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510095838469415e-05,
+      "loss": 1.0573,
+      "step": 1897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509424994573505e-05,
+      "loss": 1.082,
+      "step": 1898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95087537032354e-05,
+      "loss": 1.0587,
+      "step": 1899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950808196448668e-05,
+      "loss": 1.1083,
+      "step": 1900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507409778358962e-05,
+      "loss": 1.1335,
+      "step": 1901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950673714488386e-05,
+      "loss": 1.1572,
+      "step": 1902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506064064093033e-05,
+      "loss": 1.0664,
+      "step": 1903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505390536018147e-05,
+      "loss": 1.0384,
+      "step": 1904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504716560690898e-05,
+      "loss": 1.1408,
+      "step": 1905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504042138142993e-05,
+      "loss": 1.1343,
+      "step": 1906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503367268406164e-05,
+      "loss": 1.1501,
+      "step": 1907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950269195151217e-05,
+      "loss": 1.0908,
+      "step": 1908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502016187492784e-05,
+      "loss": 1.0959,
+      "step": 1909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501339976379804e-05,
+      "loss": 1.1847,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500663318205047e-05,
+      "loss": 1.1503,
+      "step": 1911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949998621300035e-05,
+      "loss": 1.055,
+      "step": 1912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499308660797575e-05,
+      "loss": 1.1187,
+      "step": 1913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498630661628598e-05,
+      "loss": 1.1029,
+      "step": 1914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497952215525325e-05,
+      "loss": 1.0805,
+      "step": 1915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497273322519677e-05,
+      "loss": 1.0503,
+      "step": 1916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94965939826436e-05,
+      "loss": 1.0303,
+      "step": 1917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495914195929054e-05,
+      "loss": 1.0652,
+      "step": 1918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495233962408026e-05,
+      "loss": 1.0849,
+      "step": 1919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494553282112525e-05,
+      "loss": 1.0282,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493872155074575e-05,
+      "loss": 1.0422,
+      "step": 1921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493190581326228e-05,
+      "loss": 1.1189,
+      "step": 1922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492508560899558e-05,
+      "loss": 1.0444,
+      "step": 1923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491826093826646e-05,
+      "loss": 1.1022,
+      "step": 1924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491143180139606e-05,
+      "loss": 1.0541,
+      "step": 1925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490459819870576e-05,
+      "loss": 1.0123,
+      "step": 1926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489776013051706e-05,
+      "loss": 1.057,
+      "step": 1927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489091759715175e-05,
+      "loss": 1.047,
+      "step": 1928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948840705989317e-05,
+      "loss": 0.9863,
+      "step": 1929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948772191361792e-05,
+      "loss": 1.1451,
+      "step": 1930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948703632092165e-05,
+      "loss": 1.1504,
+      "step": 1931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948635028183663e-05,
+      "loss": 1.0627,
+      "step": 1932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485663796395133e-05,
+      "loss": 1.0506,
+      "step": 1933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484976864629462e-05,
+      "loss": 1.0511,
+      "step": 1934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948428948657194e-05,
+      "loss": 1.0234,
+      "step": 1935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483601662254906e-05,
+      "loss": 1.0916,
+      "step": 1936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948291339171073e-05,
+      "loss": 1.0161,
+      "step": 1937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482224674971787e-05,
+      "loss": 1.1295,
+      "step": 1938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481535512070494e-05,
+      "loss": 1.0313,
+      "step": 1939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480845903039275e-05,
+      "loss": 1.0307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948015584791057e-05,
+      "loss": 1.0455,
+      "step": 1941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479465346716857e-05,
+      "loss": 1.0831,
+      "step": 1942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947877439949062e-05,
+      "loss": 1.0225,
+      "step": 1943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478083006264374e-05,
+      "loss": 0.9672,
+      "step": 1944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947739116707065e-05,
+      "loss": 1.0415,
+      "step": 1945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476698881942e-05,
+      "loss": 1.0748,
+      "step": 1946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476006150910996e-05,
+      "loss": 1.0052,
+      "step": 1947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475312974010233e-05,
+      "loss": 1.0258,
+      "step": 1948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947461935127233e-05,
+      "loss": 1.0581,
+      "step": 1949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947392528272992e-05,
+      "loss": 1.0197,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473230768415663e-05,
+      "loss": 1.0732,
+      "step": 1951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472535808362237e-05,
+      "loss": 1.1602,
+      "step": 1952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947184040260234e-05,
+      "loss": 1.1221,
+      "step": 1953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471144551168697e-05,
+      "loss": 1.0797,
+      "step": 1954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470448254094045e-05,
+      "loss": 1.0653,
+      "step": 1955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469751511411145e-05,
+      "loss": 1.1314,
+      "step": 1956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946905432315279e-05,
+      "loss": 1.0441,
+      "step": 1957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946835668935177e-05,
+      "loss": 1.0912,
+      "step": 1958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467658610040926e-05,
+      "loss": 1.112,
+      "step": 1959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466960085253096e-05,
+      "loss": 1.122,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466261115021145e-05,
+      "loss": 1.1235,
+      "step": 1961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465561699377966e-05,
+      "loss": 1.0797,
+      "step": 1962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464861838356467e-05,
+      "loss": 1.1469,
+      "step": 1963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464161531989575e-05,
+      "loss": 1.0328,
+      "step": 1964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463460780310246e-05,
+      "loss": 1.0734,
+      "step": 1965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462759583351452e-05,
+      "loss": 0.9949,
+      "step": 1966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462057941146182e-05,
+      "loss": 1.0225,
+      "step": 1967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461355853727455e-05,
+      "loss": 1.15,
+      "step": 1968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460653321128303e-05,
+      "loss": 1.0362,
+      "step": 1969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459950343381776e-05,
+      "loss": 1.0832,
+      "step": 1970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459246920520964e-05,
+      "loss": 1.0573,
+      "step": 1971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458543052578955e-05,
+      "loss": 1.1031,
+      "step": 1972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457838739588874e-05,
+      "loss": 1.1415,
+      "step": 1973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945713398158385e-05,
+      "loss": 1.1311,
+      "step": 1974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456428778597058e-05,
+      "loss": 1.0706,
+      "step": 1975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455723130661666e-05,
+      "loss": 1.0446,
+      "step": 1976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945501703781089e-05,
+      "loss": 1.0506,
+      "step": 1977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945431050007794e-05,
+      "loss": 1.1433,
+      "step": 1978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453603517496075e-05,
+      "loss": 1.078,
+      "step": 1979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452896090098546e-05,
+      "loss": 1.0543,
+      "step": 1980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452188217918647e-05,
+      "loss": 1.065,
+      "step": 1981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945147990098968e-05,
+      "loss": 1.0974,
+      "step": 1982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945077113934498e-05,
+      "loss": 1.0962,
+      "step": 1983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450061933017893e-05,
+      "loss": 1.0899,
+      "step": 1984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449352282041784e-05,
+      "loss": 1.1105,
+      "step": 1985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944864218645005e-05,
+      "loss": 1.0886,
+      "step": 1986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447931646276105e-05,
+      "loss": 1.1366,
+      "step": 1987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447220661553373e-05,
+      "loss": 0.9981,
+      "step": 1988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446509232315315e-05,
+      "loss": 1.0274,
+      "step": 1989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94457973585954e-05,
+      "loss": 1.0888,
+      "step": 1990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944508504042713e-05,
+      "loss": 1.0154,
+      "step": 1991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444372277844018e-05,
+      "loss": 1.0433,
+      "step": 1992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443659070879598e-05,
+      "loss": 1.1128,
+      "step": 1993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442945419567433e-05,
+      "loss": 1.0846,
+      "step": 1994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94422313239411e-05,
+      "loss": 1.0691,
+      "step": 1995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94415167840342e-05,
+      "loss": 1.004,
+      "step": 1996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440801799880352e-05,
+      "loss": 0.9502,
+      "step": 1997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94400863715132e-05,
+      "loss": 1.0661,
+      "step": 1998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439370498966405e-05,
+      "loss": 1.0561,
+      "step": 1999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438654182273645e-05,
+      "loss": 1.0582,
+      "step": 2000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943793742146864e-05,
+      "loss": 1.1115,
+      "step": 2001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94372202165851e-05,
+      "loss": 1.0632,
+      "step": 2002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943650256765678e-05,
+      "loss": 1.1178,
+      "step": 2003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435784474717444e-05,
+      "loss": 1.0486,
+      "step": 2004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943506593780088e-05,
+      "loss": 1.1348,
+      "step": 2005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434346956940897e-05,
+      "loss": 1.1061,
+      "step": 2006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433627532171326e-05,
+      "loss": 1.1012,
+      "step": 2007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432907663526018e-05,
+      "loss": 1.1029,
+      "step": 2008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943218735103884e-05,
+      "loss": 1.1286,
+      "step": 2009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943146659474369e-05,
+      "loss": 1.0966,
+      "step": 2010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430745394674483e-05,
+      "loss": 1.0073,
+      "step": 2011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430023750865148e-05,
+      "loss": 1.0099,
+      "step": 2012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942930166334964e-05,
+      "loss": 1.0163,
+      "step": 2013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942857913216194e-05,
+      "loss": 1.055,
+      "step": 2014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9427856157336042e-05,
+      "loss": 1.0693,
+      "step": 2015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942713273890596e-05,
+      "loss": 1.0696,
+      "step": 2016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426408876905735e-05,
+      "loss": 1.0401,
+      "step": 2017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942568457136943e-05,
+      "loss": 1.0859,
+      "step": 2018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424959822331128e-05,
+      "loss": 1.1093,
+      "step": 2019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942423462982492e-05,
+      "loss": 1.1378,
+      "step": 2020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423508993884933e-05,
+      "loss": 0.9929,
+      "step": 2021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422782914545312e-05,
+      "loss": 1.0686,
+      "step": 2022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422056391840224e-05,
+      "loss": 1.0681,
+      "step": 2023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421329425803845e-05,
+      "loss": 1.0603,
+      "step": 2024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420602016470383e-05,
+      "loss": 1.1125,
+      "step": 2025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419874163874067e-05,
+      "loss": 1.1316,
+      "step": 2026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419145868049147e-05,
+      "loss": 1.0643,
+      "step": 2027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418417129029884e-05,
+      "loss": 1.0112,
+      "step": 2028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417687946850572e-05,
+      "loss": 1.0399,
+      "step": 2029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416958321545517e-05,
+      "loss": 1.0393,
+      "step": 2030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416228253149053e-05,
+      "loss": 1.0547,
+      "step": 2031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415497741695533e-05,
+      "loss": 1.0411,
+      "step": 2032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414766787219322e-05,
+      "loss": 1.0774,
+      "step": 2033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414035389754824e-05,
+      "loss": 1.0403,
+      "step": 2034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413303549336446e-05,
+      "loss": 1.1153,
+      "step": 2035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941257126599862e-05,
+      "loss": 1.0223,
+      "step": 2036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411838539775808e-05,
+      "loss": 1.0658,
+      "step": 2037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941110537070249e-05,
+      "loss": 1.1497,
+      "step": 2038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941037175881315e-05,
+      "loss": 1.067,
+      "step": 2039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940963770414232e-05,
+      "loss": 1.0742,
+      "step": 2040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940890320672453e-05,
+      "loss": 1.1014,
+      "step": 2041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408168266594344e-05,
+      "loss": 1.1008,
+      "step": 2042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407432883786345e-05,
+      "loss": 1.1129,
+      "step": 2043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940669705833513e-05,
+      "loss": 1.0602,
+      "step": 2044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940596079027532e-05,
+      "loss": 1.037,
+      "step": 2045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405224079641566e-05,
+      "loss": 1.0817,
+      "step": 2046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404486926468526e-05,
+      "loss": 1.1213,
+      "step": 2047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403749330790882e-05,
+      "loss": 1.1123,
+      "step": 2048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403011292643348e-05,
+      "loss": 1.0636,
+      "step": 2049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402272812060646e-05,
+      "loss": 1.062,
+      "step": 2050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401533889077526e-05,
+      "loss": 1.0734,
+      "step": 2051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940079452372875e-05,
+      "loss": 1.0464,
+      "step": 2052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400054716049116e-05,
+      "loss": 1.0275,
+      "step": 2053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399314466073426e-05,
+      "loss": 1.0206,
+      "step": 2054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939857377383651e-05,
+      "loss": 1.1007,
+      "step": 2055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397832639373232e-05,
+      "loss": 1.0457,
+      "step": 2056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397091062718446e-05,
+      "loss": 1.066,
+      "step": 2057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396349043907057e-05,
+      "loss": 0.9886,
+      "step": 2058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395606582973978e-05,
+      "loss": 1.0338,
+      "step": 2059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394863679954145e-05,
+      "loss": 1.0853,
+      "step": 2060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93941203348825e-05,
+      "loss": 1.0375,
+      "step": 2061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939337654779404e-05,
+      "loss": 1.0374,
+      "step": 2062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392632318723745e-05,
+      "loss": 1.0977,
+      "step": 2063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939188764770664e-05,
+      "loss": 1.0903,
+      "step": 2064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391142534777764e-05,
+      "loss": 1.0649,
+      "step": 2065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390396979972175e-05,
+      "loss": 1.0791,
+      "step": 2066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938965098332495e-05,
+      "loss": 1.1347,
+      "step": 2067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93889045448712e-05,
+      "loss": 1.0133,
+      "step": 2068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938815766464604e-05,
+      "loss": 0.9978,
+      "step": 2069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387410342684608e-05,
+      "loss": 1.0885,
+      "step": 2070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386662579022075e-05,
+      "loss": 1.0439,
+      "step": 2071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385914373693625e-05,
+      "loss": 1.0364,
+      "step": 2072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385165726734455e-05,
+      "loss": 1.0946,
+      "step": 2073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93844166381798e-05,
+      "loss": 1.0332,
+      "step": 2074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93836671080649e-05,
+      "loss": 1.0626,
+      "step": 2075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382917136425024e-05,
+      "loss": 1.1097,
+      "step": 2076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382166723295463e-05,
+      "loss": 1.079,
+      "step": 2077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381415868711524e-05,
+      "loss": 1.0035,
+      "step": 2078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380664572708534e-05,
+      "loss": 1.0709,
+      "step": 2079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937991283532185e-05,
+      "loss": 1.1454,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937916065658683e-05,
+      "loss": 1.0886,
+      "step": 2081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378408036538886e-05,
+      "loss": 0.9975,
+      "step": 2082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937765497521341e-05,
+      "loss": 1.1189,
+      "step": 2083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376901472645852e-05,
+      "loss": 1.0652,
+      "step": 2084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376147528871652e-05,
+      "loss": 1.0191,
+      "step": 2085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375393143926295e-05,
+      "loss": 1.0578,
+      "step": 2086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374638317845274e-05,
+      "loss": 1.0341,
+      "step": 2087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373883050664105e-05,
+      "loss": 1.1175,
+      "step": 2088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373127342418324e-05,
+      "loss": 1.1205,
+      "step": 2089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937237119314349e-05,
+      "loss": 1.0422,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371614602875183e-05,
+      "loss": 1.0792,
+      "step": 2091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370857571649e-05,
+      "loss": 1.0664,
+      "step": 2092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370100099500563e-05,
+      "loss": 1.0545,
+      "step": 2093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936934218646551e-05,
+      "loss": 1.0301,
+      "step": 2094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936858383257951e-05,
+      "loss": 1.0347,
+      "step": 2095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367825037878235e-05,
+      "loss": 1.0709,
+      "step": 2096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93670658023974e-05,
+      "loss": 0.9651,
+      "step": 2097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366306126172718e-05,
+      "loss": 1.0303,
+      "step": 2098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936554600923994e-05,
+      "loss": 1.0894,
+      "step": 2099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936478545163483e-05,
+      "loss": 1.0522,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364024453393174e-05,
+      "loss": 1.1184,
+      "step": 2101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363263014550776e-05,
+      "loss": 1.0522,
+      "step": 2102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936250113514347e-05,
+      "loss": 1.0239,
+      "step": 2103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93617388152071e-05,
+      "loss": 1.048,
+      "step": 2104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360976054777537e-05,
+      "loss": 1.039,
+      "step": 2105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360212853890666e-05,
+      "loss": 1.085,
+      "step": 2106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93594492125824e-05,
+      "loss": 1.1363,
+      "step": 2107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358685130888678e-05,
+      "loss": 1.0156,
+      "step": 2108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935792060884544e-05,
+      "loss": 1.0975,
+      "step": 2109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357155646488666e-05,
+      "loss": 1.0144,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356390243854344e-05,
+      "loss": 1.0859,
+      "step": 2111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355624400978493e-05,
+      "loss": 1.1805,
+      "step": 2112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935485811789715e-05,
+      "loss": 1.1127,
+      "step": 2113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354091394646362e-05,
+      "loss": 1.1215,
+      "step": 2114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353324231262213e-05,
+      "loss": 0.9817,
+      "step": 2115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352556627780794e-05,
+      "loss": 1.0283,
+      "step": 2116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351788584238228e-05,
+      "loss": 1.16,
+      "step": 2117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351020100670648e-05,
+      "loss": 1.0555,
+      "step": 2118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935025117711422e-05,
+      "loss": 1.0546,
+      "step": 2119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349481813605117e-05,
+      "loss": 1.0155,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348712010179543e-05,
+      "loss": 1.0486,
+      "step": 2121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934794176687372e-05,
+      "loss": 1.1042,
+      "step": 2122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347171083723885e-05,
+      "loss": 0.9911,
+      "step": 2123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346399960766306e-05,
+      "loss": 1.1131,
+      "step": 2124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345628398037265e-05,
+      "loss": 1.1076,
+      "step": 2125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344856395573065e-05,
+      "loss": 1.1276,
+      "step": 2126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934408395341003e-05,
+      "loss": 1.081,
+      "step": 2127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343311071584504e-05,
+      "loss": 1.0274,
+      "step": 2128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342537750132857e-05,
+      "loss": 1.1318,
+      "step": 2129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341763989091476e-05,
+      "loss": 1.0001,
+      "step": 2130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340989788496763e-05,
+      "loss": 0.9872,
+      "step": 2131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340215148385152e-05,
+      "loss": 1.0322,
+      "step": 2132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339440068793088e-05,
+      "loss": 1.0965,
+      "step": 2133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933866454975704e-05,
+      "loss": 1.1079,
+      "step": 2134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93378885913135e-05,
+      "loss": 1.0501,
+      "step": 2135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933711219349898e-05,
+      "loss": 1.0999,
+      "step": 2136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933633535635001e-05,
+      "loss": 1.0949,
+      "step": 2137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933555807990314e-05,
+      "loss": 1.082,
+      "step": 2138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334780364194943e-05,
+      "loss": 1.1153,
+      "step": 2139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334002209262018e-05,
+      "loss": 1.004,
+      "step": 2140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333223615140974e-05,
+      "loss": 1.005,
+      "step": 2141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332444581868445e-05,
+      "loss": 1.0698,
+      "step": 2142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331665109481093e-05,
+      "loss": 1.1104,
+      "step": 2143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330885198015585e-05,
+      "loss": 1.0209,
+      "step": 2144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330104847508625e-05,
+      "loss": 1.1031,
+      "step": 2145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932932405799693e-05,
+      "loss": 1.0388,
+      "step": 2146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328542829517232e-05,
+      "loss": 1.11,
+      "step": 2147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327761162106295e-05,
+      "loss": 1.0785,
+      "step": 2148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93269790558009e-05,
+      "loss": 1.0832,
+      "step": 2149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932619651063784e-05,
+      "loss": 1.0397,
+      "step": 2150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932541352665394e-05,
+      "loss": 1.0534,
+      "step": 2151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324630103886046e-05,
+      "loss": 1.073,
+      "step": 2152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323846242371012e-05,
+      "loss": 1.0678,
+      "step": 2153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323061942145726e-05,
+      "loss": 1.198,
+      "step": 2154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322277203247087e-05,
+      "loss": 0.9417,
+      "step": 2155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932149202571202e-05,
+      "loss": 1.0431,
+      "step": 2156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320706409577475e-05,
+      "loss": 0.9864,
+      "step": 2157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319920354880407e-05,
+      "loss": 0.9992,
+      "step": 2158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319133861657814e-05,
+      "loss": 1.0875,
+      "step": 2159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318346929946692e-05,
+      "loss": 1.1575,
+      "step": 2160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317559559784075e-05,
+      "loss": 1.0928,
+      "step": 2161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316771751207007e-05,
+      "loss": 1.038,
+      "step": 2162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315983504252557e-05,
+      "loss": 1.1123,
+      "step": 2163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315194818957814e-05,
+      "loss": 1.0491,
+      "step": 2164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931440569535989e-05,
+      "loss": 1.0837,
+      "step": 2165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931361613349591e-05,
+      "loss": 1.0586,
+      "step": 2166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931282613340303e-05,
+      "loss": 1.0492,
+      "step": 2167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931203569511842e-05,
+      "loss": 1.0336,
+      "step": 2168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931124481867927e-05,
+      "loss": 1.0754,
+      "step": 2169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310453504122798e-05,
+      "loss": 1.0778,
+      "step": 2170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930966175148623e-05,
+      "loss": 1.1287,
+      "step": 2171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308869560806826e-05,
+      "loss": 1.0533,
+      "step": 2172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930807693212186e-05,
+      "loss": 1.0576,
+      "step": 2173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930728386546862e-05,
+      "loss": 1.0491,
+      "step": 2174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930649036088443e-05,
+      "loss": 1.0676,
+      "step": 2175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930569641840662e-05,
+      "loss": 1.0927,
+      "step": 2176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304902038072556e-05,
+      "loss": 1.1091,
+      "step": 2177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930410721991961e-05,
+      "loss": 1.1351,
+      "step": 2178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303311963985177e-05,
+      "loss": 1.0673,
+      "step": 2179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302516270306683e-05,
+      "loss": 1.1161,
+      "step": 2180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930172013892156e-05,
+      "loss": 1.1234,
+      "step": 2181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930092356986727e-05,
+      "loss": 1.0676,
+      "step": 2182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93001265631813e-05,
+      "loss": 1.0622,
+      "step": 2183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929932911890114e-05,
+      "loss": 1.0363,
+      "step": 2184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298531237064318e-05,
+      "loss": 1.09,
+      "step": 2185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297732917708384e-05,
+      "loss": 1.0955,
+      "step": 2186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296934160870887e-05,
+      "loss": 1.0786,
+      "step": 2187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296134966589418e-05,
+      "loss": 1.0198,
+      "step": 2188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295335334901578e-05,
+      "loss": 1.0296,
+      "step": 2189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294535265844992e-05,
+      "loss": 1.024,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929373475945731e-05,
+      "loss": 1.0163,
+      "step": 2191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292933815776195e-05,
+      "loss": 1.0201,
+      "step": 2192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929213243483933e-05,
+      "loss": 1.051,
+      "step": 2193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929133061668443e-05,
+      "loss": 1.0351,
+      "step": 2194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290528361349213e-05,
+      "loss": 1.0972,
+      "step": 2195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928972566887143e-05,
+      "loss": 1.0982,
+      "step": 2196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288922539288853e-05,
+      "loss": 1.0788,
+      "step": 2197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288118972639275e-05,
+      "loss": 1.0201,
+      "step": 2198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287314968960493e-05,
+      "loss": 0.984,
+      "step": 2199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286510528290346e-05,
+      "loss": 1.0562,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285705650666682e-05,
+      "loss": 1.0861,
+      "step": 2201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284900336127375e-05,
+      "loss": 0.9581,
+      "step": 2202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928409458471032e-05,
+      "loss": 1.1138,
+      "step": 2203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283288396453425e-05,
+      "loss": 1.1044,
+      "step": 2204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282481771394618e-05,
+      "loss": 1.0376,
+      "step": 2205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281674709571864e-05,
+      "loss": 1.0262,
+      "step": 2206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928086721102313e-05,
+      "loss": 1.0587,
+      "step": 2207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280059275786413e-05,
+      "loss": 1.0168,
+      "step": 2208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279250903899727e-05,
+      "loss": 1.0717,
+      "step": 2209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927844209540111e-05,
+      "loss": 1.1363,
+      "step": 2210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927763285032862e-05,
+      "loss": 1.0686,
+      "step": 2211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276823168720328e-05,
+      "loss": 1.1079,
+      "step": 2212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276013050614337e-05,
+      "loss": 1.0128,
+      "step": 2213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275202496048762e-05,
+      "loss": 1.0229,
+      "step": 2214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274391505061743e-05,
+      "loss": 1.0913,
+      "step": 2215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273580077691437e-05,
+      "loss": 1.0194,
+      "step": 2216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927276821397603e-05,
+      "loss": 0.9406,
+      "step": 2217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927195591395371e-05,
+      "loss": 1.1488,
+      "step": 2218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271143177662715e-05,
+      "loss": 0.9921,
+      "step": 2219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270330005141273e-05,
+      "loss": 1.0654,
+      "step": 2220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269516396427648e-05,
+      "loss": 1.1115,
+      "step": 2221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268702351560123e-05,
+      "loss": 1.1292,
+      "step": 2222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267887870577007e-05,
+      "loss": 1.0048,
+      "step": 2223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267072953516614e-05,
+      "loss": 1.0517,
+      "step": 2224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266257600417293e-05,
+      "loss": 1.045,
+      "step": 2225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265441811317406e-05,
+      "loss": 1.0941,
+      "step": 2226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926462558625534e-05,
+      "loss": 1.0883,
+      "step": 2227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92638089252695e-05,
+      "loss": 0.993,
+      "step": 2228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926299182839831e-05,
+      "loss": 1.0784,
+      "step": 2229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926217429568022e-05,
+      "loss": 1.0643,
+      "step": 2230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261356327153694e-05,
+      "loss": 1.1383,
+      "step": 2231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260537922857223e-05,
+      "loss": 1.0818,
+      "step": 2232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259719082829312e-05,
+      "loss": 1.1107,
+      "step": 2233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925889980710849e-05,
+      "loss": 1.0766,
+      "step": 2234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258080095733303e-05,
+      "loss": 1.0967,
+      "step": 2235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257259948742323e-05,
+      "loss": 1.0998,
+      "step": 2236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256439366174143e-05,
+      "loss": 1.0558,
+      "step": 2237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255618348067375e-05,
+      "loss": 1.0709,
+      "step": 2238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925479689446064e-05,
+      "loss": 0.975,
+      "step": 2239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253975005392595e-05,
+      "loss": 1.0501,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253152680901915e-05,
+      "loss": 1.0717,
+      "step": 2241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252329921027287e-05,
+      "loss": 1.0398,
+      "step": 2242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251506725807428e-05,
+      "loss": 1.004,
+      "step": 2243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250683095281068e-05,
+      "loss": 1.0236,
+      "step": 2244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249859029486963e-05,
+      "loss": 1.0727,
+      "step": 2245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924903452846389e-05,
+      "loss": 1.0786,
+      "step": 2246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924820959225064e-05,
+      "loss": 1.1044,
+      "step": 2247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924738422088603e-05,
+      "loss": 1.0809,
+      "step": 2248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246558414408892e-05,
+      "loss": 1.1146,
+      "step": 2249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245732172858088e-05,
+      "loss": 1.0595,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924490549627249e-05,
+      "loss": 1.1459,
+      "step": 2251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244078384690998e-05,
+      "loss": 1.0652,
+      "step": 2252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924325083815253e-05,
+      "loss": 1.1215,
+      "step": 2253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242422856696022e-05,
+      "loss": 1.0076,
+      "step": 2254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924159444036043e-05,
+      "loss": 1.087,
+      "step": 2255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240765589184742e-05,
+      "loss": 1.0598,
+      "step": 2256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239936303207953e-05,
+      "loss": 1.0565,
+      "step": 2257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239106582469077e-05,
+      "loss": 1.1352,
+      "step": 2258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238276427007164e-05,
+      "loss": 1.0483,
+      "step": 2259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237445836861267e-05,
+      "loss": 1.0915,
+      "step": 2260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236614812070473e-05,
+      "loss": 1.0168,
+      "step": 2261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235783352673884e-05,
+      "loss": 1.0013,
+      "step": 2262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234951458710615e-05,
+      "loss": 1.0851,
+      "step": 2263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234119130219814e-05,
+      "loss": 1.019,
+      "step": 2264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233286367240647e-05,
+      "loss": 1.0257,
+      "step": 2265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232453169812292e-05,
+      "loss": 1.093,
+      "step": 2266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231619537973955e-05,
+      "loss": 1.0196,
+      "step": 2267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230785471764864e-05,
+      "loss": 1.0981,
+      "step": 2268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229950971224254e-05,
+      "loss": 1.0868,
+      "step": 2269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229116036391404e-05,
+      "loss": 1.0271,
+      "step": 2270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922828066730559e-05,
+      "loss": 1.0635,
+      "step": 2271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922744486400612e-05,
+      "loss": 1.0286,
+      "step": 2272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226608626532323e-05,
+      "loss": 1.0567,
+      "step": 2273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225771954923544e-05,
+      "loss": 1.1503,
+      "step": 2274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922493484921915e-05,
+      "loss": 1.1088,
+      "step": 2275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224097309458533e-05,
+      "loss": 1.0096,
+      "step": 2276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223259335681096e-05,
+      "loss": 1.1863,
+      "step": 2277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922242092792627e-05,
+      "loss": 1.0234,
+      "step": 2278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221582086233507e-05,
+      "loss": 1.0937,
+      "step": 2279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922074281064227e-05,
+      "loss": 1.0392,
+      "step": 2280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219903101192055e-05,
+      "loss": 1.0897,
+      "step": 2281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921906295792237e-05,
+      "loss": 1.0815,
+      "step": 2282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218222380872748e-05,
+      "loss": 1.0973,
+      "step": 2283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217381370082736e-05,
+      "loss": 1.0852,
+      "step": 2284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921653992559191e-05,
+      "loss": 1.0185,
+      "step": 2285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215698047439856e-05,
+      "loss": 1.0561,
+      "step": 2286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92148557356662e-05,
+      "loss": 1.0739,
+      "step": 2287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214012990310555e-05,
+      "loss": 1.0971,
+      "step": 2288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213169811412592e-05,
+      "loss": 1.0791,
+      "step": 2289
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212326199011973e-05,
+      "loss": 1.0782,
+      "step": 2290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211482153148397e-05,
+      "loss": 1.0325,
+      "step": 2291
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921063767386158e-05,
+      "loss": 1.0614,
+      "step": 2292
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209792761191256e-05,
+      "loss": 1.0791,
+      "step": 2293
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208947415177174e-05,
+      "loss": 1.1931,
+      "step": 2294
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920810163585912e-05,
+      "loss": 1.0821,
+      "step": 2295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920725542327688e-05,
+      "loss": 1.0116,
+      "step": 2296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206408777470277e-05,
+      "loss": 1.0385,
+      "step": 2297
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205561698479148e-05,
+      "loss": 0.9678,
+      "step": 2298
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204714186343346e-05,
+      "loss": 1.0545,
+      "step": 2299
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203866241102753e-05,
+      "loss": 1.0699,
+      "step": 2300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203017862797263e-05,
+      "loss": 1.0583,
+      "step": 2301
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202169051466795e-05,
+      "loss": 1.014,
+      "step": 2302
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201319807151293e-05,
+      "loss": 1.0096,
+      "step": 2303
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200470129890708e-05,
+      "loss": 1.1187,
+      "step": 2304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199620019725022e-05,
+      "loss": 0.9869,
+      "step": 2305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919876947669424e-05,
+      "loss": 1.0749,
+      "step": 2306
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197918500838375e-05,
+      "loss": 1.098,
+      "step": 2307
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197067092197472e-05,
+      "loss": 1.0816,
+      "step": 2308
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919621525081159e-05,
+      "loss": 1.0832,
+      "step": 2309
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195362976720812e-05,
+      "loss": 0.988,
+      "step": 2310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194510269965238e-05,
+      "loss": 1.0502,
+      "step": 2311
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919365713058499e-05,
+      "loss": 1.0949,
+      "step": 2312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919280355862021e-05,
+      "loss": 1.0364,
+      "step": 2313
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9191949554111064e-05,
+      "loss": 1.0844,
+      "step": 2314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919109511709773e-05,
+      "loss": 1.0924,
+      "step": 2315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190240247620417e-05,
+      "loss": 1.1072,
+      "step": 2316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189384945719342e-05,
+      "loss": 1.0687,
+      "step": 2317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188529211434755e-05,
+      "loss": 1.035,
+      "step": 2318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187673044806917e-05,
+      "loss": 1.1145,
+      "step": 2319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918681644587612e-05,
+      "loss": 0.9637,
+      "step": 2320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185959414682655e-05,
+      "loss": 1.1242,
+      "step": 2321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185101951266856e-05,
+      "loss": 1.1204,
+      "step": 2322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918424405566907e-05,
+      "loss": 1.0575,
+      "step": 2323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183385727929665e-05,
+      "loss": 1.0552,
+      "step": 2324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918252696808902e-05,
+      "loss": 1.0871,
+      "step": 2325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181667776187548e-05,
+      "loss": 1.1545,
+      "step": 2326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180808152265675e-05,
+      "loss": 1.0242,
+      "step": 2327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179948096363845e-05,
+      "loss": 1.0194,
+      "step": 2328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917908760852253e-05,
+      "loss": 1.0499,
+      "step": 2329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178226688782212e-05,
+      "loss": 1.031,
+      "step": 2330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177365337183408e-05,
+      "loss": 1.0066,
+      "step": 2331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917650355376664e-05,
+      "loss": 1.0586,
+      "step": 2332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917564133857246e-05,
+      "loss": 1.0492,
+      "step": 2333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174778691641437e-05,
+      "loss": 1.1314,
+      "step": 2334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173915613014163e-05,
+      "loss": 1.0754,
+      "step": 2335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173052102731242e-05,
+      "loss": 1.0505,
+      "step": 2336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172188160833308e-05,
+      "loss": 1.0469,
+      "step": 2337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171323787361014e-05,
+      "loss": 0.9894,
+      "step": 2338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170458982355026e-05,
+      "loss": 1.1082,
+      "step": 2339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916959374585604e-05,
+      "loss": 0.9357,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168728077904763e-05,
+      "loss": 1.031,
+      "step": 2341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916786197854193e-05,
+      "loss": 1.0163,
+      "step": 2342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166995447808292e-05,
+      "loss": 1.1049,
+      "step": 2343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166128485744622e-05,
+      "loss": 1.058,
+      "step": 2344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165261092391712e-05,
+      "loss": 1.0692,
+      "step": 2345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164393267790374e-05,
+      "loss": 1.0072,
+      "step": 2346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163525011981446e-05,
+      "loss": 1.0517,
+      "step": 2347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162656325005774e-05,
+      "loss": 1.1109,
+      "step": 2348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161787206904238e-05,
+      "loss": 1.1141,
+      "step": 2349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160917657717734e-05,
+      "loss": 1.068,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160047677487168e-05,
+      "loss": 1.0172,
+      "step": 2351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159177266253482e-05,
+      "loss": 1.0768,
+      "step": 2352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158306424057625e-05,
+      "loss": 1.0784,
+      "step": 2353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157435150940583e-05,
+      "loss": 1.0825,
+      "step": 2354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156563446943336e-05,
+      "loss": 1.0243,
+      "step": 2355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155691312106913e-05,
+      "loss": 1.0742,
+      "step": 2356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154818746472346e-05,
+      "loss": 1.0885,
+      "step": 2357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153945750080688e-05,
+      "loss": 1.1039,
+      "step": 2358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915307232297302e-05,
+      "loss": 1.0672,
+      "step": 2359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152198465190438e-05,
+      "loss": 1.0583,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151324176774058e-05,
+      "loss": 1.0713,
+      "step": 2361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150449457765017e-05,
+      "loss": 1.0048,
+      "step": 2362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149574308204474e-05,
+      "loss": 1.1439,
+      "step": 2363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148698728133607e-05,
+      "loss": 1.0054,
+      "step": 2364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147822717593618e-05,
+      "loss": 1.0108,
+      "step": 2365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146946276625714e-05,
+      "loss": 1.1109,
+      "step": 2366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146069405271145e-05,
+      "loss": 0.9324,
+      "step": 2367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914519210357117e-05,
+      "loss": 1.1386,
+      "step": 2368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914431437156706e-05,
+      "loss": 1.0422,
+      "step": 2369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143436209300116e-05,
+      "loss": 1.0181,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142557616811667e-05,
+      "loss": 1.1438,
+      "step": 2371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914167859414304e-05,
+      "loss": 1.0388,
+      "step": 2372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140799141335608e-05,
+      "loss": 1.0847,
+      "step": 2373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139919258430743e-05,
+      "loss": 1.0607,
+      "step": 2374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913903894546985e-05,
+      "loss": 1.0981,
+      "step": 2375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138158202494343e-05,
+      "loss": 1.0282,
+      "step": 2376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913727702954567e-05,
+      "loss": 1.0377,
+      "step": 2377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913639542666529e-05,
+      "loss": 1.0469,
+      "step": 2378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135513393894687e-05,
+      "loss": 1.1179,
+      "step": 2379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913463093127536e-05,
+      "loss": 0.9878,
+      "step": 2380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913374803884883e-05,
+      "loss": 1.1114,
+      "step": 2381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132864716656643e-05,
+      "loss": 1.1247,
+      "step": 2382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913198096474036e-05,
+      "loss": 1.0515,
+      "step": 2383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131096783141567e-05,
+      "loss": 1.0722,
+      "step": 2384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130212171901855e-05,
+      "loss": 1.0432,
+      "step": 2385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129327131062862e-05,
+      "loss": 1.0327,
+      "step": 2386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912844166066622e-05,
+      "loss": 1.0611,
+      "step": 2387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127555760753602e-05,
+      "loss": 1.0521,
+      "step": 2388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126669431366686e-05,
+      "loss": 0.9798,
+      "step": 2389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912578267254718e-05,
+      "loss": 1.0921,
+      "step": 2390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124895484336804e-05,
+      "loss": 1.0913,
+      "step": 2391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91240078667773e-05,
+      "loss": 1.0717,
+      "step": 2392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912311981991044e-05,
+      "loss": 1.1191,
+      "step": 2393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122231343778004e-05,
+      "loss": 1.1324,
+      "step": 2394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121342438421805e-05,
+      "loss": 1.0399,
+      "step": 2395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120453103883654e-05,
+      "loss": 1.0931,
+      "step": 2396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119563340205407e-05,
+      "loss": 1.1117,
+      "step": 2397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118673147428927e-05,
+      "loss": 1.065,
+      "step": 2398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91177825255961e-05,
+      "loss": 1.0518,
+      "step": 2399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116891474748826e-05,
+      "loss": 1.0983,
+      "step": 2400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115999994929045e-05,
+      "loss": 1.0667,
+      "step": 2401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911510808617869e-05,
+      "loss": 1.2458,
+      "step": 2402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114215748539734e-05,
+      "loss": 1.1463,
+      "step": 2403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911332298205416e-05,
+      "loss": 0.9721,
+      "step": 2404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911242978676398e-05,
+      "loss": 1.049,
+      "step": 2405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111536162711214e-05,
+      "loss": 1.0855,
+      "step": 2406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110642109937922e-05,
+      "loss": 1.0065,
+      "step": 2407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109747628486154e-05,
+      "loss": 1.0813,
+      "step": 2408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910885271839801e-05,
+      "loss": 1.0711,
+      "step": 2409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107957379715597e-05,
+      "loss": 1.1171,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107061612481036e-05,
+      "loss": 1.1038,
+      "step": 2411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910616541673648e-05,
+      "loss": 1.1113,
+      "step": 2412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105268792524098e-05,
+      "loss": 1.0622,
+      "step": 2413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104371739886075e-05,
+      "loss": 1.0555,
+      "step": 2414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103474258864626e-05,
+      "loss": 1.0565,
+      "step": 2415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102576349501972e-05,
+      "loss": 1.024,
+      "step": 2416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101678011840362e-05,
+      "loss": 1.131,
+      "step": 2417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100779245922072e-05,
+      "loss": 1.0365,
+      "step": 2418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099880051789385e-05,
+      "loss": 1.1089,
+      "step": 2419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098980429484613e-05,
+      "loss": 0.9799,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098080379050088e-05,
+      "loss": 1.0261,
+      "step": 2421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097179900528157e-05,
+      "loss": 1.0444,
+      "step": 2422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096278993961185e-05,
+      "loss": 1.1185,
+      "step": 2423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909537765939157e-05,
+      "loss": 1.0781,
+      "step": 2424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094475896861713e-05,
+      "loss": 1.1571,
+      "step": 2425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093573706414056e-05,
+      "loss": 1.0482,
+      "step": 2426
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909267108809104e-05,
+      "loss": 1.0685,
+      "step": 2427
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909176804193514e-05,
+      "loss": 1.0644,
+      "step": 2428
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090864567988844e-05,
+      "loss": 1.0673,
+      "step": 2429
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089960666294662e-05,
+      "loss": 1.0436,
+      "step": 2430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908905633689513e-05,
+      "loss": 1.0698,
+      "step": 2431
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9088151579832793e-05,
+      "loss": 0.9851,
+      "step": 2432
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087246395150223e-05,
+      "loss": 1.0806,
+      "step": 2433
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086340782890015e-05,
+      "loss": 1.0341,
+      "step": 2434
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085434743094776e-05,
+      "loss": 1.0134,
+      "step": 2435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084528275807145e-05,
+      "loss": 1.0211,
+      "step": 2436
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908362138106976e-05,
+      "loss": 1.0427,
+      "step": 2437
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908271405892531e-05,
+      "loss": 1.1084,
+      "step": 2438
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081806309416467e-05,
+      "loss": 1.0366,
+      "step": 2439
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080898132585962e-05,
+      "loss": 1.0271,
+      "step": 2440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079989528476514e-05,
+      "loss": 1.0376,
+      "step": 2441
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907908049713088e-05,
+      "loss": 1.1073,
+      "step": 2442
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907817103859183e-05,
+      "loss": 1.0932,
+      "step": 2443
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907726115290216e-05,
+      "loss": 1.0615,
+      "step": 2444
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907635084010468e-05,
+      "loss": 1.089,
+      "step": 2445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075440100242226e-05,
+      "loss": 1.0657,
+      "step": 2446
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074528933357644e-05,
+      "loss": 1.0636,
+      "step": 2447
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073617339493807e-05,
+      "loss": 1.0653,
+      "step": 2448
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072705318693617e-05,
+      "loss": 1.0252,
+      "step": 2449
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071792870999976e-05,
+      "loss": 1.0801,
+      "step": 2450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070879996455824e-05,
+      "loss": 0.9922,
+      "step": 2451
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069966695104113e-05,
+      "loss": 1.1213,
+      "step": 2452
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906905296698781e-05,
+      "loss": 1.0411,
+      "step": 2453
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068138812149917e-05,
+      "loss": 1.0046,
+      "step": 2454
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067224230633443e-05,
+      "loss": 1.0026,
+      "step": 2455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906630922248142e-05,
+      "loss": 0.9971,
+      "step": 2456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065393787736903e-05,
+      "loss": 0.9809,
+      "step": 2457
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064477926442964e-05,
+      "loss": 1.0551,
+      "step": 2458
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90635616386427e-05,
+      "loss": 1.0912,
+      "step": 2459
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062644924379225e-05,
+      "loss": 1.0406,
+      "step": 2460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906172778369567e-05,
+      "loss": 1.1022,
+      "step": 2461
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060810216635184e-05,
+      "loss": 1.0848,
+      "step": 2462
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.905989222324095e-05,
+      "loss": 1.0309,
+      "step": 2463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058973803556157e-05,
+      "loss": 1.0517,
+      "step": 2464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905805495762402e-05,
+      "loss": 1.0689,
+      "step": 2465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057135685487772e-05,
+      "loss": 1.0384,
+      "step": 2466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905621598719067e-05,
+      "loss": 1.0942,
+      "step": 2467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055295862775983e-05,
+      "loss": 1.004,
+      "step": 2468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054375312287007e-05,
+      "loss": 1.0059,
+      "step": 2469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053454335767063e-05,
+      "loss": 1.0269,
+      "step": 2470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052532933259477e-05,
+      "loss": 1.0238,
+      "step": 2471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051611104807604e-05,
+      "loss": 1.055,
+      "step": 2472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905068885045482e-05,
+      "loss": 1.0621,
+      "step": 2473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904976617024452e-05,
+      "loss": 1.1183,
+      "step": 2474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904884306422012e-05,
+      "loss": 1.0198,
+      "step": 2475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904791953242505e-05,
+      "loss": 1.0647,
+      "step": 2476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046995574902768e-05,
+      "loss": 1.0131,
+      "step": 2477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046071191696747e-05,
+      "loss": 1.0477,
+      "step": 2478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904514638285048e-05,
+      "loss": 0.9826,
+      "step": 2479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044221148407488e-05,
+      "loss": 1.0115,
+      "step": 2480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043295488411297e-05,
+      "loss": 1.0866,
+      "step": 2481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042369402905466e-05,
+      "loss": 1.0557,
+      "step": 2482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904144289193357e-05,
+      "loss": 0.9937,
+      "step": 2483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040515955539206e-05,
+      "loss": 1.0776,
+      "step": 2484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903958859376598e-05,
+      "loss": 1.0832,
+      "step": 2485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038660806657535e-05,
+      "loss": 1.0239,
+      "step": 2486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037732594257523e-05,
+      "loss": 1.0524,
+      "step": 2487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903680395660962e-05,
+      "loss": 1.0504,
+      "step": 2488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035874893757518e-05,
+      "loss": 1.0969,
+      "step": 2489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034945405744932e-05,
+      "loss": 1.0641,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90340154926156e-05,
+      "loss": 1.2017,
+      "step": 2491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033085154413274e-05,
+      "loss": 0.9982,
+      "step": 2492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903215439118173e-05,
+      "loss": 1.0693,
+      "step": 2493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031223202964762e-05,
+      "loss": 1.0216,
+      "step": 2494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030291589806184e-05,
+      "loss": 1.074,
+      "step": 2495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902935955174983e-05,
+      "loss": 1.0521,
+      "step": 2496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902842708883956e-05,
+      "loss": 1.0476,
+      "step": 2497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027494201119243e-05,
+      "loss": 1.0745,
+      "step": 2498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026560888632777e-05,
+      "loss": 1.0934,
+      "step": 2499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025627151424073e-05,
+      "loss": 1.0706,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024692989537068e-05,
+      "loss": 1.0731,
+      "step": 2501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902375840301572e-05,
+      "loss": 1.0966,
+      "step": 2502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022823391903998e-05,
+      "loss": 1.1122,
+      "step": 2503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021887956245895e-05,
+      "loss": 1.014,
+      "step": 2504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020952096085437e-05,
+      "loss": 1.082,
+      "step": 2505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902001581146665e-05,
+      "loss": 1.1405,
+      "step": 2506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019079102433586e-05,
+      "loss": 1.0823,
+      "step": 2507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018141969030322e-05,
+      "loss": 1.0515,
+      "step": 2508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017204411300957e-05,
+      "loss": 1.0471,
+      "step": 2509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016266429289603e-05,
+      "loss": 1.1292,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015328023040388e-05,
+      "loss": 0.9596,
+      "step": 2511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014389192597476e-05,
+      "loss": 1.0529,
+      "step": 2512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901344993800504e-05,
+      "loss": 1.004,
+      "step": 2513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012510259307263e-05,
+      "loss": 1.1429,
+      "step": 2514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011570156548372e-05,
+      "loss": 1.0498,
+      "step": 2515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010629629772598e-05,
+      "loss": 1.1258,
+      "step": 2516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900968867902419e-05,
+      "loss": 1.0817,
+      "step": 2517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900874730434743e-05,
+      "loss": 1.0779,
+      "step": 2518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007805505786607e-05,
+      "loss": 1.0624,
+      "step": 2519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006863283386038e-05,
+      "loss": 0.9913,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900592063719005e-05,
+      "loss": 1.0693,
+      "step": 2521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004977567243005e-05,
+      "loss": 0.997,
+      "step": 2522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004034073589273e-05,
+      "loss": 0.9997,
+      "step": 2523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900309015627325e-05,
+      "loss": 1.0671,
+      "step": 2524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002145815339345e-05,
+      "loss": 1.0117,
+      "step": 2525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001201050831995e-05,
+      "loss": 1.0298,
+      "step": 2526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000255862795653e-05,
+      "loss": 1.1509,
+      "step": 2527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999310251274794e-05,
+      "loss": 1.0812,
+      "step": 2528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998364216313908e-05,
+      "loss": 1.0707,
+      "step": 2529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997417757957514e-05,
+      "loss": 1.0273,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899647087625014e-05,
+      "loss": 1.0836,
+      "step": 2531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995523571236337e-05,
+      "loss": 0.987,
+      "step": 2532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994575842960685e-05,
+      "loss": 1.0311,
+      "step": 2533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993627691467772e-05,
+      "loss": 1.0538,
+      "step": 2534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992679116802213e-05,
+      "loss": 1.0434,
+      "step": 2535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991730119008643e-05,
+      "loss": 1.0874,
+      "step": 2536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990780698131707e-05,
+      "loss": 1.0177,
+      "step": 2537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898983085421609e-05,
+      "loss": 0.9896,
+      "step": 2538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898888058730647e-05,
+      "loss": 1.0317,
+      "step": 2539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987929897447567e-05,
+      "loss": 0.9686,
+      "step": 2540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898697878468412e-05,
+      "loss": 1.1032,
+      "step": 2541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986027249060868e-05,
+      "loss": 1.1278,
+      "step": 2542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898507529062259e-05,
+      "loss": 1.0219,
+      "step": 2543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898412290941408e-05,
+      "loss": 1.0521,
+      "step": 2544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983170105480145e-05,
+      "loss": 1.0691,
+      "step": 2545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898221687886562e-05,
+      "loss": 1.0114,
+      "step": 2546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981263229615354e-05,
+      "loss": 1.0663,
+      "step": 2547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980309157774225e-05,
+      "loss": 1.0786,
+      "step": 2548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979354663387115e-05,
+      "loss": 1.1427,
+      "step": 2549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897839974649894e-05,
+      "loss": 1.1317,
+      "step": 2550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977444407154636e-05,
+      "loss": 0.9442,
+      "step": 2551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976488645399147e-05,
+      "loss": 1.0465,
+      "step": 2552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975532461277445e-05,
+      "loss": 1.0619,
+      "step": 2553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974575854834522e-05,
+      "loss": 1.0365,
+      "step": 2554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897361882611539e-05,
+      "loss": 1.0617,
+      "step": 2555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972661375165084e-05,
+      "loss": 1.0242,
+      "step": 2556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971703502028643e-05,
+      "loss": 1.1174,
+      "step": 2557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970745206751145e-05,
+      "loss": 1.0011,
+      "step": 2558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969786489377677e-05,
+      "loss": 1.0941,
+      "step": 2559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896882734995335e-05,
+      "loss": 0.9699,
+      "step": 2560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89678677885233e-05,
+      "loss": 1.0958,
+      "step": 2561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896690780513267e-05,
+      "loss": 1.0735,
+      "step": 2562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896594739982663e-05,
+      "loss": 1.0857,
+      "step": 2563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964986572650367e-05,
+      "loss": 0.9792,
+      "step": 2564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964025323649095e-05,
+      "loss": 1.0962,
+      "step": 2565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963063652868042e-05,
+      "loss": 1.0494,
+      "step": 2566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896210156035246e-05,
+      "loss": 1.0348,
+      "step": 2567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961139046147612e-05,
+      "loss": 1.0101,
+      "step": 2568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896017611029879e-05,
+      "loss": 1.1106,
+      "step": 2569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959212752851302e-05,
+      "loss": 1.0263,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895824897385048e-05,
+      "loss": 1.0857,
+      "step": 2571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957284773341662e-05,
+      "loss": 1.0962,
+      "step": 2572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956320151370224e-05,
+      "loss": 1.0106,
+      "step": 2573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955355107981558e-05,
+      "loss": 1.066,
+      "step": 2574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895438964322106e-05,
+      "loss": 1.0927,
+      "step": 2575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895342375713417e-05,
+      "loss": 1.0963,
+      "step": 2576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895245744976632e-05,
+      "loss": 0.9947,
+      "step": 2577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951490721162993e-05,
+      "loss": 1.0919,
+      "step": 2578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895052357136967e-05,
+      "loss": 1.0582,
+      "step": 2579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949556000431853e-05,
+      "loss": 1.0233,
+      "step": 2580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948588008395073e-05,
+      "loss": 1.0015,
+      "step": 2581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894761959530488e-05,
+      "loss": 1.0405,
+      "step": 2582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8946650761206835e-05,
+      "loss": 1.0784,
+      "step": 2583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894568150614652e-05,
+      "loss": 1.1075,
+      "step": 2584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944711830169553e-05,
+      "loss": 1.058,
+      "step": 2585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894374173332155e-05,
+      "loss": 0.9911,
+      "step": 2586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8942771215648163e-05,
+      "loss": 1.0395,
+      "step": 2587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894180027719505e-05,
+      "loss": 1.0648,
+      "step": 2588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89408289180079e-05,
+      "loss": 1.0085,
+      "step": 2589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893985713813242e-05,
+      "loss": 1.0742,
+      "step": 2590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893888493761433e-05,
+      "loss": 1.0877,
+      "step": 2591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893791231649938e-05,
+      "loss": 1.1138,
+      "step": 2592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893693927483333e-05,
+      "loss": 1.0534,
+      "step": 2593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935965812661964e-05,
+      "loss": 1.1065,
+      "step": 2594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934991930031086e-05,
+      "loss": 1.0533,
+      "step": 2595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934017626986522e-05,
+      "loss": 1.1278,
+      "step": 2596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933042903574112e-05,
+      "loss": 1.0917,
+      "step": 2597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932067759839726e-05,
+      "loss": 1.0737,
+      "step": 2598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931092195829237e-05,
+      "loss": 1.0456,
+      "step": 2599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930116211588556e-05,
+      "loss": 1.0676,
+      "step": 2600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89291398071636e-05,
+      "loss": 1.0465,
+      "step": 2601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928162982600312e-05,
+      "loss": 1.0771,
+      "step": 2602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892718573794466e-05,
+      "loss": 1.0386,
+      "step": 2603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892620807324262e-05,
+      "loss": 0.9906,
+      "step": 2604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925229988540198e-05,
+      "loss": 1.0756,
+      "step": 2605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924251483883408e-05,
+      "loss": 1.0644,
+      "step": 2606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923272559318297e-05,
+      "loss": 1.1145,
+      "step": 2607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892229321489092e-05,
+      "loss": 1.1148,
+      "step": 2608
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921313450647366e-05,
+      "loss": 1.0232,
+      "step": 2609
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920333266633733e-05,
+      "loss": 0.9551,
+      "step": 2610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919352662896136e-05,
+      "loss": 1.1021,
+      "step": 2611
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891837163948072e-05,
+      "loss": 1.1372,
+      "step": 2612
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891739019643364e-05,
+      "loss": 1.1161,
+      "step": 2613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891640833380108e-05,
+      "loss": 1.0861,
+      "step": 2614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915426051629236e-05,
+      "loss": 1.0071,
+      "step": 2615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914443349964334e-05,
+      "loss": 1.0569,
+      "step": 2616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89134602288526e-05,
+      "loss": 1.0487,
+      "step": 2617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89124766883403e-05,
+      "loss": 0.972,
+      "step": 2618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911492728473713e-05,
+      "loss": 1.0727,
+      "step": 2619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891050834929913e-05,
+      "loss": 1.0336,
+      "step": 2620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909523550862882e-05,
+      "loss": 1.0742,
+      "step": 2621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890853833321129e-05,
+      "loss": 1.0497,
+      "step": 2622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890755269639072e-05,
+      "loss": 1.0291,
+      "step": 2623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906566640447547e-05,
+      "loss": 1.0232,
+      "step": 2624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890558016542817e-05,
+      "loss": 1.0505,
+      "step": 2625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904593271378998e-05,
+      "loss": 1.0407,
+      "step": 2626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903605958346477e-05,
+      "loss": 1.0491,
+      "step": 2627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902618226377053e-05,
+      "loss": 0.9838,
+      "step": 2628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901630075517206e-05,
+      "loss": 1.0595,
+      "step": 2629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890064150581343e-05,
+      "loss": 1.1469,
+      "step": 2630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889965251731224e-05,
+      "loss": 1.0933,
+      "step": 2631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898663110060173e-05,
+      "loss": 1.0827,
+      "step": 2632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897673284103778e-05,
+      "loss": 1.12,
+      "step": 2633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889668303948963e-05,
+      "loss": 1.063,
+      "step": 2634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889569237626432e-05,
+      "loss": 1.0376,
+      "step": 2635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894701294474473e-05,
+      "loss": 1.1109,
+      "step": 2636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889370979416671e-05,
+      "loss": 1.065,
+      "step": 2637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892717875387686e-05,
+      "loss": 1.0444,
+      "step": 2638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891725538184076e-05,
+      "loss": 0.9994,
+      "step": 2639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889073278260257e-05,
+      "loss": 1.0601,
+      "step": 2640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888973960868988e-05,
+      "loss": 1.0596,
+      "step": 2641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888746016492734e-05,
+      "loss": 1.1671,
+      "step": 2642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887752006057893e-05,
+      "loss": 1.0627,
+      "step": 2643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886757577432117e-05,
+      "loss": 0.9976,
+      "step": 2644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88857627306622e-05,
+      "loss": 1.1079,
+      "step": 2645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884767465794953e-05,
+      "loss": 1.0433,
+      "step": 2646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883771782877206e-05,
+      "loss": 1.026,
+      "step": 2647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882775681955804e-05,
+      "loss": 1.0376,
+      "step": 2648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881779163077626e-05,
+      "loss": 1.0855,
+      "step": 2649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888078222628955e-05,
+      "loss": 1.0557,
+      "step": 2650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887978487163849e-05,
+      "loss": 1.0492,
+      "step": 2651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887878709917137e-05,
+      "loss": 1.0427,
+      "step": 2652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877788908935143e-05,
+      "loss": 1.0931,
+      "step": 2653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887679030097677e-05,
+      "loss": 1.1165,
+      "step": 2654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875791275343244e-05,
+      "loss": 1.0199,
+      "step": 2655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874791832081572e-05,
+      "loss": 1.0273,
+      "step": 2656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873791971238773e-05,
+      "loss": 1.0753,
+      "step": 2657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88727916928619e-05,
+      "loss": 0.9833,
+      "step": 2658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871790996998017e-05,
+      "loss": 1.0928,
+      "step": 2659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870789883694206e-05,
+      "loss": 1.086,
+      "step": 2660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869788352997573e-05,
+      "loss": 0.999,
+      "step": 2661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868786404955247e-05,
+      "loss": 1.0904,
+      "step": 2662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867784039614364e-05,
+      "loss": 1.1737,
+      "step": 2663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88667812570221e-05,
+      "loss": 1.0522,
+      "step": 2664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865778057225625e-05,
+      "loss": 1.0424,
+      "step": 2665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864774440272152e-05,
+      "loss": 1.1134,
+      "step": 2666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88637704062089e-05,
+      "loss": 1.0124,
+      "step": 2667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886276595508311e-05,
+      "loss": 0.9715,
+      "step": 2668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861761086942043e-05,
+      "loss": 1.0153,
+      "step": 2669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860755801832983e-05,
+      "loss": 1.0869,
+      "step": 2670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859750099803234e-05,
+      "loss": 1.1359,
+      "step": 2671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885874398090011e-05,
+      "loss": 1.0381,
+      "step": 2672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857737445170958e-05,
+      "loss": 1.1452,
+      "step": 2673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856730492663135e-05,
+      "loss": 1.132,
+      "step": 2674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885572312342402e-05,
+      "loss": 1.0319,
+      "step": 2675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885471533750101e-05,
+      "loss": 1.0858,
+      "step": 2676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885370713494153e-05,
+      "loss": 1.1263,
+      "step": 2677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852698515793017e-05,
+      "loss": 1.0428,
+      "step": 2678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851689480102924e-05,
+      "loss": 1.0053,
+      "step": 2679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850680027918732e-05,
+      "loss": 1.0591,
+      "step": 2680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849670159287942e-05,
+      "loss": 0.9665,
+      "step": 2681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848659874258063e-05,
+      "loss": 0.9994,
+      "step": 2682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847649172876636e-05,
+      "loss": 1.0253,
+      "step": 2683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846638055191214e-05,
+      "loss": 1.0876,
+      "step": 2684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884562652124938e-05,
+      "loss": 0.9685,
+      "step": 2685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884461457109872e-05,
+      "loss": 1.0064,
+      "step": 2686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843602204786857e-05,
+      "loss": 1.0045,
+      "step": 2687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884258942236142e-05,
+      "loss": 1.0956,
+      "step": 2688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841576223870063e-05,
+      "loss": 1.0864,
+      "step": 2689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840562609360466e-05,
+      "loss": 0.9972,
+      "step": 2690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883954857888031e-05,
+      "loss": 1.0622,
+      "step": 2691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838534132477317e-05,
+      "loss": 1.0174,
+      "step": 2692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883751927019922e-05,
+      "loss": 1.0907,
+      "step": 2693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836503992093763e-05,
+      "loss": 1.0535,
+      "step": 2694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835488298208723e-05,
+      "loss": 1.1397,
+      "step": 2695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834472188591893e-05,
+      "loss": 1.0735,
+      "step": 2696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883345566329108e-05,
+      "loss": 1.1035,
+      "step": 2697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832438722354112e-05,
+      "loss": 0.9826,
+      "step": 2698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831421365828842e-05,
+      "loss": 1.0252,
+      "step": 2699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883040359376314e-05,
+      "loss": 0.9495,
+      "step": 2700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882938540620489e-05,
+      "loss": 1.0443,
+      "step": 2701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828366803202005e-05,
+      "loss": 1.0528,
+      "step": 2702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882734778480241e-05,
+      "loss": 1.0412,
+      "step": 2703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826328351054055e-05,
+      "loss": 1.1613,
+      "step": 2704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825308502004907e-05,
+      "loss": 1.0453,
+      "step": 2705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824288237702947e-05,
+      "loss": 1.0564,
+      "step": 2706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882326755819619e-05,
+      "loss": 0.971,
+      "step": 2707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882224646353265e-05,
+      "loss": 0.9961,
+      "step": 2708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821224953760384e-05,
+      "loss": 1.0869,
+      "step": 2709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820203028927447e-05,
+      "loss": 1.1024,
+      "step": 2710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881918068908193e-05,
+      "loss": 1.0541,
+      "step": 2711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881815793427193e-05,
+      "loss": 1.0308,
+      "step": 2712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817134764545578e-05,
+      "loss": 1.027,
+      "step": 2713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881611117995101e-05,
+      "loss": 1.037,
+      "step": 2714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815087180536392e-05,
+      "loss": 1.0674,
+      "step": 2715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814062766349904e-05,
+      "loss": 1.0599,
+      "step": 2716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813037937439747e-05,
+      "loss": 1.0333,
+      "step": 2717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812012693854145e-05,
+      "loss": 1.1208,
+      "step": 2718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881098703564133e-05,
+      "loss": 1.1132,
+      "step": 2719
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809960962849577e-05,
+      "loss": 1.0457,
+      "step": 2720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880893447552715e-05,
+      "loss": 1.0439,
+      "step": 2721
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807907573722353e-05,
+      "loss": 1.1208,
+      "step": 2722
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806880257483508e-05,
+      "loss": 1.0779,
+      "step": 2723
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805852526858946e-05,
+      "loss": 1.0349,
+      "step": 2724
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880482438189703e-05,
+      "loss": 1.073,
+      "step": 2725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803795822646134e-05,
+      "loss": 1.0291,
+      "step": 2726
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802766849154656e-05,
+      "loss": 1.1232,
+      "step": 2727
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880173746147101e-05,
+      "loss": 0.9866,
+      "step": 2728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800707659643635e-05,
+      "loss": 1.0259,
+      "step": 2729
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799677443720984e-05,
+      "loss": 1.0239,
+      "step": 2730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798646813751524e-05,
+      "loss": 1.0491,
+      "step": 2731
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879761576978376e-05,
+      "loss": 1.0863,
+      "step": 2732
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87965843118662e-05,
+      "loss": 0.966,
+      "step": 2733
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795552440047374e-05,
+      "loss": 1.0546,
+      "step": 2734
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794520154375836e-05,
+      "loss": 1.0655,
+      "step": 2735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793487454900162e-05,
+      "loss": 1.0207,
+      "step": 2736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792454341668938e-05,
+      "loss": 1.0864,
+      "step": 2737
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791420814730776e-05,
+      "loss": 1.151,
+      "step": 2738
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790386874134306e-05,
+      "loss": 0.976,
+      "step": 2739
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789352519928178e-05,
+      "loss": 0.9515,
+      "step": 2740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878831775216106e-05,
+      "loss": 1.0481,
+      "step": 2741
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878728257088164e-05,
+      "loss": 1.0836,
+      "step": 2742
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878624697613863e-05,
+      "loss": 0.9875,
+      "step": 2743
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785210967980753e-05,
+      "loss": 0.9864,
+      "step": 2744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784174546456754e-05,
+      "loss": 1.0581,
+      "step": 2745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783137711615406e-05,
+      "loss": 1.0383,
+      "step": 2746
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878210046350549e-05,
+      "loss": 1.0452,
+      "step": 2747
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781062802175812e-05,
+      "loss": 1.0447,
+      "step": 2748
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780024727675196e-05,
+      "loss": 1.0677,
+      "step": 2749
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778986240052487e-05,
+      "loss": 1.1126,
+      "step": 2750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777947339356544e-05,
+      "loss": 1.0253,
+      "step": 2751
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776908025636257e-05,
+      "loss": 1.0774,
+      "step": 2752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775868298940526e-05,
+      "loss": 1.061,
+      "step": 2753
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774828159318273e-05,
+      "loss": 1.0248,
+      "step": 2754
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773787606818436e-05,
+      "loss": 1.0628,
+      "step": 2755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772746641489977e-05,
+      "loss": 1.0363,
+      "step": 2756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877170526338188e-05,
+      "loss": 1.0856,
+      "step": 2757
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770663472543135e-05,
+      "loss": 1.0177,
+      "step": 2758
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876962126902277e-05,
+      "loss": 1.052,
+      "step": 2759
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768578652869826e-05,
+      "loss": 1.0344,
+      "step": 2760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767535624133346e-05,
+      "loss": 1.0217,
+      "step": 2761
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876649218286242e-05,
+      "loss": 1.0333,
+      "step": 2762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876544832910614e-05,
+      "loss": 1.0619,
+      "step": 2763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764404062913627e-05,
+      "loss": 1.0801,
+      "step": 2764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763359384334005e-05,
+      "loss": 1.0398,
+      "step": 2765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762314293416438e-05,
+      "loss": 1.065,
+      "step": 2766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87612687902101e-05,
+      "loss": 1.0482,
+      "step": 2767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760222874764176e-05,
+      "loss": 1.1021,
+      "step": 2768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759176547127894e-05,
+      "loss": 1.0455,
+      "step": 2769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758129807350472e-05,
+      "loss": 1.01,
+      "step": 2770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875708265548117e-05,
+      "loss": 1.0294,
+      "step": 2771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756035091569253e-05,
+      "loss": 1.0986,
+      "step": 2772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754987115664013e-05,
+      "loss": 1.0193,
+      "step": 2773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753938727814768e-05,
+      "loss": 1.0436,
+      "step": 2774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752889928070837e-05,
+      "loss": 1.0206,
+      "step": 2775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751840716481573e-05,
+      "loss": 1.0704,
+      "step": 2776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750791093096343e-05,
+      "loss": 1.1008,
+      "step": 2777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749741057964538e-05,
+      "loss": 1.1148,
+      "step": 2778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874869061113556e-05,
+      "loss": 1.017,
+      "step": 2779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747639752658833e-05,
+      "loss": 1.0294,
+      "step": 2780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746588482583813e-05,
+      "loss": 1.0222,
+      "step": 2781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745536800959956e-05,
+      "loss": 1.0765,
+      "step": 2782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874448470783675e-05,
+      "loss": 1.0429,
+      "step": 2783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743432203263694e-05,
+      "loss": 1.1102,
+      "step": 2784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742379287290314e-05,
+      "loss": 1.0405,
+      "step": 2785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741325959966157e-05,
+      "loss": 1.0024,
+      "step": 2786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874027222134078e-05,
+      "loss": 1.1316,
+      "step": 2787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739218071463757e-05,
+      "loss": 1.0761,
+      "step": 2788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87381635103847e-05,
+      "loss": 1.0331,
+      "step": 2789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737108538153227e-05,
+      "loss": 0.9983,
+      "step": 2790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873605315481897e-05,
+      "loss": 1.1384,
+      "step": 2791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87349973604316e-05,
+      "loss": 1.0508,
+      "step": 2792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733941155040778e-05,
+      "loss": 1.0216,
+      "step": 2793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873288453869621e-05,
+      "loss": 1.023,
+      "step": 2794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731827511447614e-05,
+      "loss": 0.9733,
+      "step": 2795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730770073344727e-05,
+      "loss": 1.0076,
+      "step": 2796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87297122244373e-05,
+      "loss": 1.0303,
+      "step": 2797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728653964775105e-05,
+      "loss": 1.0202,
+      "step": 2798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727595294407944e-05,
+      "loss": 1.0672,
+      "step": 2799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872653621338562e-05,
+      "loss": 1.0282,
+      "step": 2800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725476721757974e-05,
+      "loss": 0.9459,
+      "step": 2801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724416819574853e-05,
+      "loss": 1.0933,
+      "step": 2802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872335650688613e-05,
+      "loss": 1.0185,
+      "step": 2803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722295783741695e-05,
+      "loss": 1.0749,
+      "step": 2804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721234650191458e-05,
+      "loss": 1.1188,
+      "step": 2805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720173106285345e-05,
+      "loss": 1.0951,
+      "step": 2806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871911115207331e-05,
+      "loss": 0.992,
+      "step": 2807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718048787605316e-05,
+      "loss": 1.0965,
+      "step": 2808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716986012931353e-05,
+      "loss": 1.0663,
+      "step": 2809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715922828101424e-05,
+      "loss": 1.1237,
+      "step": 2810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714859233165555e-05,
+      "loss": 1.0682,
+      "step": 2811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713795228173793e-05,
+      "loss": 0.9693,
+      "step": 2812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87127308131762e-05,
+      "loss": 1.0707,
+      "step": 2813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711665988222863e-05,
+      "loss": 1.0746,
+      "step": 2814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871060075336388e-05,
+      "loss": 1.101,
+      "step": 2815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709535108649374e-05,
+      "loss": 1.0661,
+      "step": 2816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708469054129485e-05,
+      "loss": 1.0786,
+      "step": 2817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707402589854383e-05,
+      "loss": 1.1031,
+      "step": 2818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706335715874235e-05,
+      "loss": 1.0008,
+      "step": 2819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705268432239244e-05,
+      "loss": 0.9834,
+      "step": 2820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704200738999634e-05,
+      "loss": 1.0607,
+      "step": 2821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870313263620564e-05,
+      "loss": 1.0627,
+      "step": 2822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702064123907513e-05,
+      "loss": 1.1231,
+      "step": 2823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700995202155535e-05,
+      "loss": 1.0806,
+      "step": 2824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699925871e-05,
+      "loss": 1.0251,
+      "step": 2825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698856130491224e-05,
+      "loss": 1.0018,
+      "step": 2826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869778598067954e-05,
+      "loss": 1.0352,
+      "step": 2827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86967154216153e-05,
+      "loss": 1.0428,
+      "step": 2828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695644453348877e-05,
+      "loss": 1.0227,
+      "step": 2829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694573075930664e-05,
+      "loss": 1.0459,
+      "step": 2830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693501289411072e-05,
+      "loss": 1.0478,
+      "step": 2831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692429093840527e-05,
+      "loss": 1.1738,
+      "step": 2832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691356489269485e-05,
+      "loss": 1.0739,
+      "step": 2833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869028347574841e-05,
+      "loss": 0.9909,
+      "step": 2834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689210053327793e-05,
+      "loss": 1.0797,
+      "step": 2835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868813622205814e-05,
+      "loss": 1.0898,
+      "step": 2836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687061981989974e-05,
+      "loss": 0.9045,
+      "step": 2837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868598733317385e-05,
+      "loss": 0.9956,
+      "step": 2838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868491227566032e-05,
+      "loss": 1.0575,
+      "step": 2839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868383680949998e-05,
+      "loss": 0.9696,
+      "step": 2840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682760934743423e-05,
+      "loss": 0.9689,
+      "step": 2841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681684651441284e-05,
+      "loss": 1.0877,
+      "step": 2842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680607959644193e-05,
+      "loss": 1.0971,
+      "step": 2843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679530859402818e-05,
+      "loss": 1.0139,
+      "step": 2844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678453350767835e-05,
+      "loss": 1.0858,
+      "step": 2845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677375433789943e-05,
+      "loss": 1.0783,
+      "step": 2846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676297108519867e-05,
+      "loss": 1.0905,
+      "step": 2847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675218375008343e-05,
+      "loss": 0.9735,
+      "step": 2848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674139233306123e-05,
+      "loss": 1.0453,
+      "step": 2849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673059683463987e-05,
+      "loss": 1.1126,
+      "step": 2850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867197972553273e-05,
+      "loss": 0.9886,
+      "step": 2851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670899359563167e-05,
+      "loss": 1.0397,
+      "step": 2852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8669818585606133e-05,
+      "loss": 1.0283,
+      "step": 2853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668737403712478e-05,
+      "loss": 1.1036,
+      "step": 2854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667655813933076e-05,
+      "loss": 1.0472,
+      "step": 2855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866657381631882e-05,
+      "loss": 1.0494,
+      "step": 2856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665491410920617e-05,
+      "loss": 1.0062,
+      "step": 2857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664408597789402e-05,
+      "loss": 1.0081,
+      "step": 2858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866332537697612e-05,
+      "loss": 1.0012,
+      "step": 2859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662241748531738e-05,
+      "loss": 1.048,
+      "step": 2860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661157712507248e-05,
+      "loss": 1.104,
+      "step": 2861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660073268953656e-05,
+      "loss": 1.1138,
+      "step": 2862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658988417921983e-05,
+      "loss": 1.0408,
+      "step": 2863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865790315946328e-05,
+      "loss": 1.0715,
+      "step": 2864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656817493628607e-05,
+      "loss": 1.0335,
+      "step": 2865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865573142046905e-05,
+      "loss": 1.0232,
+      "step": 2866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654644940035708e-05,
+      "loss": 0.9991,
+      "step": 2867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653558052379707e-05,
+      "loss": 1.0366,
+      "step": 2868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652470757552183e-05,
+      "loss": 1.0452,
+      "step": 2869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86513830556043e-05,
+      "loss": 1.0748,
+      "step": 2870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650294946587234e-05,
+      "loss": 1.083,
+      "step": 2871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649206430552187e-05,
+      "loss": 0.9846,
+      "step": 2872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648117507550374e-05,
+      "loss": 1.1253,
+      "step": 2873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647028177633027e-05,
+      "loss": 1.0512,
+      "step": 2874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864593844085141e-05,
+      "loss": 1.0015,
+      "step": 2875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644848297256797e-05,
+      "loss": 1.0355,
+      "step": 2876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643757746900478e-05,
+      "loss": 0.9924,
+      "step": 2877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642666789833765e-05,
+      "loss": 1.0539,
+      "step": 2878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864157542610799e-05,
+      "loss": 1.0447,
+      "step": 2879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640483655774513e-05,
+      "loss": 1.0594,
+      "step": 2880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639391478884695e-05,
+      "loss": 1.0219,
+      "step": 2881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638298895489932e-05,
+      "loss": 1.102,
+      "step": 2882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863720590564163e-05,
+      "loss": 1.0632,
+      "step": 2883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636112509391214e-05,
+      "loss": 1.0934,
+      "step": 2884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635018706790132e-05,
+      "loss": 1.1462,
+      "step": 2885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633924497889856e-05,
+      "loss": 1.036,
+      "step": 2886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632829882741866e-05,
+      "loss": 1.1409,
+      "step": 2887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631734861397667e-05,
+      "loss": 1.0748,
+      "step": 2888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630639433908783e-05,
+      "loss": 1.0187,
+      "step": 2889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629543600326757e-05,
+      "loss": 1.0988,
+      "step": 2890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862844736070315e-05,
+      "loss": 0.9702,
+      "step": 2891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627350715089545e-05,
+      "loss": 1.0069,
+      "step": 2892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862625366353754e-05,
+      "loss": 1.1071,
+      "step": 2893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625156206098754e-05,
+      "loss": 1.0734,
+      "step": 2894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8624058342824824e-05,
+      "loss": 1.1068,
+      "step": 2895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862296007376741e-05,
+      "loss": 1.0439,
+      "step": 2896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621861398978182e-05,
+      "loss": 1.0788,
+      "step": 2897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620762318508846e-05,
+      "loss": 1.0769,
+      "step": 2898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861966283241111e-05,
+      "loss": 1.0523,
+      "step": 2899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618562940736703e-05,
+      "loss": 1.0514,
+      "step": 2900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617462643537387e-05,
+      "loss": 1.0038,
+      "step": 2901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861636194086493e-05,
+      "loss": 1.0876,
+      "step": 2902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615260832771123e-05,
+      "loss": 1.0752,
+      "step": 2903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614159319307775e-05,
+      "loss": 1.0856,
+      "step": 2904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613057400526714e-05,
+      "loss": 1.0033,
+      "step": 2905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611955076479788e-05,
+      "loss": 1.1219,
+      "step": 2906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610852347218867e-05,
+      "loss": 1.0196,
+      "step": 2907
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609749212795835e-05,
+      "loss": 1.0435,
+      "step": 2908
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86086456732626e-05,
+      "loss": 1.047,
+      "step": 2909
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607541728671084e-05,
+      "loss": 1.0411,
+      "step": 2910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8606437379073227e-05,
+      "loss": 0.9768,
+      "step": 2911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605332624520997e-05,
+      "loss": 1.0085,
+      "step": 2912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604227465066375e-05,
+      "loss": 1.1028,
+      "step": 2913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8603121900761357e-05,
+      "loss": 1.0751,
+      "step": 2914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602015931657966e-05,
+      "loss": 1.0435,
+      "step": 2915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860090955780824e-05,
+      "loss": 1.086,
+      "step": 2916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599802779264236e-05,
+      "loss": 1.0579,
+      "step": 2917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598695596078032e-05,
+      "loss": 1.1084,
+      "step": 2918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597588008301723e-05,
+      "loss": 1.1058,
+      "step": 2919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596480015987428e-05,
+      "loss": 1.0574,
+      "step": 2920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859537161918727e-05,
+      "loss": 1.0723,
+      "step": 2921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594262817953414e-05,
+      "loss": 0.9519,
+      "step": 2922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593153612338022e-05,
+      "loss": 1.0832,
+      "step": 2923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859204400239329e-05,
+      "loss": 1.0201,
+      "step": 2924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859093398817143e-05,
+      "loss": 1.0262,
+      "step": 2925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589823569724666e-05,
+      "loss": 1.0084,
+      "step": 2926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588712747105248e-05,
+      "loss": 0.9789,
+      "step": 2927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587601520365442e-05,
+      "loss": 1.0182,
+      "step": 2928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586489889557537e-05,
+      "loss": 1.0962,
+      "step": 2929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585377854733836e-05,
+      "loss": 1.039,
+      "step": 2930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584265415946664e-05,
+      "loss": 1.1243,
+      "step": 2931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858315257324836e-05,
+      "loss": 1.0693,
+      "step": 2932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582039326691293e-05,
+      "loss": 1.0425,
+      "step": 2933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580925676327838e-05,
+      "loss": 1.0355,
+      "step": 2934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579811622210394e-05,
+      "loss": 1.0309,
+      "step": 2935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857869716439139e-05,
+      "loss": 1.1423,
+      "step": 2936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857758230292325e-05,
+      "loss": 1.0359,
+      "step": 2937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576467037858443e-05,
+      "loss": 1.028,
+      "step": 2938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857535136924944e-05,
+      "loss": 1.0841,
+      "step": 2939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574235297148734e-05,
+      "loss": 1.0262,
+      "step": 2940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8573118821608843e-05,
+      "loss": 1.0038,
+      "step": 2941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572001942682297e-05,
+      "loss": 1.092,
+      "step": 2942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857088466042165e-05,
+      "loss": 1.1072,
+      "step": 2943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569766974879472e-05,
+      "loss": 0.9793,
+      "step": 2944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856864888610835e-05,
+      "loss": 1.0625,
+      "step": 2945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85675303941609e-05,
+      "loss": 1.0862,
+      "step": 2946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566411499089745e-05,
+      "loss": 1.0544,
+      "step": 2947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565292200947533e-05,
+      "loss": 1.0055,
+      "step": 2948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564172499786926e-05,
+      "loss": 1.0774,
+      "step": 2949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563052395660617e-05,
+      "loss": 1.1825,
+      "step": 2950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85619318886213e-05,
+      "loss": 1.0644,
+      "step": 2951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560810978721703e-05,
+      "loss": 1.0172,
+      "step": 2952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559689666014567e-05,
+      "loss": 1.0531,
+      "step": 2953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558567950552654e-05,
+      "loss": 1.0836,
+      "step": 2954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557445832388743e-05,
+      "loss": 1.0262,
+      "step": 2955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556323311575632e-05,
+      "loss": 1.0614,
+      "step": 2956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555200388166134e-05,
+      "loss": 1.0804,
+      "step": 2957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554077062213092e-05,
+      "loss": 1.0444,
+      "step": 2958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552953333769357e-05,
+      "loss": 1.0186,
+      "step": 2959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551829202887806e-05,
+      "loss": 1.0612,
+      "step": 2960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855070466962133e-05,
+      "loss": 1.1055,
+      "step": 2961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854957973402284e-05,
+      "loss": 0.9982,
+      "step": 2962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548454396145273e-05,
+      "loss": 1.066,
+      "step": 2963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854732865604157e-05,
+      "loss": 1.0544,
+      "step": 2964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854620251376471e-05,
+      "loss": 1.0827,
+      "step": 2965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545075969367678e-05,
+      "loss": 1.08,
+      "step": 2966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854394902290347e-05,
+      "loss": 1.0389,
+      "step": 2967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542821674425128e-05,
+      "loss": 1.0574,
+      "step": 2968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541693923985684e-05,
+      "loss": 1.0478,
+      "step": 2969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540565771638207e-05,
+      "loss": 1.0308,
+      "step": 2970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539437217435782e-05,
+      "loss": 1.0858,
+      "step": 2971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538308261431502e-05,
+      "loss": 1.0596,
+      "step": 2972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537178903678495e-05,
+      "loss": 1.0315,
+      "step": 2973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536049144229895e-05,
+      "loss": 0.9931,
+      "step": 2974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853491898313886e-05,
+      "loss": 1.0648,
+      "step": 2975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533788420458575e-05,
+      "loss": 1.0261,
+      "step": 2976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532657456242224e-05,
+      "loss": 1.0708,
+      "step": 2977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531526090543033e-05,
+      "loss": 1.0513,
+      "step": 2978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530394323414227e-05,
+      "loss": 1.0552,
+      "step": 2979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529262154909063e-05,
+      "loss": 1.0215,
+      "step": 2980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528129585080808e-05,
+      "loss": 1.0973,
+      "step": 2981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526996613982758e-05,
+      "loss": 1.1144,
+      "step": 2982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525863241668217e-05,
+      "loss": 1.0073,
+      "step": 2983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524729468190514e-05,
+      "loss": 1.0167,
+      "step": 2984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523595293603e-05,
+      "loss": 1.0094,
+      "step": 2985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522460717959036e-05,
+      "loss": 1.0592,
+      "step": 2986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521325741312014e-05,
+      "loss": 1.0082,
+      "step": 2987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520190363715326e-05,
+      "loss": 1.0718,
+      "step": 2988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519054585222403e-05,
+      "loss": 1.0291,
+      "step": 2989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517918405886682e-05,
+      "loss": 1.1116,
+      "step": 2990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516781825761626e-05,
+      "loss": 1.0364,
+      "step": 2991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851564484490071e-05,
+      "loss": 1.0762,
+      "step": 2992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514507463357437e-05,
+      "loss": 1.0547,
+      "step": 2993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851336968118532e-05,
+      "loss": 1.099,
+      "step": 2994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512231498437897e-05,
+      "loss": 1.0288,
+      "step": 2995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851109291516872e-05,
+      "loss": 1.1034,
+      "step": 2996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509953931431363e-05,
+      "loss": 1.0636,
+      "step": 2997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508814547279415e-05,
+      "loss": 1.0992,
+      "step": 2998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507674762766496e-05,
+      "loss": 1.0714,
+      "step": 2999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506534577946226e-05,
+      "loss": 1.0623,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505393992872256e-05,
+      "loss": 0.9537,
+      "step": 3001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504253007598255e-05,
+      "loss": 1.0009,
+      "step": 3002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850311162217791e-05,
+      "loss": 1.0488,
+      "step": 3003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501969836664927e-05,
+      "loss": 1.0842,
+      "step": 3004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500827651113025e-05,
+      "loss": 1.0755,
+      "step": 3005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499685065575946e-05,
+      "loss": 0.9775,
+      "step": 3006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849854208010746e-05,
+      "loss": 1.0638,
+      "step": 3007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849739869476134e-05,
+      "loss": 1.0895,
+      "step": 3008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496254909591387e-05,
+      "loss": 1.1048,
+      "step": 3009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849511072465142e-05,
+      "loss": 1.0595,
+      "step": 3010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493966139995278e-05,
+      "loss": 1.0678,
+      "step": 3011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849282115567681e-05,
+      "loss": 1.0634,
+      "step": 3012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849167577174989e-05,
+      "loss": 1.1412,
+      "step": 3013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849052998826842e-05,
+      "loss": 1.0262,
+      "step": 3014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8489383805286308e-05,
+      "loss": 1.0019,
+      "step": 3015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488237222857484e-05,
+      "loss": 1.0195,
+      "step": 3016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487090241035894e-05,
+      "loss": 0.9871,
+      "step": 3017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8485942859875515e-05,
+      "loss": 0.9838,
+      "step": 3018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484795079430324e-05,
+      "loss": 1.0342,
+      "step": 3019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483646899754334e-05,
+      "loss": 1.1138,
+      "step": 3020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482498320901565e-05,
+      "loss": 1.0706,
+      "step": 3021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481349342926065e-05,
+      "loss": 1.1913,
+      "step": 3022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84801999658819e-05,
+      "loss": 1.0786,
+      "step": 3023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479050189823137e-05,
+      "loss": 0.9744,
+      "step": 3024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477900014803888e-05,
+      "loss": 1.0853,
+      "step": 3025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476749440878267e-05,
+      "loss": 0.9815,
+      "step": 3026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475598468100413e-05,
+      "loss": 0.9983,
+      "step": 3027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847444709652448e-05,
+      "loss": 1.0763,
+      "step": 3028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473295326204644e-05,
+      "loss": 1.1008,
+      "step": 3029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472143157195098e-05,
+      "loss": 0.9864,
+      "step": 3030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847099058955006e-05,
+      "loss": 1.1659,
+      "step": 3031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469837623323755e-05,
+      "loss": 0.966,
+      "step": 3032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8468684258570432e-05,
+      "loss": 1.0255,
+      "step": 3033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467530495344367e-05,
+      "loss": 0.9847,
+      "step": 3034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846637633369984e-05,
+      "loss": 1.0035,
+      "step": 3035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465221773691158e-05,
+      "loss": 1.0727,
+      "step": 3036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846406681537265e-05,
+      "loss": 1.0952,
+      "step": 3037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462911458798654e-05,
+      "loss": 1.0531,
+      "step": 3038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846175570402354e-05,
+      "loss": 1.0342,
+      "step": 3039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460599551101683e-05,
+      "loss": 1.0467,
+      "step": 3040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459443000087487e-05,
+      "loss": 1.0583,
+      "step": 3041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845828605103537e-05,
+      "loss": 1.1576,
+      "step": 3042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457128703999762e-05,
+      "loss": 1.0031,
+      "step": 3043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455970959035127e-05,
+      "loss": 1.1298,
+      "step": 3044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454812816195938e-05,
+      "loss": 0.9972,
+      "step": 3045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845365427553669e-05,
+      "loss": 1.0386,
+      "step": 3046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452495337111892e-05,
+      "loss": 1.0651,
+      "step": 3047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845133600097608e-05,
+      "loss": 0.9938,
+      "step": 3048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450176267183796e-05,
+      "loss": 1.0826,
+      "step": 3049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449016135789613e-05,
+      "loss": 1.0753,
+      "step": 3050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844785560684812e-05,
+      "loss": 1.0557,
+      "step": 3051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446694680413914e-05,
+      "loss": 1.033,
+      "step": 3052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844553335654163e-05,
+      "loss": 1.0637,
+      "step": 3053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444371635285908e-05,
+      "loss": 0.967,
+      "step": 3054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844320951670141e-05,
+      "loss": 1.0493,
+      "step": 3055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844204700084281e-05,
+      "loss": 0.9899,
+      "step": 3056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440884087764815e-05,
+      "loss": 1.0528,
+      "step": 3057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439720777522144e-05,
+      "loss": 1.0164,
+      "step": 3058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438557070169526e-05,
+      "loss": 1.0619,
+      "step": 3059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437392965761724e-05,
+      "loss": 1.0201,
+      "step": 3060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843622846435351e-05,
+      "loss": 1.0728,
+      "step": 3061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843506356599967e-05,
+      "loss": 1.0384,
+      "step": 3062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433898270755028e-05,
+      "loss": 1.0819,
+      "step": 3063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432732578674398e-05,
+      "loss": 1.0747,
+      "step": 3064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431566489812643e-05,
+      "loss": 1.0884,
+      "step": 3065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430400004224623e-05,
+      "loss": 1.0515,
+      "step": 3066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842923312196523e-05,
+      "loss": 1.0192,
+      "step": 3067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428065843089363e-05,
+      "loss": 1.0035,
+      "step": 3068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426898167651946e-05,
+      "loss": 1.1489,
+      "step": 3069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842573009570792e-05,
+      "loss": 1.0292,
+      "step": 3070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842456162731225e-05,
+      "loss": 1.0645,
+      "step": 3071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423392762519915e-05,
+      "loss": 1.1235,
+      "step": 3072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842222350138591e-05,
+      "loss": 1.078,
+      "step": 3073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8421053843965252e-05,
+      "loss": 1.0946,
+      "step": 3074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419883790312978e-05,
+      "loss": 1.1016,
+      "step": 3075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418713340484144e-05,
+      "loss": 1.0101,
+      "step": 3076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417542494533812e-05,
+      "loss": 1.038,
+      "step": 3077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416371252517086e-05,
+      "loss": 1.0395,
+      "step": 3078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841519961448907e-05,
+      "loss": 1.1432,
+      "step": 3079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414027580504897e-05,
+      "loss": 1.0492,
+      "step": 3080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412855150619705e-05,
+      "loss": 1.0966,
+      "step": 3081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411682324888668e-05,
+      "loss": 1.0722,
+      "step": 3082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841050910336697e-05,
+      "loss": 1.0735,
+      "step": 3083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409335486109805e-05,
+      "loss": 1.0712,
+      "step": 3084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408161473172406e-05,
+      "loss": 1.1091,
+      "step": 3085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406987064610006e-05,
+      "loss": 1.0954,
+      "step": 3086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840581226047787e-05,
+      "loss": 1.039,
+      "step": 3087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404637060831268e-05,
+      "loss": 1.0599,
+      "step": 3088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403461465725502e-05,
+      "loss": 1.0743,
+      "step": 3089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402285475215886e-05,
+      "loss": 1.0853,
+      "step": 3090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401109089357745e-05,
+      "loss": 1.0261,
+      "step": 3091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8399932308206444e-05,
+      "loss": 0.9926,
+      "step": 3092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398755131817345e-05,
+      "loss": 1.0191,
+      "step": 3093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397577560245842e-05,
+      "loss": 1.0929,
+      "step": 3094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396399593547336e-05,
+      "loss": 1.0667,
+      "step": 3095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839522123177726e-05,
+      "loss": 1.0852,
+      "step": 3096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394042474991054e-05,
+      "loss": 0.9843,
+      "step": 3097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392863323244185e-05,
+      "loss": 1.0946,
+      "step": 3098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391683776592135e-05,
+      "loss": 1.0747,
+      "step": 3099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83905038350904e-05,
+      "loss": 1.1255,
+      "step": 3100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389323498794507e-05,
+      "loss": 0.9978,
+      "step": 3101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388142767759983e-05,
+      "loss": 1.0596,
+      "step": 3102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386961642042397e-05,
+      "loss": 1.0277,
+      "step": 3103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838578012169731e-05,
+      "loss": 1.0258,
+      "step": 3104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838459820678033e-05,
+      "loss": 1.0419,
+      "step": 3105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838341589734706e-05,
+      "loss": 1.0495,
+      "step": 3106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838223319345313e-05,
+      "loss": 1.0714,
+      "step": 3107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381050095154193e-05,
+      "loss": 1.0236,
+      "step": 3108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379866602505916e-05,
+      "loss": 1.0276,
+      "step": 3109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378682715563986e-05,
+      "loss": 1.0709,
+      "step": 3110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377498434384106e-05,
+      "loss": 1.0465,
+      "step": 3111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376313759022e-05,
+      "loss": 0.9898,
+      "step": 3112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837512868953341e-05,
+      "loss": 1.0131,
+      "step": 3113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373943225974097e-05,
+      "loss": 1.1059,
+      "step": 3114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8372757368399837e-05,
+      "loss": 1.0127,
+      "step": 3115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371571116866433e-05,
+      "loss": 1.0807,
+      "step": 3116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370384471429698e-05,
+      "loss": 1.1287,
+      "step": 3117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836919743214546e-05,
+      "loss": 1.0847,
+      "step": 3118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368009999069587e-05,
+      "loss": 1.0774,
+      "step": 3119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836682217225794e-05,
+      "loss": 0.9686,
+      "step": 3120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8365633951766412e-05,
+      "loss": 1.0234,
+      "step": 3121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836444533765091e-05,
+      "loss": 1.0365,
+      "step": 3122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363256329967364e-05,
+      "loss": 1.0508,
+      "step": 3123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836206692877172e-05,
+      "loss": 1.0219,
+      "step": 3124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836087713411994e-05,
+      "loss": 1.0766,
+      "step": 3125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359686946068007e-05,
+      "loss": 1.0137,
+      "step": 3126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358496364671922e-05,
+      "loss": 1.0785,
+      "step": 3127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835730538998771e-05,
+      "loss": 1.0046,
+      "step": 3128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356114022071402e-05,
+      "loss": 1.0262,
+      "step": 3129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354922260979057e-05,
+      "loss": 1.0222,
+      "step": 3130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353730106766754e-05,
+      "loss": 1.089,
+      "step": 3131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352537559490583e-05,
+      "loss": 1.062,
+      "step": 3132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351344619206653e-05,
+      "loss": 0.9949,
+      "step": 3133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835015128597111e-05,
+      "loss": 1.0233,
+      "step": 3134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348957559840083e-05,
+      "loss": 0.9986,
+      "step": 3135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347763440869752e-05,
+      "loss": 1.0978,
+      "step": 3136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346568929116305e-05,
+      "loss": 1.0263,
+      "step": 3137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345374024635937e-05,
+      "loss": 1.0283,
+      "step": 3138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834417872748488e-05,
+      "loss": 0.9663,
+      "step": 3139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834298303771937e-05,
+      "loss": 1.0953,
+      "step": 3140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834178695539567e-05,
+      "loss": 1.117,
+      "step": 3141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834059048057006e-05,
+      "loss": 0.9804,
+      "step": 3142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339393613298834e-05,
+      "loss": 1.0813,
+      "step": 3143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833819635363831e-05,
+      "loss": 1.0268,
+      "step": 3144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336998701644823e-05,
+      "loss": 1.1129,
+      "step": 3145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833580065737472e-05,
+      "loss": 1.0445,
+      "step": 3146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334602220884378e-05,
+      "loss": 1.0438,
+      "step": 3147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833340339223018e-05,
+      "loss": 1.073,
+      "step": 3148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332204171468543e-05,
+      "loss": 1.0078,
+      "step": 3149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331004558655887e-05,
+      "loss": 1.0186,
+      "step": 3150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329804553848657e-05,
+      "loss": 1.0007,
+      "step": 3151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328604157103318e-05,
+      "loss": 0.9926,
+      "step": 3152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327403368476352e-05,
+      "loss": 1.0177,
+      "step": 3153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326202188024255e-05,
+      "loss": 1.0073,
+      "step": 3154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832500061580355e-05,
+      "loss": 1.1554,
+      "step": 3155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323798651870773e-05,
+      "loss": 1.036,
+      "step": 3156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322596296282474e-05,
+      "loss": 1.1158,
+      "step": 3157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832139354909524e-05,
+      "loss": 1.1025,
+      "step": 3158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320190410365647e-05,
+      "loss": 1.1016,
+      "step": 3159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831898688015032e-05,
+      "loss": 1.0562,
+      "step": 3160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317782958505878e-05,
+      "loss": 1.0008,
+      "step": 3161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316578645488972e-05,
+      "loss": 1.0697,
+      "step": 3162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315373941156275e-05,
+      "loss": 1.108,
+      "step": 3163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314168845564457e-05,
+      "loss": 1.0243,
+      "step": 3164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312963358770232e-05,
+      "loss": 1.0062,
+      "step": 3165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8311757480830316e-05,
+      "loss": 0.9962,
+      "step": 3166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310551211801452e-05,
+      "loss": 1.1458,
+      "step": 3167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830934455174039e-05,
+      "loss": 1.0284,
+      "step": 3168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308137500703916e-05,
+      "loss": 0.9891,
+      "step": 3169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306930058748828e-05,
+      "loss": 1.0153,
+      "step": 3170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305722225931927e-05,
+      "loss": 1.0734,
+      "step": 3171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304514002310044e-05,
+      "loss": 1.0529,
+      "step": 3172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830330538794004e-05,
+      "loss": 1.0245,
+      "step": 3173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302096382878782e-05,
+      "loss": 1.0902,
+      "step": 3174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830088698718315e-05,
+      "loss": 1.0254,
+      "step": 3175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829967720091005e-05,
+      "loss": 1.0619,
+      "step": 3176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298467024116408e-05,
+      "loss": 1.0918,
+      "step": 3177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297256456859168e-05,
+      "loss": 1.1356,
+      "step": 3178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8296045499195285e-05,
+      "loss": 1.04,
+      "step": 3179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829483415118174e-05,
+      "loss": 1.0582,
+      "step": 3180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293622412875528e-05,
+      "loss": 1.1337,
+      "step": 3181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292410284333672e-05,
+      "loss": 1.1183,
+      "step": 3182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291197765613197e-05,
+      "loss": 1.0209,
+      "step": 3183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828998485677116e-05,
+      "loss": 1.1078,
+      "step": 3184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288771557864624e-05,
+      "loss": 1.0456,
+      "step": 3185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287557868950687e-05,
+      "loss": 1.0609,
+      "step": 3186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828634379008645e-05,
+      "loss": 1.0384,
+      "step": 3187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828512932132904e-05,
+      "loss": 0.9923,
+      "step": 3188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283914462735608e-05,
+      "loss": 1.0683,
+      "step": 3189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282699214363308e-05,
+      "loss": 1.0436,
+      "step": 3190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828148357626932e-05,
+      "loss": 1.0719,
+      "step": 3191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280267548510844e-05,
+      "loss": 1.0687,
+      "step": 3192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8279051131145097e-05,
+      "loss": 1.0565,
+      "step": 3193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277834324229317e-05,
+      "loss": 0.996,
+      "step": 3194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827661712782076e-05,
+      "loss": 1.042,
+      "step": 3195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827539954197669e-05,
+      "loss": 1.0666,
+      "step": 3196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82741815667544e-05,
+      "loss": 1.0651,
+      "step": 3197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272963202211205e-05,
+      "loss": 1.0473,
+      "step": 3198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271744448404426e-05,
+      "loss": 1.0671,
+      "step": 3199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827052530539141e-05,
+      "loss": 1.025,
+      "step": 3200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826930577322952e-05,
+      "loss": 1.0729,
+      "step": 3201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826808585197614e-05,
+      "loss": 1.0832,
+      "step": 3202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266865541688673e-05,
+      "loss": 1.0536,
+      "step": 3203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826564484242453e-05,
+      "loss": 1.0332,
+      "step": 3204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8264423754241152e-05,
+      "loss": 1.0565,
+      "step": 3205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263202277195998e-05,
+      "loss": 1.0899,
+      "step": 3206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261980411346532e-05,
+      "loss": 0.9746,
+      "step": 3207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260758156750252e-05,
+      "loss": 1.0683,
+      "step": 3208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259535513464674e-05,
+      "loss": 1.0263,
+      "step": 3209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258312481547315e-05,
+      "loss": 1.0567,
+      "step": 3210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257089061055726e-05,
+      "loss": 1.0596,
+      "step": 3211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255865252047474e-05,
+      "loss": 1.0741,
+      "step": 3212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254641054580142e-05,
+      "loss": 1.0837,
+      "step": 3213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825341646871133e-05,
+      "loss": 1.1043,
+      "step": 3214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825219149449866e-05,
+      "loss": 1.0013,
+      "step": 3215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250966131999767e-05,
+      "loss": 1.1065,
+      "step": 3216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824974038127231e-05,
+      "loss": 1.0804,
+      "step": 3217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824851424237396e-05,
+      "loss": 1.0849,
+      "step": 3218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247287715362413e-05,
+      "loss": 1.1516,
+      "step": 3219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824606080029538e-05,
+      "loss": 1.1133,
+      "step": 3220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824483349723059e-05,
+      "loss": 1.048,
+      "step": 3221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243605806225793e-05,
+      "loss": 1.0398,
+      "step": 3222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824237772733875e-05,
+      "loss": 1.0754,
+      "step": 3223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824114926062725e-05,
+      "loss": 1.0526,
+      "step": 3224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823992040614909e-05,
+      "loss": 1.004,
+      "step": 3225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238691163962093e-05,
+      "loss": 1.0603,
+      "step": 3226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237461534124102e-05,
+      "loss": 1.0647,
+      "step": 3227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236231516692972e-05,
+      "loss": 1.0338,
+      "step": 3228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235001111726577e-05,
+      "loss": 1.0106,
+      "step": 3229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233770319282808e-05,
+      "loss": 0.9773,
+      "step": 3230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8232539139419582e-05,
+      "loss": 1.0083,
+      "step": 3231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231307572194827e-05,
+      "loss": 1.0778,
+      "step": 3232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823007561766649e-05,
+      "loss": 1.1027,
+      "step": 3233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228843275892537e-05,
+      "loss": 1.1337,
+      "step": 3234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822761054693096e-05,
+      "loss": 1.0702,
+      "step": 3235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226377430839754e-05,
+      "loss": 1.0734,
+      "step": 3236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225143927676942e-05,
+      "loss": 1.0399,
+      "step": 3237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223910037500562e-05,
+      "loss": 1.0687,
+      "step": 3238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222675760368676e-05,
+      "loss": 1.0791,
+      "step": 3239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221441096339356e-05,
+      "loss": 0.976,
+      "step": 3240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220206045470702e-05,
+      "loss": 1.0811,
+      "step": 3241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218970607820815e-05,
+      "loss": 0.9962,
+      "step": 3242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217734783447837e-05,
+      "loss": 1.0797,
+      "step": 3243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821649857240991e-05,
+      "loss": 1.0503,
+      "step": 3244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215261974765206e-05,
+      "loss": 1.0899,
+      "step": 3245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214024990571904e-05,
+      "loss": 1.0826,
+      "step": 3246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212787619888212e-05,
+      "loss": 0.9885,
+      "step": 3247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211549862772345e-05,
+      "loss": 1.0056,
+      "step": 3248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210311719282548e-05,
+      "loss": 1.1051,
+      "step": 3249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820907318947708e-05,
+      "loss": 1.0288,
+      "step": 3250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820783427341421e-05,
+      "loss": 0.9552,
+      "step": 3251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206594971152242e-05,
+      "loss": 1.0569,
+      "step": 3252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205355282749478e-05,
+      "loss": 0.9989,
+      "step": 3253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204115208264258e-05,
+      "loss": 1.0889,
+      "step": 3254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202874747754923e-05,
+      "loss": 1.0814,
+      "step": 3255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201633901279843e-05,
+      "loss": 1.063,
+      "step": 3256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200392668897403e-05,
+      "loss": 1.087,
+      "step": 3257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199151050666004e-05,
+      "loss": 1.0738,
+      "step": 3258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197909046644074e-05,
+      "loss": 1.1079,
+      "step": 3259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8196666656890044e-05,
+      "loss": 1.065,
+      "step": 3260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819542388146237e-05,
+      "loss": 0.9429,
+      "step": 3261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194180720419538e-05,
+      "loss": 1.0251,
+      "step": 3262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192937173820035e-05,
+      "loss": 1.0269,
+      "step": 3263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191693241722374e-05,
+      "loss": 0.9688,
+      "step": 3264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190448924185088e-05,
+      "loss": 1.1194,
+      "step": 3265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818920422126672e-05,
+      "loss": 1.0677,
+      "step": 3266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818795913302584e-05,
+      "loss": 1.0102,
+      "step": 3267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818671365952103e-05,
+      "loss": 1.0725,
+      "step": 3268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185467800810895e-05,
+      "loss": 1.0357,
+      "step": 3269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184221556954055e-05,
+      "loss": 1.0571,
+      "step": 3270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182974928009152e-05,
+      "loss": 1.0914,
+      "step": 3271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181727914034835e-05,
+      "loss": 0.9951,
+      "step": 3272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180480515089787e-05,
+      "loss": 0.9747,
+      "step": 3273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81792327312327e-05,
+      "loss": 0.9991,
+      "step": 3274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817798456252228e-05,
+      "loss": 1.0841,
+      "step": 3275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817673600901726e-05,
+      "loss": 1.0243,
+      "step": 3276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175487070776386e-05,
+      "loss": 0.9903,
+      "step": 3277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817423774785843e-05,
+      "loss": 1.0925,
+      "step": 3278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817298804032217e-05,
+      "loss": 1.1119,
+      "step": 3279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171737948226407e-05,
+      "loss": 1.1158,
+      "step": 3280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170487471629964e-05,
+      "loss": 1.0395,
+      "step": 3281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169236610591676e-05,
+      "loss": 1.0979,
+      "step": 3282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167985365170403e-05,
+      "loss": 1.0477,
+      "step": 3283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8166733735425017e-05,
+      "loss": 1.0193,
+      "step": 3284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816548172141441e-05,
+      "loss": 1.1288,
+      "step": 3285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164229323197494e-05,
+      "loss": 0.9698,
+      "step": 3286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162976540833193e-05,
+      "loss": 1.0596,
+      "step": 3287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816172337438046e-05,
+      "loss": 1.0856,
+      "step": 3288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160469823898253e-05,
+      "loss": 1.0343,
+      "step": 3289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159215889445562e-05,
+      "loss": 0.9899,
+      "step": 3290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157961571081385e-05,
+      "loss": 1.1016,
+      "step": 3291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8156706868864735e-05,
+      "loss": 1.0394,
+      "step": 3292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815545178285466e-05,
+      "loss": 0.9982,
+      "step": 3293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154196313110203e-05,
+      "loss": 1.0065,
+      "step": 3294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152940459690445e-05,
+      "loss": 1.0295,
+      "step": 3295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151684222654476e-05,
+      "loss": 1.0449,
+      "step": 3296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81504276020614e-05,
+      "loss": 0.9965,
+      "step": 3297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814917059797035e-05,
+      "loss": 1.0139,
+      "step": 3298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814791321044047e-05,
+      "loss": 1.0735,
+      "step": 3299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814665543953092e-05,
+      "loss": 1.094,
+      "step": 3300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145397285300882e-05,
+      "loss": 0.9816,
+      "step": 3301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814413874780956e-05,
+      "loss": 0.9351,
+      "step": 3302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142879827116168e-05,
+      "loss": 1.1377,
+      "step": 3303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814162052327994e-05,
+      "loss": 1.0645,
+      "step": 3304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140360836360136e-05,
+      "loss": 1.018,
+      "step": 3305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8139100766416014e-05,
+      "loss": 1.0817,
+      "step": 3306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137840313506877e-05,
+      "loss": 1.0608,
+      "step": 3307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136579477692028e-05,
+      "loss": 1.1008,
+      "step": 3308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135318259030788e-05,
+      "loss": 1.0986,
+      "step": 3309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813405665758251e-05,
+      "loss": 0.9927,
+      "step": 3310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132794673406545e-05,
+      "loss": 1.0967,
+      "step": 3311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131532306562276e-05,
+      "loss": 1.1012,
+      "step": 3312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130269557109104e-05,
+      "loss": 1.0225,
+      "step": 3313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812900642510644e-05,
+      "loss": 1.0204,
+      "step": 3314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812774291061372e-05,
+      "loss": 1.0912,
+      "step": 3315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81264790136904e-05,
+      "loss": 1.0205,
+      "step": 3316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812521473439594e-05,
+      "loss": 1.0822,
+      "step": 3317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123950072789832e-05,
+      "loss": 1.0455,
+      "step": 3318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122685028931587e-05,
+      "loss": 0.9787,
+      "step": 3319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812141960288072e-05,
+      "loss": 1.131,
+      "step": 3320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812015379469677e-05,
+      "loss": 1.097,
+      "step": 3321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811888760443931e-05,
+      "loss": 1.0777,
+      "step": 3322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8117621032167907e-05,
+      "loss": 1.049,
+      "step": 3323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116354077942158e-05,
+      "loss": 1.0906,
+      "step": 3324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115086741821678e-05,
+      "loss": 1.0264,
+      "step": 3325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81138190238661e-05,
+      "loss": 1.0398,
+      "step": 3326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8112550924135066e-05,
+      "loss": 1.0419,
+      "step": 3327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111282442688253e-05,
+      "loss": 1.0625,
+      "step": 3328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8110013579585337e-05,
+      "loss": 0.9956,
+      "step": 3329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108744334886028e-05,
+      "loss": 1.0211,
+      "step": 3330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107474708650042e-05,
+      "loss": 1.0392,
+      "step": 3331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106204700937125e-05,
+      "loss": 1.0818,
+      "step": 3332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104934311807027e-05,
+      "loss": 1.0788,
+      "step": 3333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810366354131953e-05,
+      "loss": 1.0678,
+      "step": 3334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8102392389534416e-05,
+      "loss": 1.0059,
+      "step": 3335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101120856511503e-05,
+      "loss": 1.1013,
+      "step": 3336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099848942310625e-05,
+      "loss": 1.0529,
+      "step": 3337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809857664699162e-05,
+      "loss": 1.0407,
+      "step": 3338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8097303970614355e-05,
+      "loss": 1.0898,
+      "step": 3339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096030913238712e-05,
+      "loss": 1.0434,
+      "step": 3340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094757474924598e-05,
+      "loss": 1.1038,
+      "step": 3341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8093483655731924e-05,
+      "loss": 1.0418,
+      "step": 3342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809220945572063e-05,
+      "loss": 1.0597,
+      "step": 3343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090934874950665e-05,
+      "loss": 1.0325,
+      "step": 3344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089659913482006e-05,
+      "loss": 1.021,
+      "step": 3345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088384571374647e-05,
+      "loss": 1.0269,
+      "step": 3346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808710884868859e-05,
+      "loss": 1.0105,
+      "step": 3347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808583274548386e-05,
+      "loss": 0.9975,
+      "step": 3348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084556261820502e-05,
+      "loss": 1.0677,
+      "step": 3349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808327939775858e-05,
+      "loss": 1.0962,
+      "step": 3350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082002153358174e-05,
+      "loss": 1.1158,
+      "step": 3351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8080724528679382e-05,
+      "loss": 0.9935,
+      "step": 3352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079446523782312e-05,
+      "loss": 1.0397,
+      "step": 3353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078168138727105e-05,
+      "loss": 1.1867,
+      "step": 3354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8076889373573913e-05,
+      "loss": 1.0713,
+      "step": 3355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80756102283829e-05,
+      "loss": 0.9856,
+      "step": 3356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074330703214257e-05,
+      "loss": 1.0473,
+      "step": 3357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073050798128183e-05,
+      "loss": 0.9544,
+      "step": 3358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071770513184906e-05,
+      "loss": 1.0195,
+      "step": 3359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.807048984844467e-05,
+      "loss": 1.0574,
+      "step": 3360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069208803967724e-05,
+      "loss": 1.0775,
+      "step": 3361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067927379814348e-05,
+      "loss": 1.041,
+      "step": 3362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806664557604484e-05,
+      "loss": 0.968,
+      "step": 3363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806536339271951e-05,
+      "loss": 1.0747,
+      "step": 3364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064080829898688e-05,
+      "loss": 0.998,
+      "step": 3365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062797887642716e-05,
+      "loss": 1.0842,
+      "step": 3366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061514566011974e-05,
+      "loss": 0.9552,
+      "step": 3367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806023086506683e-05,
+      "loss": 1.0248,
+      "step": 3368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805894678486769e-05,
+      "loss": 1.0156,
+      "step": 3369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805766232547498e-05,
+      "loss": 1.0893,
+      "step": 3370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8056377486949127e-05,
+      "loss": 1.078,
+      "step": 3371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055092269350593e-05,
+      "loss": 1.0394,
+      "step": 3372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053806672739848e-05,
+      "loss": 1.0049,
+      "step": 3373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805252069717738e-05,
+      "loss": 0.997,
+      "step": 3374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051234342723704e-05,
+      "loss": 0.9685,
+      "step": 3375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049947609439345e-05,
+      "loss": 1.043,
+      "step": 3376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048660497384837e-05,
+      "loss": 1.0592,
+      "step": 3377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047373006620754e-05,
+      "loss": 1.1169,
+      "step": 3378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804608513720767e-05,
+      "loss": 1.0156,
+      "step": 3379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044796889206185e-05,
+      "loss": 1.0342,
+      "step": 3380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804350826267691e-05,
+      "loss": 1.0455,
+      "step": 3381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804221925768048e-05,
+      "loss": 1.0579,
+      "step": 3382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804092987427755e-05,
+      "loss": 1.0113,
+      "step": 3383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039640112528783e-05,
+      "loss": 0.9982,
+      "step": 3384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803834997249487e-05,
+      "loss": 1.0065,
+      "step": 3385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037059454236514e-05,
+      "loss": 1.0536,
+      "step": 3386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803576855781444e-05,
+      "loss": 1.1234,
+      "step": 3387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034477283289375e-05,
+      "loss": 1.0539,
+      "step": 3388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803318563072209e-05,
+      "loss": 1.0755,
+      "step": 3389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031893600173357e-05,
+      "loss": 1.0342,
+      "step": 3390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030601191703965e-05,
+      "loss": 1.0597,
+      "step": 3391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029308405374734e-05,
+      "loss": 0.9622,
+      "step": 3392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028015241246485e-05,
+      "loss": 0.9638,
+      "step": 3393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8026721699380067e-05,
+      "loss": 0.9883,
+      "step": 3394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802542777983635e-05,
+      "loss": 1.0372,
+      "step": 3395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024133482676202e-05,
+      "loss": 1.0529,
+      "step": 3396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022838807960537e-05,
+      "loss": 1.0581,
+      "step": 3397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8021543755750264e-05,
+      "loss": 1.104,
+      "step": 3398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020248326106325e-05,
+      "loss": 1.0208,
+      "step": 3399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018952519089668e-05,
+      "loss": 1.0185,
+      "step": 3400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017656334761263e-05,
+      "loss": 1.0737,
+      "step": 3401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016359773182107e-05,
+      "loss": 1.0476,
+      "step": 3402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80150628344132e-05,
+      "loss": 1.0518,
+      "step": 3403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013765518515563e-05,
+      "loss": 1.1404,
+      "step": 3404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012467825550247e-05,
+      "loss": 1.0077,
+      "step": 3405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011169755578305e-05,
+      "loss": 1.1002,
+      "step": 3406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009871308660818e-05,
+      "loss": 1.1503,
+      "step": 3407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008572484858877e-05,
+      "loss": 0.9809,
+      "step": 3408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80072732842336e-05,
+      "loss": 1.0153,
+      "step": 3409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005973706846114e-05,
+      "loss": 1.0052,
+      "step": 3410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004673752757567e-05,
+      "loss": 1.1117,
+      "step": 3411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800337342202913e-05,
+      "loss": 1.0246,
+      "step": 3412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002072714721978e-05,
+      "loss": 1.0624,
+      "step": 3413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800077163089732e-05,
+      "loss": 1.03,
+      "step": 3414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999470170616377e-05,
+      "loss": 1.0623,
+      "step": 3415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998168333940377e-05,
+      "loss": 1.0953,
+      "step": 3416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799686612093058e-05,
+      "loss": 1.02,
+      "step": 3417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995563531648262e-05,
+      "loss": 1.0254,
+      "step": 3418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994260566154706e-05,
+      "loss": 0.9915,
+      "step": 3419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992957224511224e-05,
+      "loss": 1.0075,
+      "step": 3420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799165350677914e-05,
+      "loss": 1.0218,
+      "step": 3421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7990349413019798e-05,
+      "loss": 1.0318,
+      "step": 3422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798904494329456e-05,
+      "loss": 1.0462,
+      "step": 3423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79877400976648e-05,
+      "loss": 1.036,
+      "step": 3424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7986434876191915e-05,
+      "loss": 1.0216,
+      "step": 3425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798512927893733e-05,
+      "loss": 1.0458,
+      "step": 3426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983823305962458e-05,
+      "loss": 1.0202,
+      "step": 3427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982516957328763e-05,
+      "loss": 1.0016,
+      "step": 3428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981210233097708e-05,
+      "loss": 1.0155,
+      "step": 3429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979903133330774e-05,
+      "loss": 1.0035,
+      "step": 3430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797859565808947e-05,
+      "loss": 1.0332,
+      "step": 3431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977287807435305e-05,
+      "loss": 1.0649,
+      "step": 3432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797597958142983e-05,
+      "loss": 1.0448,
+      "step": 3433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797467098013459e-05,
+      "loss": 1.0285,
+      "step": 3434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973362003611165e-05,
+      "loss": 1.0412,
+      "step": 3435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972052651921143e-05,
+      "loss": 1.2041,
+      "step": 3436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970742925126128e-05,
+      "loss": 1.0178,
+      "step": 3437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969432823287753e-05,
+      "loss": 1.0468,
+      "step": 3438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968122346467658e-05,
+      "loss": 1.0128,
+      "step": 3439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79668114947275e-05,
+      "loss": 1.0322,
+      "step": 3440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965500268128967e-05,
+      "loss": 1.01,
+      "step": 3441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796418866673375e-05,
+      "loss": 1.0617,
+      "step": 3442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962876690603565e-05,
+      "loss": 1.0168,
+      "step": 3443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796156433980014e-05,
+      "loss": 1.0576,
+      "step": 3444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796025161438523e-05,
+      "loss": 1.0505,
+      "step": 3445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958938514420597e-05,
+      "loss": 1.0479,
+      "step": 3446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957625039968026e-05,
+      "loss": 1.0957,
+      "step": 3447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956311191089325e-05,
+      "loss": 1.0714,
+      "step": 3448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954996967846306e-05,
+      "loss": 1.07,
+      "step": 3449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953682370300814e-05,
+      "loss": 1.0868,
+      "step": 3450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952367398514698e-05,
+      "loss": 1.0002,
+      "step": 3451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951052052549833e-05,
+      "loss": 1.0789,
+      "step": 3452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794973633246811e-05,
+      "loss": 0.9965,
+      "step": 3453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948420238331436e-05,
+      "loss": 1.0334,
+      "step": 3454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947103770201737e-05,
+      "loss": 1.0464,
+      "step": 3455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945786928140955e-05,
+      "loss": 1.0303,
+      "step": 3456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7944469712211054e-05,
+      "loss": 1.0916,
+      "step": 3457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943152122474008e-05,
+      "loss": 1.0337,
+      "step": 3458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794183415899182e-05,
+      "loss": 0.9939,
+      "step": 3459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940515821826493e-05,
+      "loss": 1.03,
+      "step": 3460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939197111040064e-05,
+      "loss": 1.0229,
+      "step": 3461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937878026694582e-05,
+      "loss": 1.0238,
+      "step": 3462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793655856885211e-05,
+      "loss": 0.9672,
+      "step": 3463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7935238737574736e-05,
+      "loss": 1.0469,
+      "step": 3464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793391853292456e-05,
+      "loss": 0.9898,
+      "step": 3465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932597954963698e-05,
+      "loss": 1.0489,
+      "step": 3466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793127700375429e-05,
+      "loss": 0.9975,
+      "step": 3467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792995567935849e-05,
+      "loss": 0.996,
+      "step": 3468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928633981838468e-05,
+      "loss": 1.1057,
+      "step": 3469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927311911256412e-05,
+      "loss": 1.0933,
+      "step": 3470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925989467674535e-05,
+      "loss": 1.0901,
+      "step": 3471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924666651155052e-05,
+      "loss": 1.0571,
+      "step": 3472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923343461760207e-05,
+      "loss": 1.0572,
+      "step": 3473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7922019899552267e-05,
+      "loss": 1.0124,
+      "step": 3474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920695964593503e-05,
+      "loss": 1.0183,
+      "step": 3475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791937165694621e-05,
+      "loss": 1.0809,
+      "step": 3476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918046976672698e-05,
+      "loss": 1.0959,
+      "step": 3477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7916721923835303e-05,
+      "loss": 0.987,
+      "step": 3478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915396498496366e-05,
+      "loss": 1.0154,
+      "step": 3479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791407070071825e-05,
+      "loss": 1.0422,
+      "step": 3480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912744530563345e-05,
+      "loss": 0.9879,
+      "step": 3481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911417988094042e-05,
+      "loss": 0.9971,
+      "step": 3482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910091073372764e-05,
+      "loss": 1.0288,
+      "step": 3483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908763786461946e-05,
+      "loss": 1.0083,
+      "step": 3484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907436127424038e-05,
+      "loss": 1.0452,
+      "step": 3485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7906108096321503e-05,
+      "loss": 1.0556,
+      "step": 3486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790477969321684e-05,
+      "loss": 0.9292,
+      "step": 3487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903450918172546e-05,
+      "loss": 0.9862,
+      "step": 3488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790212177125115e-05,
+      "loss": 0.9967,
+      "step": 3489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900792252515183e-05,
+      "loss": 0.982,
+      "step": 3490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899462362027207e-05,
+      "loss": 1.0943,
+      "step": 3491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898132099849797e-05,
+      "loss": 1.0996,
+      "step": 3492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896801466045548e-05,
+      "loss": 1.0637,
+      "step": 3493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7895470460677062e-05,
+      "loss": 0.9698,
+      "step": 3494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894139083806973e-05,
+      "loss": 1.0471,
+      "step": 3495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789280733549792e-05,
+      "loss": 1.0522,
+      "step": 3496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891475215812568e-05,
+      "loss": 1.0447,
+      "step": 3497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890142724813596e-05,
+      "loss": 0.9832,
+      "step": 3498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78888098625637e-05,
+      "loss": 1.0462,
+      "step": 3499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887476629125606e-05,
+      "loss": 1.0553,
+      "step": 3500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886143024562023e-05,
+      "loss": 1.1685,
+      "step": 3501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788480904893572e-05,
+      "loss": 1.0178,
+      "step": 3502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883474702309458e-05,
+      "loss": 1.0167,
+      "step": 3503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882139984746016e-05,
+      "loss": 1.0759,
+      "step": 3504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880804896308202e-05,
+      "loss": 1.0243,
+      "step": 3505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879469437058835e-05,
+      "loss": 1.0981,
+      "step": 3506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7878133607060752e-05,
+      "loss": 1.0941,
+      "step": 3507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876797406376802e-05,
+      "loss": 1.0662,
+      "step": 3508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875460835069863e-05,
+      "loss": 1.0524,
+      "step": 3509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874123893202818e-05,
+      "loss": 1.0,
+      "step": 3510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872786580838577e-05,
+      "loss": 0.9869,
+      "step": 3511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871448898040066e-05,
+      "loss": 1.0403,
+      "step": 3512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870110844870222e-05,
+      "loss": 0.9569,
+      "step": 3513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868772421392004e-05,
+      "loss": 1.0308,
+      "step": 3514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786743362766839e-05,
+      "loss": 1.0499,
+      "step": 3515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866094463762372e-05,
+      "loss": 1.1254,
+      "step": 3516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864754929736966e-05,
+      "loss": 1.0382,
+      "step": 3517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786341502565519e-05,
+      "loss": 1.0475,
+      "step": 3518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7862074751580105e-05,
+      "loss": 1.0483,
+      "step": 3519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860734107574763e-05,
+      "loss": 1.0784,
+      "step": 3520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859393093702245e-05,
+      "loss": 1.009,
+      "step": 3521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858051710025653e-05,
+      "loss": 1.065,
+      "step": 3522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78567099566081e-05,
+      "loss": 1.0488,
+      "step": 3523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785536783351272e-05,
+      "loss": 1.0452,
+      "step": 3524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7854025340802663e-05,
+      "loss": 1.0796,
+      "step": 3525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852682478541096e-05,
+      "loss": 0.9511,
+      "step": 3526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851339246791204e-05,
+      "loss": 1.031,
+      "step": 3527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849995645616193e-05,
+      "loss": 1.0576,
+      "step": 3528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848651675079276e-05,
+      "loss": 1.0312,
+      "step": 3529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7847307335243694e-05,
+      "loss": 1.0515,
+      "step": 3530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7845962626172702e-05,
+      "loss": 1.054,
+      "step": 3531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844617547929576e-05,
+      "loss": 1.0161,
+      "step": 3532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843272100577596e-05,
+      "loss": 1.0571,
+      "step": 3533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841926284180076e-05,
+      "loss": 1.0973,
+      "step": 3534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7840580098800337e-05,
+      "loss": 1.0842,
+      "step": 3535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839233544501717e-05,
+      "loss": 0.9996,
+      "step": 3536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837886621347584e-05,
+      "loss": 1.0716,
+      "step": 3537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7836539329401306e-05,
+      "loss": 1.0481,
+      "step": 3538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835191668726278e-05,
+      "loss": 1.0368,
+      "step": 3539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833843639385913e-05,
+      "loss": 1.0939,
+      "step": 3540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783249524144364e-05,
+      "loss": 1.0634,
+      "step": 3541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78311464749629e-05,
+      "loss": 1.009,
+      "step": 3542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782979734000716e-05,
+      "loss": 1.0514,
+      "step": 3543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7828447836639896e-05,
+      "loss": 1.0439,
+      "step": 3544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7827097964924608e-05,
+      "loss": 1.0053,
+      "step": 3545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825747724924813e-05,
+      "loss": 1.1235,
+      "step": 3546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782439711670404e-05,
+      "loss": 1.0573,
+      "step": 3547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782304614032584e-05,
+      "loss": 1.0728,
+      "step": 3548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782169479585378e-05,
+      "loss": 1.0333,
+      "step": 3549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782034308335144e-05,
+      "loss": 0.9941,
+      "step": 3550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818991002882428e-05,
+      "loss": 1.0895,
+      "step": 3551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781763855451036e-05,
+      "loss": 1.027,
+      "step": 3552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781628573829887e-05,
+      "loss": 1.044,
+      "step": 3553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814932554311614e-05,
+      "loss": 1.045,
+      "step": 3554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781357900261226e-05,
+      "loss": 1.1311,
+      "step": 3555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78122250832645e-05,
+      "loss": 1.0007,
+      "step": 3556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781087079633204e-05,
+      "loss": 1.0056,
+      "step": 3557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7809516141878595e-05,
+      "loss": 1.0919,
+      "step": 3558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808161119967912e-05,
+      "loss": 0.9654,
+      "step": 3559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806805730663743e-05,
+      "loss": 1.0596,
+      "step": 3560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805449974029866e-05,
+      "loss": 1.0718,
+      "step": 3561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804093850130074e-05,
+      "loss": 1.0072,
+      "step": 3562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802737359028172e-05,
+      "loss": 1.067,
+      "step": 3563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780138050078799e-05,
+      "loss": 1.0527,
+      "step": 3564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7800023275473365e-05,
+      "loss": 0.9995,
+      "step": 3565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779866568314817e-05,
+      "loss": 1.0451,
+      "step": 3566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779730772387627e-05,
+      "loss": 1.1007,
+      "step": 3567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779594939772157e-05,
+      "loss": 1.1332,
+      "step": 3568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794590704747977e-05,
+      "loss": 1.005,
+      "step": 3569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779323164501942e-05,
+      "loss": 0.9847,
+      "step": 3570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791872218599852e-05,
+      "loss": 1.0487,
+      "step": 3571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790512425553233e-05,
+      "loss": 0.9774,
+      "step": 3572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789152265943545e-05,
+      "loss": 1.0893,
+      "step": 3573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778779173983479e-05,
+      "loss": 1.0182,
+      "step": 3574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786430847290982e-05,
+      "loss": 0.9978,
+      "step": 3575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785069588376155e-05,
+      "loss": 1.069,
+      "step": 3576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7783707963154354e-05,
+      "loss": 1.0369,
+      "step": 3577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782345971689655e-05,
+      "loss": 1.0152,
+      "step": 3578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778098361404614e-05,
+      "loss": 1.0681,
+      "step": 3579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777962089028791e-05,
+      "loss": 1.0755,
+      "step": 3580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777825780047909e-05,
+      "loss": 1.0293,
+      "step": 3581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776894344683804e-05,
+      "loss": 1.0899,
+      "step": 3582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775530522966224e-05,
+      "loss": 1.0222,
+      "step": 3583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777416633539051e-05,
+      "loss": 1.0244,
+      "step": 3584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7772801782020847e-05,
+      "loss": 1.1141,
+      "step": 3585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7771436862921448e-05,
+      "loss": 1.1195,
+      "step": 3586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777007157815654e-05,
+      "loss": 1.0312,
+      "step": 3587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768705927790348e-05,
+      "loss": 1.0334,
+      "step": 3588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776733991188714e-05,
+      "loss": 1.0326,
+      "step": 3589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776597353051119e-05,
+      "loss": 1.0802,
+      "step": 3590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764606783726788e-05,
+      "loss": 1.007,
+      "step": 3591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763239671598244e-05,
+      "loss": 1.013,
+      "step": 3592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761872194189883e-05,
+      "loss": 0.9725,
+      "step": 3593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7760504351566047e-05,
+      "loss": 1.0602,
+      "step": 3594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77591361437911e-05,
+      "loss": 1.1663,
+      "step": 3595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757767570929413e-05,
+      "loss": 1.0658,
+      "step": 3596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756398633045387e-05,
+      "loss": 1.0611,
+      "step": 3597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755029330203436e-05,
+      "loss": 1.0667,
+      "step": 3598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7753659662467982e-05,
+      "loss": 1.0531,
+      "step": 3599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752289629903474e-05,
+      "loss": 1.1067,
+      "step": 3600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750919232574375e-05,
+      "loss": 1.0842,
+      "step": 3601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749548470545167e-05,
+      "loss": 1.0096,
+      "step": 3602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748177343880348e-05,
+      "loss": 1.0102,
+      "step": 3603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746805852644433e-05,
+      "loss": 1.0416,
+      "step": 3604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774543399690195e-05,
+      "loss": 1.0569,
+      "step": 3605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7744061776717454e-05,
+      "loss": 1.077,
+      "step": 3606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742689192155508e-05,
+      "loss": 0.9985,
+      "step": 3607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741316243280696e-05,
+      "loss": 1.0643,
+      "step": 3608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773994293015762e-05,
+      "loss": 1.0453,
+      "step": 3609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77385692528509e-05,
+      "loss": 1.0515,
+      "step": 3610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737195211425163e-05,
+      "loss": 1.0165,
+      "step": 3611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735820805945068e-05,
+      "loss": 1.0502,
+      "step": 3612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734446036475282e-05,
+      "loss": 1.0688,
+      "step": 3613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773307090308049e-05,
+      "loss": 1.1511,
+      "step": 3614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77316954058254e-05,
+      "loss": 1.0158,
+      "step": 3615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773031954477473e-05,
+      "loss": 1.0681,
+      "step": 3616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728943319993215e-05,
+      "loss": 0.9503,
+      "step": 3617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727566731545612e-05,
+      "loss": 1.0338,
+      "step": 3618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7726189779496693e-05,
+      "loss": 1.0586,
+      "step": 3619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724812463911247e-05,
+      "loss": 1.0812,
+      "step": 3620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772343478485408e-05,
+      "loss": 1.0901,
+      "step": 3621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722056742390016e-05,
+      "loss": 0.9683,
+      "step": 3622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7720678336583898e-05,
+      "loss": 1.0171,
+      "step": 3623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7719299567500578e-05,
+      "loss": 1.0217,
+      "step": 3624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771792043520493e-05,
+      "loss": 1.0602,
+      "step": 3625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716540939761847e-05,
+      "loss": 1.1084,
+      "step": 3626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7715161081236246e-05,
+      "loss": 0.9397,
+      "step": 3627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771378085969304e-05,
+      "loss": 1.0539,
+      "step": 3628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771240027519718e-05,
+      "loss": 1.0426,
+      "step": 3629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711019327813625e-05,
+      "loss": 1.0415,
+      "step": 3630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709638017607343e-05,
+      "loss": 0.9851,
+      "step": 3631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708256344643345e-05,
+      "loss": 1.0958,
+      "step": 3632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770687430898663e-05,
+      "loss": 1.1288,
+      "step": 3633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770549191070223e-05,
+      "loss": 0.9822,
+      "step": 3634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704109149855185e-05,
+      "loss": 1.0263,
+      "step": 3635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702726026510566e-05,
+      "loss": 1.0564,
+      "step": 3636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701342540733448e-05,
+      "loss": 1.0036,
+      "step": 3637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699958692588922e-05,
+      "loss": 1.0823,
+      "step": 3638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698574482142113e-05,
+      "loss": 1.0482,
+      "step": 3639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697189909458144e-05,
+      "loss": 1.1129,
+      "step": 3640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695804974602166e-05,
+      "loss": 1.0439,
+      "step": 3641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769441967763934e-05,
+      "loss": 1.0488,
+      "step": 3642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769303401863485e-05,
+      "loss": 1.0763,
+      "step": 3643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691647997653898e-05,
+      "loss": 1.0825,
+      "step": 3644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690261614761692e-05,
+      "loss": 1.0537,
+      "step": 3645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768887487002347e-05,
+      "loss": 1.095,
+      "step": 3646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768748776350448e-05,
+      "loss": 1.069,
+      "step": 3647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768610029526999e-05,
+      "loss": 1.1233,
+      "step": 3648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684712465385285e-05,
+      "loss": 0.9901,
+      "step": 3649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683324273915662e-05,
+      "loss": 1.0336,
+      "step": 3650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768193572092644e-05,
+      "loss": 1.0162,
+      "step": 3651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768054680648296e-05,
+      "loss": 1.0415,
+      "step": 3652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679157530650567e-05,
+      "loss": 1.0638,
+      "step": 3653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677767893494627e-05,
+      "loss": 1.0117,
+      "step": 3654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676377895080535e-05,
+      "loss": 0.9964,
+      "step": 3655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767498753547369e-05,
+      "loss": 1.0814,
+      "step": 3656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767359681473951e-05,
+      "loss": 1.0273,
+      "step": 3657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7672205732943434e-05,
+      "loss": 1.0684,
+      "step": 3658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670814290150918e-05,
+      "loss": 1.0117,
+      "step": 3659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669422486427424e-05,
+      "loss": 1.0119,
+      "step": 3660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766803032183845e-05,
+      "loss": 1.0869,
+      "step": 3661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666637796449494e-05,
+      "loss": 1.0532,
+      "step": 3662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665244910326083e-05,
+      "loss": 1.0121,
+      "step": 3663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7663851663533754e-05,
+      "loss": 1.0496,
+      "step": 3664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766245805613806e-05,
+      "loss": 1.0136,
+      "step": 3665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661064088204577e-05,
+      "loss": 1.0258,
+      "step": 3666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659669759798896e-05,
+      "loss": 1.1121,
+      "step": 3667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658275070986618e-05,
+      "loss": 1.1224,
+      "step": 3668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656880021833367e-05,
+      "loss": 1.0125,
+      "step": 3669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765548461240479e-05,
+      "loss": 1.0536,
+      "step": 3670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654088842766542e-05,
+      "loss": 1.037,
+      "step": 3671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765269271298429e-05,
+      "loss": 1.0269,
+      "step": 3672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651296223123736e-05,
+      "loss": 1.0668,
+      "step": 3673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649899373250587e-05,
+      "loss": 1.0652,
+      "step": 3674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764850216343056e-05,
+      "loss": 1.0451,
+      "step": 3675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76471045937294e-05,
+      "loss": 0.9736,
+      "step": 3676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645706664212875e-05,
+      "loss": 1.0599,
+      "step": 3677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644308374946756e-05,
+      "loss": 1.0832,
+      "step": 3678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764290972599683e-05,
+      "loss": 1.083,
+      "step": 3679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764151071742891e-05,
+      "loss": 1.0663,
+      "step": 3680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640111349308826e-05,
+      "loss": 0.8996,
+      "step": 3681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763871162170242e-05,
+      "loss": 1.0043,
+      "step": 3682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7637311534675556e-05,
+      "loss": 1.0443,
+      "step": 3683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635911088294106e-05,
+      "loss": 1.0127,
+      "step": 3684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763451028262397e-05,
+      "loss": 1.0777,
+      "step": 3685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763310911773105e-05,
+      "loss": 1.1151,
+      "step": 3686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631707593681288e-05,
+      "loss": 1.0235,
+      "step": 3687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630305710540615e-05,
+      "loss": 1.0776,
+      "step": 3688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628903468375e-05,
+      "loss": 1.008,
+      "step": 3689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627500867250428e-05,
+      "loss": 1.102,
+      "step": 3690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626097907232886e-05,
+      "loss": 1.0443,
+      "step": 3691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762469458838839e-05,
+      "loss": 1.0468,
+      "step": 3692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7623290910782968e-05,
+      "loss": 1.0597,
+      "step": 3693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762188687448267e-05,
+      "loss": 1.0226,
+      "step": 3694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620482479553554e-05,
+      "loss": 1.0084,
+      "step": 3695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619077726061707e-05,
+      "loss": 1.0255,
+      "step": 3696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761767261407322e-05,
+      "loss": 1.0782,
+      "step": 3697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761626714365421e-05,
+      "loss": 0.9936,
+      "step": 3698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614861314870806e-05,
+      "loss": 1.1286,
+      "step": 3699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761345512778916e-05,
+      "loss": 1.0,
+      "step": 3700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612048582475435e-05,
+      "loss": 1.0328,
+      "step": 3701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761064167899581e-05,
+      "loss": 1.0024,
+      "step": 3702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760923441741648e-05,
+      "loss": 0.9799,
+      "step": 3703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7607826797803672e-05,
+      "loss": 1.0509,
+      "step": 3704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606418820223606e-05,
+      "loss": 0.9657,
+      "step": 3705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760501048474254e-05,
+      "loss": 1.0137,
+      "step": 3706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603601791426737e-05,
+      "loss": 0.9786,
+      "step": 3707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602192740342473e-05,
+      "loss": 0.9642,
+      "step": 3708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600783331556054e-05,
+      "loss": 1.0513,
+      "step": 3709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7599373565133795e-05,
+      "loss": 1.0345,
+      "step": 3710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759796344114203e-05,
+      "loss": 1.0586,
+      "step": 3711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7596552959647105e-05,
+      "loss": 1.0188,
+      "step": 3712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759514212071539e-05,
+      "loss": 1.0644,
+      "step": 3713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759373092441327e-05,
+      "loss": 1.0245,
+      "step": 3714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592319370807138e-05,
+      "loss": 1.0521,
+      "step": 3715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590907459963417e-05,
+      "loss": 1.076,
+      "step": 3716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7589495191948543e-05,
+      "loss": 0.9808,
+      "step": 3717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588082566828962e-05,
+      "loss": 1.0463,
+      "step": 3718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586669584671147e-05,
+      "loss": 1.0728,
+      "step": 3719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585256245541575e-05,
+      "loss": 1.0023,
+      "step": 3720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583842549506754e-05,
+      "loss": 1.081,
+      "step": 3721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7582428496633194e-05,
+      "loss": 1.0984,
+      "step": 3722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758101408698744e-05,
+      "loss": 1.0472,
+      "step": 3723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7579599320636032e-05,
+      "loss": 1.0682,
+      "step": 3724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7578184197645547e-05,
+      "loss": 1.0659,
+      "step": 3725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576768718082566e-05,
+      "loss": 1.0655,
+      "step": 3726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575352882013693e-05,
+      "loss": 1.0695,
+      "step": 3727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573936689505547e-05,
+      "loss": 0.993,
+      "step": 3728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572520140624762e-05,
+      "loss": 0.968,
+      "step": 3729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571103235437988e-05,
+      "loss": 1.1167,
+      "step": 3730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75696859740119e-05,
+      "loss": 1.0711,
+      "step": 3731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568268356413177e-05,
+      "loss": 1.005,
+      "step": 3732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566850382708526e-05,
+      "loss": 1.0253,
+      "step": 3733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565432052964662e-05,
+      "loss": 1.0856,
+      "step": 3734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564013367248328e-05,
+      "loss": 0.9363,
+      "step": 3735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756259432562627e-05,
+      "loss": 1.038,
+      "step": 3736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561174928165262e-05,
+      "loss": 1.0549,
+      "step": 3737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755975517493208e-05,
+      "loss": 1.0749,
+      "step": 3738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7558335065993543e-05,
+      "loss": 1.0702,
+      "step": 3739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755691460141646e-05,
+      "loss": 1.049,
+      "step": 3740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755549378126767e-05,
+      "loss": 1.0905,
+      "step": 3741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554072605614028e-05,
+      "loss": 1.0499,
+      "step": 3742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7552651074522397e-05,
+      "loss": 1.0459,
+      "step": 3743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551229188059673e-05,
+      "loss": 1.0847,
+      "step": 3744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7549806946292753e-05,
+      "loss": 1.0838,
+      "step": 3745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754838434928856e-05,
+      "loss": 1.0151,
+      "step": 3746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754696139711403e-05,
+      "loss": 0.992,
+      "step": 3747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545538089836115e-05,
+      "loss": 0.9893,
+      "step": 3748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7544114427521783e-05,
+      "loss": 1.0528,
+      "step": 3749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754269041023803e-05,
+      "loss": 1.0844,
+      "step": 3750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754126603805185e-05,
+      "loss": 1.056,
+      "step": 3751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539841311030268e-05,
+      "loss": 1.0441,
+      "step": 3752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753841622924032e-05,
+      "loss": 0.9761,
+      "step": 3753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753699079274906e-05,
+      "loss": 0.9992,
+      "step": 3754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753556500162356e-05,
+      "loss": 1.1259,
+      "step": 3755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75341388559309e-05,
+      "loss": 1.0139,
+      "step": 3756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753271235573819e-05,
+      "loss": 0.9606,
+      "step": 3757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753128550111255e-05,
+      "loss": 1.0321,
+      "step": 3758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529858292121116e-05,
+      "loss": 1.028,
+      "step": 3759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752843072883104e-05,
+      "loss": 1.059,
+      "step": 3760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527002811309498e-05,
+      "loss": 1.0752,
+      "step": 3761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752557453962367e-05,
+      "loss": 1.0642,
+      "step": 3762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524145913840767e-05,
+      "loss": 1.0103,
+      "step": 3763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522716934028003e-05,
+      "loss": 0.9649,
+      "step": 3764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7521287600252618e-05,
+      "loss": 1.0394,
+      "step": 3765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7519857912581866e-05,
+      "loss": 1.0132,
+      "step": 3766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7518427871083013e-05,
+      "loss": 0.9941,
+      "step": 3767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516997475823354e-05,
+      "loss": 1.0954,
+      "step": 3768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515566726870186e-05,
+      "loss": 1.0338,
+      "step": 3769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514135624290832e-05,
+      "loss": 0.9745,
+      "step": 3770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512704168152632e-05,
+      "loss": 1.1069,
+      "step": 3771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7511272358522934e-05,
+      "loss": 1.018,
+      "step": 3772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750984019546911e-05,
+      "loss": 1.0237,
+      "step": 3773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508407679058547e-05,
+      "loss": 1.0412,
+      "step": 3774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750697480935865e-05,
+      "loss": 1.0949,
+      "step": 3775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750554158643684e-05,
+      "loss": 1.0893,
+      "step": 3776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504108010360547e-05,
+      "loss": 1.0616,
+      "step": 3777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502674081197232e-05,
+      "loss": 1.0542,
+      "step": 3778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750123979901436e-05,
+      "loss": 1.0271,
+      "step": 3779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749980516387942e-05,
+      "loss": 1.0207,
+      "step": 3780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7498370175859913e-05,
+      "loss": 1.0135,
+      "step": 3781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7496934835023363e-05,
+      "loss": 1.0452,
+      "step": 3782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495499141437303e-05,
+      "loss": 1.1091,
+      "step": 3783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7494063095169286e-05,
+      "loss": 1.055,
+      "step": 3784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492626696286884e-05,
+      "loss": 1.0215,
+      "step": 3785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749118994485768e-05,
+      "loss": 0.9891,
+      "step": 3786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489752840949277e-05,
+      "loss": 1.0319,
+      "step": 3787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74883153846293e-05,
+      "loss": 1.0903,
+      "step": 3788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7486877575965376e-05,
+      "loss": 1.0661,
+      "step": 3789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485439415025165e-05,
+      "loss": 1.1234,
+      "step": 3790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7484000901876332e-05,
+      "loss": 1.0097,
+      "step": 3791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482562036586562e-05,
+      "loss": 1.0071,
+      "step": 3792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481122819223563e-05,
+      "loss": 1.0608,
+      "step": 3793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479683249855043e-05,
+      "loss": 0.9731,
+      "step": 3794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478243328548748e-05,
+      "loss": 1.0221,
+      "step": 3795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476803055372428e-05,
+      "loss": 1.0811,
+      "step": 3796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475362430393844e-05,
+      "loss": 0.9775,
+      "step": 3797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747392145368079e-05,
+      "loss": 1.0448,
+      "step": 3798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472480125301063e-05,
+      "loss": 0.9996,
+      "step": 3799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7471038445322483e-05,
+      "loss": 1.0491,
+      "step": 3800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469596413812883e-05,
+      "loss": 1.0257,
+      "step": 3801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746815403084011e-05,
+      "loss": 1.0905,
+      "step": 3802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746671129647204e-05,
+      "loss": 1.0816,
+      "step": 3803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465268210776556e-05,
+      "loss": 0.9927,
+      "step": 3804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463824773821555e-05,
+      "loss": 1.0787,
+      "step": 3805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7462380985674957e-05,
+      "loss": 0.9859,
+      "step": 3806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7460936846404695e-05,
+      "loss": 1.0645,
+      "step": 3807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7459492356078715e-05,
+      "loss": 1.0047,
+      "step": 3808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745804751476499e-05,
+      "loss": 0.9927,
+      "step": 3809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456602322531502e-05,
+      "loss": 1.0187,
+      "step": 3810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7455156779446252e-05,
+      "loss": 1.0934,
+      "step": 3811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453710885577252e-05,
+      "loss": 1.1038,
+      "step": 3812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745226464099254e-05,
+      "loss": 1.0632,
+      "step": 3813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745081804576016e-05,
+      "loss": 1.0844,
+      "step": 3814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744937109994818e-05,
+      "loss": 1.0773,
+      "step": 3815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7447923803624686e-05,
+      "loss": 1.076,
+      "step": 3816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446476156857773e-05,
+      "loss": 1.0122,
+      "step": 3817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744502815971556e-05,
+      "loss": 1.0228,
+      "step": 3818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443579812266174e-05,
+      "loss": 1.1131,
+      "step": 3819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7442131114577766e-05,
+      "loss": 1.0076,
+      "step": 3820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440682066718503e-05,
+      "loss": 1.043,
+      "step": 3821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743923266875656e-05,
+      "loss": 1.0015,
+      "step": 3822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743778292076014e-05,
+      "loss": 1.0364,
+      "step": 3823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743633282279746e-05,
+      "loss": 1.0793,
+      "step": 3824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434882374936744e-05,
+      "loss": 0.9772,
+      "step": 3825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743343157724624e-05,
+      "loss": 1.0223,
+      "step": 3826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7431980429794215e-05,
+      "loss": 1.0127,
+      "step": 3827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743052893264895e-05,
+      "loss": 1.0356,
+      "step": 3828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7429077085878735e-05,
+      "loss": 1.0124,
+      "step": 3829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427624889551892e-05,
+      "loss": 1.011,
+      "step": 3830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426172343736744e-05,
+      "loss": 1.0149,
+      "step": 3831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742471944850164e-05,
+      "loss": 0.9892,
+      "step": 3832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423266203914935e-05,
+      "loss": 1.0104,
+      "step": 3833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742181261004502e-05,
+      "loss": 1.0387,
+      "step": 3834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420358666960284e-05,
+      "loss": 1.0245,
+      "step": 3835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741890437472914e-05,
+      "loss": 1.0072,
+      "step": 3836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7417449733420012e-05,
+      "loss": 1.024,
+      "step": 3837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7415994743101348e-05,
+      "loss": 1.1103,
+      "step": 3838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414539403841607e-05,
+      "loss": 0.9948,
+      "step": 3839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413083715709265e-05,
+      "loss": 1.0382,
+      "step": 3840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411627678772825e-05,
+      "loss": 0.9852,
+      "step": 3841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410171293100785e-05,
+      "loss": 0.9701,
+      "step": 3842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408714558761682e-05,
+      "loss": 1.0111,
+      "step": 3843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407257475824052e-05,
+      "loss": 1.0111,
+      "step": 3844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405800044356454e-05,
+      "loss": 0.9632,
+      "step": 3845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740434226442747e-05,
+      "loss": 1.13,
+      "step": 3846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402884136105688e-05,
+      "loss": 1.0667,
+      "step": 3847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740142565945971e-05,
+      "loss": 0.954,
+      "step": 3848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739996683455818e-05,
+      "loss": 1.0285,
+      "step": 3849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7398507661469718e-05,
+      "loss": 1.0323,
+      "step": 3850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7397048140262995e-05,
+      "loss": 1.048,
+      "step": 3851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395588271006684e-05,
+      "loss": 1.0137,
+      "step": 3852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739412805376947e-05,
+      "loss": 1.0302,
+      "step": 3853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739266748862006e-05,
+      "loss": 0.9704,
+      "step": 3854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739120657562718e-05,
+      "loss": 1.0355,
+      "step": 3855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389745314859572e-05,
+      "loss": 1.0453,
+      "step": 3856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738828370638599e-05,
+      "loss": 1.0618,
+      "step": 3857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386821750275203e-05,
+      "loss": 1.0816,
+      "step": 3858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385359446596003e-05,
+      "loss": 1.004,
+      "step": 3859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383896795417196e-05,
+      "loss": 1.0261,
+      "step": 3860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382433796807602e-05,
+      "loss": 1.0218,
+      "step": 3861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380970450836057e-05,
+      "loss": 1.0179,
+      "step": 3862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737950675757142e-05,
+      "loss": 1.0768,
+      "step": 3863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378042717082555e-05,
+      "loss": 1.0364,
+      "step": 3864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376578329438353e-05,
+      "loss": 1.1147,
+      "step": 3865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375113594707715e-05,
+      "loss": 0.9642,
+      "step": 3866
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373648512959562e-05,
+      "loss": 1.0275,
+      "step": 3867
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737218308426283e-05,
+      "loss": 1.0028,
+      "step": 3868
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737071730868647e-05,
+      "loss": 1.0157,
+      "step": 3869
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736925118629945e-05,
+      "loss": 0.9713,
+      "step": 3870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367784717170755e-05,
+      "loss": 0.9922,
+      "step": 3871
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366317901369386e-05,
+      "loss": 1.0308,
+      "step": 3872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364850738964364e-05,
+      "loss": 1.0071,
+      "step": 3873
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363383230024716e-05,
+      "loss": 1.0395,
+      "step": 3874
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361915374619496e-05,
+      "loss": 1.0121,
+      "step": 3875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736044717281777e-05,
+      "loss": 0.983,
+      "step": 3876
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735897862468862e-05,
+      "loss": 1.0523,
+      "step": 3877
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357509730301144e-05,
+      "loss": 1.1345,
+      "step": 3878
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7356040489724462e-05,
+      "loss": 1.0231,
+      "step": 3879
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7354570903027696e-05,
+      "loss": 0.9622,
+      "step": 3880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735310097028e-05,
+      "loss": 1.048,
+      "step": 3881
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351630691550543e-05,
+      "loss": 1.088,
+      "step": 3882
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350160066908497e-05,
+      "loss": 1.0778,
+      "step": 3883
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734868909642306e-05,
+      "loss": 1.0728,
+      "step": 3884
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7347217780163445e-05,
+      "loss": 1.0308,
+      "step": 3885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345746118198885e-05,
+      "loss": 1.0564,
+      "step": 3886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734427411059862e-05,
+      "loss": 1.0112,
+      "step": 3887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342801757431917e-05,
+      "loss": 1.0859,
+      "step": 3888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734132905876805e-05,
+      "loss": 1.0368,
+      "step": 3889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339856014676312e-05,
+      "loss": 1.0732,
+      "step": 3890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7338382625226017e-05,
+      "loss": 1.1154,
+      "step": 3891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733690889048649e-05,
+      "loss": 1.0452,
+      "step": 3892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335434810527077e-05,
+      "loss": 1.0373,
+      "step": 3893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733396038541713e-05,
+      "loss": 1.038,
+      "step": 3894
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733248561522603e-05,
+      "loss": 1.0067,
+      "step": 3895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733101050002317e-05,
+      "loss": 0.9573,
+      "step": 3896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7329535039877955e-05,
+      "loss": 1.0055,
+      "step": 3897
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328059234859806e-05,
+      "loss": 0.9933,
+      "step": 3898
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326583085038165e-05,
+      "loss": 0.9674,
+      "step": 3899
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325106590482494e-05,
+      "loss": 1.0377,
+      "step": 3900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323629751262263e-05,
+      "loss": 0.9603,
+      "step": 3901
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732215256744696e-05,
+      "loss": 0.963,
+      "step": 3902
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320675039106088e-05,
+      "loss": 1.0554,
+      "step": 3903
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7319197166309166e-05,
+      "loss": 1.0501,
+      "step": 3904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731771894912574e-05,
+      "loss": 1.07,
+      "step": 3905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731624038762536e-05,
+      "loss": 1.0509,
+      "step": 3906
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7314761481877598e-05,
+      "loss": 1.033,
+      "step": 3907
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313282231952034e-05,
+      "loss": 1.087,
+      "step": 3908
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311802637918278e-05,
+      "loss": 1.009,
+      "step": 3909
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310322699845942e-05,
+      "loss": 1.0302,
+      "step": 3910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308842417804667e-05,
+      "loss": 1.1309,
+      "step": 3911
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73073617918641e-05,
+      "loss": 1.0075,
+      "step": 3912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305880822093907e-05,
+      "loss": 1.023,
+      "step": 3913
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7304399508563773e-05,
+      "loss": 1.0315,
+      "step": 3914
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302917851343403e-05,
+      "loss": 0.9513,
+      "step": 3915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301435850502502e-05,
+      "loss": 1.0298,
+      "step": 3916
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729995350611081e-05,
+      "loss": 1.0863,
+      "step": 3917
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298470818238067e-05,
+      "loss": 1.061,
+      "step": 3918
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7296987786954047e-05,
+      "loss": 1.0763,
+      "step": 3919
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729550441232853e-05,
+      "loss": 1.0505,
+      "step": 3920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72940206944313e-05,
+      "loss": 1.01,
+      "step": 3921
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292536633332182e-05,
+      "loss": 1.0702,
+      "step": 3922
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7291052229101e-05,
+      "loss": 1.0256,
+      "step": 3923
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72895674818076e-05,
+      "loss": 1.01,
+      "step": 3924
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288082391521842e-05,
+      "loss": 1.1208,
+      "step": 3925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286596958313605e-05,
+      "loss": 0.9901,
+      "step": 3926
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728511118225278e-05,
+      "loss": 1.0837,
+      "step": 3927
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728362506340928e-05,
+      "loss": 1.0008,
+      "step": 3928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7282138601853026e-05,
+      "loss": 1.027,
+      "step": 3929
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7280651797653964e-05,
+      "loss": 1.0786,
+      "step": 3930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727916465088205e-05,
+      "loss": 1.0896,
+      "step": 3931
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277677161607255e-05,
+      "loss": 1.0755,
+      "step": 3932
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276189329899572e-05,
+      "loss": 0.9954,
+      "step": 3933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727470115582901e-05,
+      "loss": 0.961,
+      "step": 3934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273212639465588e-05,
+      "loss": 1.0463,
+      "step": 3935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271723780879344e-05,
+      "loss": 1.079,
+      "step": 3936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727023458014033e-05,
+      "loss": 1.0212,
+      "step": 3937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268745037318628e-05,
+      "loss": 1.06,
+      "step": 3938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7267255152484308e-05,
+      "loss": 0.9916,
+      "step": 3939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265764925707486e-05,
+      "loss": 1.018,
+      "step": 3940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264274357058278e-05,
+      "loss": 1.0287,
+      "step": 3941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262783446606812e-05,
+      "loss": 1.0509,
+      "step": 3942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726129219442325e-05,
+      "loss": 1.031,
+      "step": 3943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725980060057775e-05,
+      "loss": 1.0179,
+      "step": 3944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72583086651405e-05,
+      "loss": 1.0239,
+      "step": 3945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72568163881817e-05,
+      "loss": 1.0328,
+      "step": 3946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7255323769771562e-05,
+      "loss": 1.0667,
+      "step": 3947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253830809980317e-05,
+      "loss": 1.0752,
+      "step": 3948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7252337508878217e-05,
+      "loss": 1.0229,
+      "step": 3949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725084386653552e-05,
+      "loss": 1.0396,
+      "step": 3950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724934988302251e-05,
+      "loss": 1.0238,
+      "step": 3951
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7247855558409484e-05,
+      "loss": 1.1132,
+      "step": 3952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724636089276675e-05,
+      "loss": 1.0437,
+      "step": 3953
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724486588616463e-05,
+      "loss": 0.9454,
+      "step": 3954
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243370538673482e-05,
+      "loss": 0.9539,
+      "step": 3955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241874850363654e-05,
+      "loss": 1.0362,
+      "step": 3956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.724037882130553e-05,
+      "loss": 1.0261,
+      "step": 3957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7238882451569493e-05,
+      "loss": 1.0424,
+      "step": 3958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723738574122596e-05,
+      "loss": 1.0345,
+      "step": 3959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723588869034535e-05,
+      "loss": 1.1682,
+      "step": 3960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72343912989981e-05,
+      "loss": 1.0906,
+      "step": 3961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7232893567254675e-05,
+      "loss": 0.9257,
+      "step": 3962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231395495185543e-05,
+      "loss": 0.9383,
+      "step": 3963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229897082861185e-05,
+      "loss": 0.9929,
+      "step": 3964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7228398330352116e-05,
+      "loss": 1.0177,
+      "step": 3965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722689923772885e-05,
+      "loss": 1.0613,
+      "step": 3966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7225399805061923e-05,
+      "loss": 1.0317,
+      "step": 3967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223900032421886e-05,
+      "loss": 1.0301,
+      "step": 3968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222399919879312e-05,
+      "loss": 1.0315,
+      "step": 3969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220899467504782e-05,
+      "loss": 1.0251,
+      "step": 3970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7219398675368897e-05,
+      "loss": 1.0337,
+      "step": 3971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217897543542274e-05,
+      "loss": 1.1094,
+      "step": 3972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216396072095537e-05,
+      "loss": 1.0324,
+      "step": 3973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214894261099346e-05,
+      "loss": 0.9831,
+      "step": 3974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213392110624355e-05,
+      "loss": 1.0159,
+      "step": 3975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721188962074125e-05,
+      "loss": 1.0672,
+      "step": 3976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210386791520728e-05,
+      "loss": 1.0463,
+      "step": 3977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208883623033496e-05,
+      "loss": 0.9789,
+      "step": 3978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207380115350284e-05,
+      "loss": 1.0036,
+      "step": 3979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7205876268541833e-05,
+      "loss": 1.0608,
+      "step": 3980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7204372082678907e-05,
+      "loss": 1.0546,
+      "step": 3981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202867557832278e-05,
+      "loss": 1.2057,
+      "step": 3982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201362694072746e-05,
+      "loss": 0.9299,
+      "step": 3983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7199857491471105e-05,
+      "loss": 1.0296,
+      "step": 3984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719835195009819e-05,
+      "loss": 1.0291,
+      "step": 3985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7196846070024834e-05,
+      "loss": 1.0533,
+      "step": 3986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195339851321896e-05,
+      "loss": 1.0289,
+      "step": 3987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193833294060246e-05,
+      "loss": 1.0627,
+      "step": 3988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719232639831077e-05,
+      "loss": 0.9754,
+      "step": 3989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190819164144374e-05,
+      "loss": 1.0177,
+      "step": 3990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718931159163198e-05,
+      "loss": 0.9595,
+      "step": 3991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187803680844514e-05,
+      "loss": 0.9984,
+      "step": 3992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718629543185293e-05,
+      "loss": 1.0102,
+      "step": 3993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184786844728195e-05,
+      "loss": 0.9907,
+      "step": 3994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183277919541298e-05,
+      "loss": 1.0234,
+      "step": 3995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181768656363233e-05,
+      "loss": 1.0122,
+      "step": 3996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180259055265014e-05,
+      "loss": 1.0402,
+      "step": 3997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717874911631767e-05,
+      "loss": 1.0418,
+      "step": 3998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717723883959225e-05,
+      "loss": 1.0168,
+      "step": 3999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717572822515982e-05,
+      "loss": 1.041,
+      "step": 4000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7174217273091447e-05,
+      "loss": 0.9927,
+      "step": 4001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717270598345824e-05,
+      "loss": 1.0177,
+      "step": 4002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171194356331297e-05,
+      "loss": 0.9735,
+      "step": 4003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7169682391781747e-05,
+      "loss": 1.0522,
+      "step": 4004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168170089880737e-05,
+      "loss": 0.9935,
+      "step": 4005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166657450699416e-05,
+      "loss": 1.0437,
+      "step": 4006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716514447430896e-05,
+      "loss": 1.0367,
+      "step": 4007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7163631160780566e-05,
+      "loss": 1.0672,
+      "step": 4008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162117510185426e-05,
+      "loss": 1.1443,
+      "step": 4009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7160603522594776e-05,
+      "loss": 1.085,
+      "step": 4010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715908919807984e-05,
+      "loss": 1.0557,
+      "step": 4011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157574536711875e-05,
+      "loss": 0.9589,
+      "step": 4012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156059538562153e-05,
+      "loss": 1.0368,
+      "step": 4013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154544203701954e-05,
+      "loss": 1.1076,
+      "step": 4014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715302853220258e-05,
+      "loss": 0.9811,
+      "step": 4015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715151252413535e-05,
+      "loss": 1.0667,
+      "step": 4016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149996179571594e-05,
+      "loss": 1.0172,
+      "step": 4017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7148479498582657e-05,
+      "loss": 0.9566,
+      "step": 4018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146962481239905e-05,
+      "loss": 1.0403,
+      "step": 4019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145445127614716e-05,
+      "loss": 1.1212,
+      "step": 4020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143927437778487e-05,
+      "loss": 0.9939,
+      "step": 4021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714240941180263e-05,
+      "loss": 1.0526,
+      "step": 4022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714089104975857e-05,
+      "loss": 0.8832,
+      "step": 4023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139372351717747e-05,
+      "loss": 1.0658,
+      "step": 4024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7137853317751626e-05,
+      "loss": 0.9894,
+      "step": 4025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713633394793168e-05,
+      "loss": 1.0524,
+      "step": 4026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134814242329393e-05,
+      "loss": 1.0931,
+      "step": 4027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133294201016274e-05,
+      "loss": 1.0153,
+      "step": 4028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713177382406385e-05,
+      "loss": 1.0367,
+      "step": 4029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713025311154365e-05,
+      "loss": 0.9777,
+      "step": 4030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7128732063527236e-05,
+      "loss": 1.0614,
+      "step": 4031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7127210680086174e-05,
+      "loss": 1.0343,
+      "step": 4032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125688961292042e-05,
+      "loss": 1.0017,
+      "step": 4033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124166907216452e-05,
+      "loss": 1.0537,
+      "step": 4034
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712264451793101e-05,
+      "loss": 1.0734,
+      "step": 4035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7121121793507355e-05,
+      "loss": 1.0869,
+      "step": 4036
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119598734017135e-05,
+      "loss": 1.04,
+      "step": 4037
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118075339532008e-05,
+      "loss": 0.9856,
+      "step": 4038
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711655161012366e-05,
+      "loss": 1.0722,
+      "step": 4039
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711502754586378e-05,
+      "loss": 1.0445,
+      "step": 4040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113503146824088e-05,
+      "loss": 1.0715,
+      "step": 4041
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111978413076303e-05,
+      "loss": 0.9774,
+      "step": 4042
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110453344692173e-05,
+      "loss": 0.9957,
+      "step": 4043
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710892794174345e-05,
+      "loss": 1.0654,
+      "step": 4044
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710740220430191e-05,
+      "loss": 1.0365,
+      "step": 4045
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7105876132439344e-05,
+      "loss": 1.0209,
+      "step": 4046
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710434972622756e-05,
+      "loss": 1.0173,
+      "step": 4047
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102822985738375e-05,
+      "loss": 0.9805,
+      "step": 4048
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101295911043628e-05,
+      "loss": 0.971,
+      "step": 4049
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099768502215176e-05,
+      "loss": 1.0512,
+      "step": 4050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7098240759324875e-05,
+      "loss": 1.0987,
+      "step": 4051
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709671268244462e-05,
+      "loss": 1.1265,
+      "step": 4052
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709518427164631e-05,
+      "loss": 1.0629,
+      "step": 4053
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093655527001852e-05,
+      "loss": 0.9967,
+      "step": 4054
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709212644858319e-05,
+      "loss": 0.9735,
+      "step": 4055
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709059703646226e-05,
+      "loss": 0.9283,
+      "step": 4056
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089067290711033e-05,
+      "loss": 0.9285,
+      "step": 4057
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087537211401483e-05,
+      "loss": 0.9956,
+      "step": 4058
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086006798605602e-05,
+      "loss": 1.0135,
+      "step": 4059
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084476052395405e-05,
+      "loss": 1.0828,
+      "step": 4060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082944972842913e-05,
+      "loss": 1.0502,
+      "step": 4061
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7081413560020168e-05,
+      "loss": 1.0577,
+      "step": 4062
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707988181399923e-05,
+      "loss": 1.0068,
+      "step": 4063
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7078349734852167e-05,
+      "loss": 1.0323,
+      "step": 4064
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707681732265107e-05,
+      "loss": 0.9788,
+      "step": 4065
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707528457746804e-05,
+      "loss": 0.9312,
+      "step": 4066
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70737514993752e-05,
+      "loss": 1.0563,
+      "step": 4067
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7072218088444685e-05,
+      "loss": 1.0579,
+      "step": 4068
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7070684344748646e-05,
+      "loss": 1.0907,
+      "step": 4069
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7069150268359244e-05,
+      "loss": 1.063,
+      "step": 4070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7067615859348667e-05,
+      "loss": 1.0909,
+      "step": 4071
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706608111778911e-05,
+      "loss": 1.0943,
+      "step": 4072
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706454604375279e-05,
+      "loss": 1.056,
+      "step": 4073
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063010637311933e-05,
+      "loss": 1.046,
+      "step": 4074
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061474898538783e-05,
+      "loss": 0.9482,
+      "step": 4075
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70599388275056e-05,
+      "loss": 0.9679,
+      "step": 4076
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7058402424284663e-05,
+      "loss": 1.1224,
+      "step": 4077
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705686568894826e-05,
+      "loss": 1.074,
+      "step": 4078
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70553286215687e-05,
+      "loss": 1.0938,
+      "step": 4079
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705379122221831e-05,
+      "loss": 0.9874,
+      "step": 4080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705225349096942e-05,
+      "loss": 1.048,
+      "step": 4081
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705071542789439e-05,
+      "loss": 1.0167,
+      "step": 4082
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7049177033065594e-05,
+      "loss": 0.9923,
+      "step": 4083
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047638306555406e-05,
+      "loss": 0.9622,
+      "step": 4084
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704609924843623e-05,
+      "loss": 1.0229,
+      "step": 4085
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7044559858780492e-05,
+      "loss": 1.06,
+      "step": 4086
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043020137660615e-05,
+      "loss": 1.05,
+      "step": 4087
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704148008514905e-05,
+      "loss": 1.0336,
+      "step": 4088
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039939701318255e-05,
+      "loss": 1.0723,
+      "step": 4089
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038398986240715e-05,
+      "loss": 1.1271,
+      "step": 4090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036857939988923e-05,
+      "loss": 1.0359,
+      "step": 4091
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703531656263539e-05,
+      "loss": 0.9906,
+      "step": 4092
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703377485425264e-05,
+      "loss": 1.0477,
+      "step": 4093
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032232814913218e-05,
+      "loss": 1.0519,
+      "step": 4094
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030690444689674e-05,
+      "loss": 1.1039,
+      "step": 4095
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7029147743654583e-05,
+      "loss": 1.0213,
+      "step": 4096
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7027604711880537e-05,
+      "loss": 1.093,
+      "step": 4097
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702606134944013e-05,
+      "loss": 1.0434,
+      "step": 4098
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7024517656405993e-05,
+      "loss": 1.0789,
+      "step": 4099
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7022973632850755e-05,
+      "loss": 1.0309,
+      "step": 4100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021429278847063e-05,
+      "loss": 0.9966,
+      "step": 4101
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701988459446759e-05,
+      "loss": 1.0051,
+      "step": 4102
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701833957978501e-05,
+      "loss": 0.9335,
+      "step": 4103
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7016794234872024e-05,
+      "loss": 1.0468,
+      "step": 4104
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701524855980134e-05,
+      "loss": 1.0123,
+      "step": 4105
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7013702554645693e-05,
+      "loss": 1.0535,
+      "step": 4106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012156219477823e-05,
+      "loss": 1.0247,
+      "step": 4107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7010609554370488e-05,
+      "loss": 1.0068,
+      "step": 4108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009062559396465e-05,
+      "loss": 0.9755,
+      "step": 4109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007515234628537e-05,
+      "loss": 1.0267,
+      "step": 4110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700596758013952e-05,
+      "loss": 1.0126,
+      "step": 4111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004419596002222e-05,
+      "loss": 1.0692,
+      "step": 4112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700287128228949e-05,
+      "loss": 1.0371,
+      "step": 4113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001322639074174e-05,
+      "loss": 0.9972,
+      "step": 4114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699977366642914e-05,
+      "loss": 1.0516,
+      "step": 4115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699822436442727e-05,
+      "loss": 0.9915,
+      "step": 4116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6996674733141468e-05,
+      "loss": 1.0203,
+      "step": 4117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699512477264464e-05,
+      "loss": 1.1123,
+      "step": 4118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699357448300972e-05,
+      "loss": 1.0254,
+      "step": 4119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992023864309657e-05,
+      "loss": 0.9906,
+      "step": 4120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6990472916617405e-05,
+      "loss": 1.1097,
+      "step": 4121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698892164000594e-05,
+      "loss": 1.05,
+      "step": 4122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6987370034548255e-05,
+      "loss": 1.0834,
+      "step": 4123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698581810031736e-05,
+      "loss": 1.0,
+      "step": 4124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984265837386272e-05,
+      "loss": 0.9098,
+      "step": 4125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982713245828035e-05,
+      "loss": 1.0003,
+      "step": 4126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981160325715698e-05,
+      "loss": 1.1179,
+      "step": 4127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979607077122328e-05,
+      "loss": 1.073,
+      "step": 4128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978053500121016e-05,
+      "loss": 1.0654,
+      "step": 4129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976499594784855e-05,
+      "loss": 1.0173,
+      "step": 4130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6974945361186963e-05,
+      "loss": 1.0725,
+      "step": 4131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697339079940047e-05,
+      "loss": 1.0056,
+      "step": 4132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697183590949852e-05,
+      "loss": 1.0806,
+      "step": 4133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697028069155428e-05,
+      "loss": 0.9769,
+      "step": 4134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6968725145640926e-05,
+      "loss": 0.9813,
+      "step": 4135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6967169271831644e-05,
+      "loss": 0.9605,
+      "step": 4136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965613070199646e-05,
+      "loss": 0.9963,
+      "step": 4137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696405654081816e-05,
+      "loss": 0.9662,
+      "step": 4138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6962499683760414e-05,
+      "loss": 1.0268,
+      "step": 4139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696094249909967e-05,
+      "loss": 1.0559,
+      "step": 4140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959384986909192e-05,
+      "loss": 1.0293,
+      "step": 4141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957827147262268e-05,
+      "loss": 1.0731,
+      "step": 4142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69562689802322e-05,
+      "loss": 0.9556,
+      "step": 4143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69547104858923e-05,
+      "loss": 1.0137,
+      "step": 4144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953151664315904e-05,
+      "loss": 1.0585,
+      "step": 4145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695159251557635e-05,
+      "loss": 1.0373,
+      "step": 4146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695003303974701e-05,
+      "loss": 1.0575,
+      "step": 4147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948473236901256e-05,
+      "loss": 0.9495,
+      "step": 4148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694691310711248e-05,
+      "loss": 1.0118,
+      "step": 4149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694535265045409e-05,
+      "loss": 1.0613,
+      "step": 4150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6943791866999507e-05,
+      "loss": 1.0244,
+      "step": 4151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942230756822177e-05,
+      "loss": 1.0552,
+      "step": 4152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694066931999555e-05,
+      "loss": 1.0219,
+      "step": 4153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6939107556593096e-05,
+      "loss": 1.019,
+      "step": 4154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937545466688298e-05,
+      "loss": 1.0683,
+      "step": 4155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6935983050354657e-05,
+      "loss": 1.1332,
+      "step": 4156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693442030766569e-05,
+      "loss": 1.0196,
+      "step": 4157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932857238694932e-05,
+      "loss": 1.0929,
+      "step": 4158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693129384351592e-05,
+      "loss": 1.0573,
+      "step": 4159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6929730122202223e-05,
+      "loss": 0.9938,
+      "step": 4160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692816607482741e-05,
+      "loss": 0.9938,
+      "step": 4161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692660170146508e-05,
+      "loss": 1.0852,
+      "step": 4162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692503700218884e-05,
+      "loss": 1.1063,
+      "step": 4163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6923471977072314e-05,
+      "loss": 1.041,
+      "step": 4164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921906626189138e-05,
+      "loss": 1.0761,
+      "step": 4165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920340949612968e-05,
+      "loss": 1.0447,
+      "step": 4166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6918774947417468e-05,
+      "loss": 1.0375,
+      "step": 4167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917208619676327e-05,
+      "loss": 1.0418,
+      "step": 4168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6915641966463244e-05,
+      "loss": 1.0825,
+      "step": 4169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691407498785193e-05,
+      "loss": 1.0893,
+      "step": 4170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6912507683916123e-05,
+      "loss": 1.0478,
+      "step": 4171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691094005472956e-05,
+      "loss": 1.0374,
+      "step": 4172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690937210036601e-05,
+      "loss": 1.0598,
+      "step": 4173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690780382089924e-05,
+      "loss": 1.0044,
+      "step": 4174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6906235216403047e-05,
+      "loss": 1.029,
+      "step": 4175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690466628695124e-05,
+      "loss": 1.056,
+      "step": 4176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690309703261764e-05,
+      "loss": 0.9668,
+      "step": 4177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901527453476078e-05,
+      "loss": 1.0191,
+      "step": 4178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899957549600414e-05,
+      "loss": 0.969,
+      "step": 4179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689838732106451e-05,
+      "loss": 1.02,
+      "step": 4180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896816767942253e-05,
+      "loss": 1.0402,
+      "step": 4181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6895245890307546e-05,
+      "loss": 1.021,
+      "step": 4182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6893674688234293e-05,
+      "loss": 1.0672,
+      "step": 4183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892103161796427e-05,
+      "loss": 1.0369,
+      "step": 4184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890531311067895e-05,
+      "loss": 1.0567,
+      "step": 4185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688895913612265e-05,
+      "loss": 0.9965,
+      "step": 4186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887386637034678e-05,
+      "loss": 1.0175,
+      "step": 4187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688581381387796e-05,
+      "loss": 1.0211,
+      "step": 4188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884240666726497e-05,
+      "loss": 1.0657,
+      "step": 4189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882667195654318e-05,
+      "loss": 1.0254,
+      "step": 4190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881093400735462e-05,
+      "loss": 0.9727,
+      "step": 4191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687951928204397e-05,
+      "loss": 1.0891,
+      "step": 4192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687794483965391e-05,
+      "loss": 1.0529,
+      "step": 4193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876370073639375e-05,
+      "loss": 1.017,
+      "step": 4194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874794984074445e-05,
+      "loss": 1.0285,
+      "step": 4195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873219571033243e-05,
+      "loss": 1.0143,
+      "step": 4196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687164383458989e-05,
+      "loss": 1.0148,
+      "step": 4197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870067774818536e-05,
+      "loss": 1.0778,
+      "step": 4198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868491391793333e-05,
+      "loss": 0.9965,
+      "step": 4199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6866914685588448e-05,
+      "loss": 0.9379,
+      "step": 4200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6865337656278082e-05,
+      "loss": 1.0761,
+      "step": 4201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6863760303936428e-05,
+      "loss": 1.1006,
+      "step": 4202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862182628637713e-05,
+      "loss": 0.9578,
+      "step": 4203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860604630456157e-05,
+      "loss": 1.0409,
+      "step": 4204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6859026309466022e-05,
+      "loss": 1.1164,
+      "step": 4205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6857447665741568e-05,
+      "loss": 1.0106,
+      "step": 4206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855868699357074e-05,
+      "loss": 0.9433,
+      "step": 4207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685428941038683e-05,
+      "loss": 0.9885,
+      "step": 4208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852709798905152e-05,
+      "loss": 1.0859,
+      "step": 4209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685112986498636e-05,
+      "loss": 1.0562,
+      "step": 4210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6849549608704797e-05,
+      "loss": 1.062,
+      "step": 4211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6847969030134817e-05,
+      "loss": 1.0223,
+      "step": 4212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684638812935079e-05,
+      "loss": 1.0063,
+      "step": 4213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68448069064271e-05,
+      "loss": 1.0858,
+      "step": 4214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684322536143815e-05,
+      "loss": 1.01,
+      "step": 4215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841643494458352e-05,
+      "loss": 1.0972,
+      "step": 4216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684006130556214e-05,
+      "loss": 1.0041,
+      "step": 4217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683847879482396e-05,
+      "loss": 1.0916,
+      "step": 4218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836895962318273e-05,
+      "loss": 0.9452,
+      "step": 4219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835312808119553e-05,
+      "loss": 1.0295,
+      "step": 4220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683372933230229e-05,
+      "loss": 1.0696,
+      "step": 4221
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6832145534940996e-05,
+      "loss": 1.0231,
+      "step": 4222
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683056141611019e-05,
+      "loss": 1.056,
+      "step": 4223
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828976975884408e-05,
+      "loss": 1.033,
+      "step": 4224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827392214338203e-05,
+      "loss": 1.0135,
+      "step": 4225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682580713154614e-05,
+      "loss": 1.0528,
+      "step": 4226
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824221727582807e-05,
+      "loss": 0.9629,
+      "step": 4227
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682263600252279e-05,
+      "loss": 0.9577,
+      "step": 4228
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682104995644071e-05,
+      "loss": 0.9981,
+      "step": 4229
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819463589411196e-05,
+      "loss": 1.0351,
+      "step": 4230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6817876901508883e-05,
+      "loss": 1.0381,
+      "step": 4231
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6816289892808434e-05,
+      "loss": 1.0351,
+      "step": 4232
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6814702563384523e-05,
+      "loss": 0.9392,
+      "step": 4233
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813114913311834e-05,
+      "loss": 0.9842,
+      "step": 4234
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681152694266507e-05,
+      "loss": 1.012,
+      "step": 4235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809938651518946e-05,
+      "loss": 1.086,
+      "step": 4236
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808350039948205e-05,
+      "loss": 1.0271,
+      "step": 4237
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806761108027587e-05,
+      "loss": 1.0855,
+      "step": 4238
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805171855831858e-05,
+      "loss": 1.0207,
+      "step": 4239
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803582283435795e-05,
+      "loss": 1.0012,
+      "step": 4240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801992390914193e-05,
+      "loss": 0.9708,
+      "step": 4241
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680040217834186e-05,
+      "loss": 1.0213,
+      "step": 4242
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679881164579362e-05,
+      "loss": 0.9723,
+      "step": 4243
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6797220793344306e-05,
+      "loss": 1.0304,
+      "step": 4244
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679562962106878e-05,
+      "loss": 1.0095,
+      "step": 4245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794038129041905e-05,
+      "loss": 1.0408,
+      "step": 4246
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6792446317338567e-05,
+      "loss": 1.0738,
+      "step": 4247
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790854186033666e-05,
+      "loss": 1.001,
+      "step": 4248
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678926173520211e-05,
+      "loss": 1.0814,
+      "step": 4249
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6787668964918837e-05,
+      "loss": 1.0387,
+      "step": 4250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678607587525878e-05,
+      "loss": 1.067,
+      "step": 4251
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784482466296905e-05,
+      "loss": 1.0671,
+      "step": 4252
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782888738108185e-05,
+      "loss": 1.081,
+      "step": 4253
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6781294690767608e-05,
+      "loss": 1.0338,
+      "step": 4254
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779700324350177e-05,
+      "loss": 1.0044,
+      "step": 4255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6778105638930916e-05,
+      "loss": 1.1194,
+      "step": 4256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776510634584848e-05,
+      "loss": 0.9564,
+      "step": 4257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774915311387037e-05,
+      "loss": 1.0453,
+      "step": 4258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773319669412533e-05,
+      "loss": 1.0224,
+      "step": 4259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677172370873642e-05,
+      "loss": 1.0653,
+      "step": 4260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770127429433796e-05,
+      "loss": 1.1444,
+      "step": 4261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768530831579765e-05,
+      "loss": 0.9259,
+      "step": 4262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676693391524945e-05,
+      "loss": 1.0401,
+      "step": 4263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765336680517998e-05,
+      "loss": 0.9342,
+      "step": 4264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6763739127460556e-05,
+      "loss": 1.0607,
+      "step": 4265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676214125615229e-05,
+      "loss": 0.9863,
+      "step": 4266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760543066668392e-05,
+      "loss": 0.9998,
+      "step": 4267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6758944559084056e-05,
+      "loss": 0.9951,
+      "step": 4268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6757345733474495e-05,
+      "loss": 0.9739,
+      "step": 4269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675574658991494e-05,
+      "loss": 1.0464,
+      "step": 4270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6754147128480632e-05,
+      "loss": 1.028,
+      "step": 4271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752547349246837e-05,
+      "loss": 1.0073,
+      "step": 4272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750947252288822e-05,
+      "loss": 1.0198,
+      "step": 4273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674934683768187e-05,
+      "loss": 0.9664,
+      "step": 4274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67477461055013e-05,
+      "loss": 1.0995,
+      "step": 4275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6746145055822417e-05,
+      "loss": 1.0312,
+      "step": 4276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674454368872056e-05,
+      "loss": 1.1108,
+      "step": 4277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6742942004271076e-05,
+      "loss": 1.0254,
+      "step": 4278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674134000254933e-05,
+      "loss": 1.1018,
+      "step": 4279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739737683630696e-05,
+      "loss": 1.0578,
+      "step": 4280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6738135047590574e-05,
+      "loss": 1.059,
+      "step": 4281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736532094504366e-05,
+      "loss": 0.9806,
+      "step": 4282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67349288244475e-05,
+      "loss": 1.0646,
+      "step": 4283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6733325237495407e-05,
+      "loss": 0.9524,
+      "step": 4284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6731721333723543e-05,
+      "loss": 1.0004,
+      "step": 4285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730117113207377e-05,
+      "loss": 1.0508,
+      "step": 4286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728512576022395e-05,
+      "loss": 0.9972,
+      "step": 4287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726907722244082e-05,
+      "loss": 1.0616,
+      "step": 4288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725302551947966e-05,
+      "loss": 1.0623,
+      "step": 4289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6723697065209566e-05,
+      "loss": 1.0877,
+      "step": 4290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722091262104424e-05,
+      "loss": 1.1191,
+      "step": 4291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67204851427081e-05,
+      "loss": 1.0225,
+      "step": 4292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671887870709616e-05,
+      "loss": 1.0244,
+      "step": 4293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717271955344196e-05,
+      "loss": 1.0865,
+      "step": 4294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671566488752781e-05,
+      "loss": 1.0292,
+      "step": 4295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671405750372262e-05,
+      "loss": 1.0044,
+      "step": 4296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712449804004253e-05,
+      "loss": 1.0023,
+      "step": 4297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710841788448352e-05,
+      "loss": 1.075,
+      "step": 4298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670923345713059e-05,
+      "loss": 1.1315,
+      "step": 4299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670762481012663e-05,
+      "loss": 0.9879,
+      "step": 4300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706015847512175e-05,
+      "loss": 1.0549,
+      "step": 4301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6704406569362918e-05,
+      "loss": 0.982,
+      "step": 4302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670279697575459e-05,
+      "loss": 1.0457,
+      "step": 4303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6701187066762923e-05,
+      "loss": 1.0496,
+      "step": 4304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699576842463665e-05,
+      "loss": 1.0262,
+      "step": 4305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6697966302932584e-05,
+      "loss": 1.019,
+      "step": 4306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669635544824546e-05,
+      "loss": 1.0161,
+      "step": 4307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6694744278478085e-05,
+      "loss": 1.0706,
+      "step": 4308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669313279370627e-05,
+      "loss": 1.0426,
+      "step": 4309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691520994005838e-05,
+      "loss": 1.0191,
+      "step": 4310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668990887945263e-05,
+      "loss": 0.9137,
+      "step": 4311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6688296450122507e-05,
+      "loss": 0.9812,
+      "step": 4312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668668370609132e-05,
+      "loss": 0.9315,
+      "step": 4313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6685070647434968e-05,
+      "loss": 0.976,
+      "step": 4314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683457274229345e-05,
+      "loss": 1.014,
+      "step": 4315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681843586550364e-05,
+      "loss": 1.0316,
+      "step": 4316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6680229584473956e-05,
+      "loss": 1.0129,
+      "step": 4317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678615268076056e-05,
+      "loss": 1.0368,
+      "step": 4318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677000637432625e-05,
+      "loss": 1.074,
+      "step": 4319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667538569261964e-05,
+      "loss": 1.0723,
+      "step": 4320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673770433713084e-05,
+      "loss": 1.0191,
+      "step": 4321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667215486078896e-05,
+      "loss": 0.9897,
+      "step": 4322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670538973923283e-05,
+      "loss": 1.1207,
+      "step": 4323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666892277319209e-05,
+      "loss": 1.0349,
+      "step": 4324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666730625867142e-05,
+      "loss": 1.084,
+      "step": 4325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665689430437338e-05,
+      "loss": 1.0003,
+      "step": 4326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666407228856592e-05,
+      "loss": 1.0995,
+      "step": 4327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662454833133254e-05,
+      "loss": 1.0626,
+      "step": 4328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666083706421545e-05,
+      "loss": 1.0442,
+      "step": 4329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6659218981888626e-05,
+      "loss": 1.0379,
+      "step": 4330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6657600586228912e-05,
+      "loss": 0.9907,
+      "step": 4331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655981877312463e-05,
+      "loss": 1.0057,
+      "step": 4332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6654362855215442e-05,
+      "loss": 0.9844,
+      "step": 4333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665274352001403e-05,
+      "loss": 1.0572,
+      "step": 4334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651123871784416e-05,
+      "loss": 1.0702,
+      "step": 4335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649503910602813e-05,
+      "loss": 0.97,
+      "step": 4336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664788363654544e-05,
+      "loss": 1.0667,
+      "step": 4337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664626304968854e-05,
+      "loss": 1.0585,
+      "step": 4338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664464215010836e-05,
+      "loss": 0.9566,
+      "step": 4339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643020937881172e-05,
+      "loss": 1.056,
+      "step": 4340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641399413083257e-05,
+      "loss": 1.0219,
+      "step": 4341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663977757579091e-05,
+      "loss": 1.0599,
+      "step": 4342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663815542608045e-05,
+      "loss": 0.9626,
+      "step": 4343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663653296402819e-05,
+      "loss": 1.0364,
+      "step": 4344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663491018971048e-05,
+      "loss": 0.9845,
+      "step": 4345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633287103203673e-05,
+      "loss": 1.1027,
+      "step": 4346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631663704584138e-05,
+      "loss": 1.0754,
+      "step": 4347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6630039993928266e-05,
+      "loss": 1.033,
+      "step": 4348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662841597131245e-05,
+      "loss": 0.9529,
+      "step": 4349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6626791636813105e-05,
+      "loss": 1.0714,
+      "step": 4350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625166990506666e-05,
+      "loss": 1.0337,
+      "step": 4351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6623542032469567e-05,
+      "loss": 1.0317,
+      "step": 4352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621916762778273e-05,
+      "loss": 1.0495,
+      "step": 4353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6620291181509257e-05,
+      "loss": 1.0285,
+      "step": 4354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6618665288739005e-05,
+      "loss": 1.0719,
+      "step": 4355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661703908454402e-05,
+      "loss": 1.0356,
+      "step": 4356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615412569000816e-05,
+      "loss": 0.9953,
+      "step": 4357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613785742185934e-05,
+      "loss": 0.9671,
+      "step": 4358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6612158604175904e-05,
+      "loss": 1.0031,
+      "step": 4359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610531155047302e-05,
+      "loss": 0.9781,
+      "step": 4360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6608903394876695e-05,
+      "loss": 1.0603,
+      "step": 4361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607275323740683e-05,
+      "loss": 1.0891,
+      "step": 4362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605646941715857e-05,
+      "loss": 0.9871,
+      "step": 4363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604018248878844e-05,
+      "loss": 1.0814,
+      "step": 4364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660238924530628e-05,
+      "loss": 0.9584,
+      "step": 4365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600759931074807e-05,
+      "loss": 1.0837,
+      "step": 4366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599130306261097e-05,
+      "loss": 1.108,
+      "step": 4367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659750037094182e-05,
+      "loss": 1.0692,
+      "step": 4368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595870125193675e-05,
+      "loss": 0.9931,
+      "step": 4369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659423956909336e-05,
+      "loss": 1.038,
+      "step": 4370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659260870271761e-05,
+      "loss": 1.0102,
+      "step": 4371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659097752614315e-05,
+      "loss": 1.0798,
+      "step": 4372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589346039446737e-05,
+      "loss": 0.9736,
+      "step": 4373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587714242705133e-05,
+      "loss": 1.0829,
+      "step": 4374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586082135995122e-05,
+      "loss": 1.049,
+      "step": 4375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584449719393496e-05,
+      "loss": 0.9519,
+      "step": 4376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658281699297706e-05,
+      "loss": 1.0622,
+      "step": 4377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581183956822647e-05,
+      "loss": 1.1523,
+      "step": 4378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657955061100709e-05,
+      "loss": 1.0421,
+      "step": 4379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657791695560724e-05,
+      "loss": 1.0851,
+      "step": 4380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657628299069997e-05,
+      "loss": 0.9976,
+      "step": 4381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6574648716362157e-05,
+      "loss": 1.0907,
+      "step": 4382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573014132670703e-05,
+      "loss": 1.0088,
+      "step": 4383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571379239702516e-05,
+      "loss": 0.9206,
+      "step": 4384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6569744037534522e-05,
+      "loss": 0.9655,
+      "step": 4385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568108526243664e-05,
+      "loss": 1.041,
+      "step": 4386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566472705906892e-05,
+      "loss": 1.0102,
+      "step": 4387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564836576601178e-05,
+      "loss": 0.984,
+      "step": 4388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6563200138403507e-05,
+      "loss": 0.9547,
+      "step": 4389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6561563391390875e-05,
+      "loss": 1.106,
+      "step": 4390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6559926335640297e-05,
+      "loss": 0.9571,
+      "step": 4391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65582889712288e-05,
+      "loss": 1.0414,
+      "step": 4392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655665129823343e-05,
+      "loss": 1.1144,
+      "step": 4393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6555013316731235e-05,
+      "loss": 1.0425,
+      "step": 4394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655337502679929e-05,
+      "loss": 1.0434,
+      "step": 4395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551736428514688e-05,
+      "loss": 1.0178,
+      "step": 4396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550097521954517e-05,
+      "loss": 0.9773,
+      "step": 4397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548458307195896e-05,
+      "loss": 0.9894,
+      "step": 4398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546818784315956e-05,
+      "loss": 1.019,
+      "step": 4399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545178953391844e-05,
+      "loss": 1.0025,
+      "step": 4400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654353881450071e-05,
+      "loss": 1.0629,
+      "step": 4401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6541898367719736e-05,
+      "loss": 1.0559,
+      "step": 4402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540257613126104e-05,
+      "loss": 1.0905,
+      "step": 4403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.653861655079701e-05,
+      "loss": 1.0659,
+      "step": 4404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653697518080968e-05,
+      "loss": 1.0168,
+      "step": 4405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535333503241337e-05,
+      "loss": 1.0126,
+      "step": 4406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653369151816923e-05,
+      "loss": 1.0348,
+      "step": 4407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653204922567062e-05,
+      "loss": 1.0382,
+      "step": 4408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530406625822775e-05,
+      "loss": 1.1167,
+      "step": 4409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6528763718702992e-05,
+      "loss": 0.9705,
+      "step": 4410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6527120504388565e-05,
+      "loss": 1.0644,
+      "step": 4411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525476982956816e-05,
+      "loss": 1.0794,
+      "step": 4412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523833154485077e-05,
+      "loss": 1.063,
+      "step": 4413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6522189019050695e-05,
+      "loss": 1.0431,
+      "step": 4414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6520544576731028e-05,
+      "loss": 0.926,
+      "step": 4415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651889982760345e-05,
+      "loss": 1.0365,
+      "step": 4416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6517254771745355e-05,
+      "loss": 1.0896,
+      "step": 4417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515609409234142e-05,
+      "loss": 1.0677,
+      "step": 4418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513963740147235e-05,
+      "loss": 0.9459,
+      "step": 4419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512317764562063e-05,
+      "loss": 1.0554,
+      "step": 4420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6510671482556074e-05,
+      "loss": 1.1194,
+      "step": 4421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650902489420673e-05,
+      "loss": 1.0909,
+      "step": 4422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507377999591503e-05,
+      "loss": 1.0622,
+      "step": 4423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505730798787894e-05,
+      "loss": 1.0652,
+      "step": 4424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65040832918734e-05,
+      "loss": 1.037,
+      "step": 4425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502435478925543e-05,
+      "loss": 1.0795,
+      "step": 4426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650078736002185e-05,
+      "loss": 1.0957,
+      "step": 4427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649913893523988e-05,
+      "loss": 1.0458,
+      "step": 4428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649749020465719e-05,
+      "loss": 1.0943,
+      "step": 4429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6495841168351353e-05,
+      "loss": 1.0695,
+      "step": 4430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6494191826399964e-05,
+      "loss": 1.0527,
+      "step": 4431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6492542178880633e-05,
+      "loss": 1.0007,
+      "step": 4432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6490892225870974e-05,
+      "loss": 1.0448,
+      "step": 4433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489241967448625e-05,
+      "loss": 1.0668,
+      "step": 4434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6487591403691227e-05,
+      "loss": 1.0223,
+      "step": 4435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6485940534676456e-05,
+      "loss": 1.1241,
+      "step": 4436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648428936048198e-05,
+      "loss": 1.0086,
+      "step": 4437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6482637881185493e-05,
+      "loss": 0.9567,
+      "step": 4438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6480986096864704e-05,
+      "loss": 1.0897,
+      "step": 4439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479334007597333e-05,
+      "loss": 1.0199,
+      "step": 4440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477681613461106e-05,
+      "loss": 0.973,
+      "step": 4441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476028914533785e-05,
+      "loss": 1.0355,
+      "step": 4442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647437591089313e-05,
+      "loss": 1.0107,
+      "step": 4443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6472722602616914e-05,
+      "loss": 1.0905,
+      "step": 4444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471068989782934e-05,
+      "loss": 1.0207,
+      "step": 4445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469415072468994e-05,
+      "loss": 1.005,
+      "step": 4446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467760850752915e-05,
+      "loss": 1.0265,
+      "step": 4447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6466106324712538e-05,
+      "loss": 0.993,
+      "step": 4448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6464451494425705e-05,
+      "loss": 1.0193,
+      "step": 4449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646279635997028e-05,
+      "loss": 1.1069,
+      "step": 4450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6461140921424144e-05,
+      "loss": 1.0583,
+      "step": 4451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6459485178865192e-05,
+      "loss": 1.0432,
+      "step": 4452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457829132371334e-05,
+      "loss": 1.0194,
+      "step": 4453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645617278202048e-05,
+      "loss": 1.0033,
+      "step": 4454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645451612789057e-05,
+      "loss": 0.9753,
+      "step": 4455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6452859170059554e-05,
+      "loss": 0.9622,
+      "step": 4456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6451201908605397e-05,
+      "loss": 1.0109,
+      "step": 4457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644954434360608e-05,
+      "loss": 1.0606,
+      "step": 4458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644788647513959e-05,
+      "loss": 1.0656,
+      "step": 4459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446228303283937e-05,
+      "loss": 1.049,
+      "step": 4460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644456982811714e-05,
+      "loss": 0.9526,
+      "step": 4461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442911049717237e-05,
+      "loss": 1.0647,
+      "step": 4462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441251968162278e-05,
+      "loss": 1.0847,
+      "step": 4463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439592583530324e-05,
+      "loss": 1.0381,
+      "step": 4464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437932895899455e-05,
+      "loss": 1.0429,
+      "step": 4465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6436272905347765e-05,
+      "loss": 1.0122,
+      "step": 4466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6434612611953355e-05,
+      "loss": 1.02,
+      "step": 4467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643295201579435e-05,
+      "loss": 0.9728,
+      "step": 4468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6431291116948885e-05,
+      "loss": 0.9976,
+      "step": 4469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6429629915495112e-05,
+      "loss": 1.0777,
+      "step": 4470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427968411511193e-05,
+      "loss": 1.0739,
+      "step": 4471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64263066050753e-05,
+      "loss": 0.989,
+      "step": 4472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6424644496265635e-05,
+      "loss": 1.0449,
+      "step": 4473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422982085160394e-05,
+      "loss": 1.0322,
+      "step": 4474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6421319371837808e-05,
+      "loss": 1.0451,
+      "step": 4475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419656356376107e-05,
+      "loss": 1.0009,
+      "step": 4476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417993038853535e-05,
+      "loss": 1.0155,
+      "step": 4477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6416329419348363e-05,
+      "loss": 0.964,
+      "step": 4478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641466549793887e-05,
+      "loss": 1.0247,
+      "step": 4479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641300127470334e-05,
+      "loss": 1.0072,
+      "step": 4480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641133674972008e-05,
+      "loss": 1.0003,
+      "step": 4481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409671923067412e-05,
+      "loss": 1.0298,
+      "step": 4482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408006794823675e-05,
+      "loss": 0.9628,
+      "step": 4483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640634136506721e-05,
+      "loss": 1.078,
+      "step": 4484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640467563387638e-05,
+      "loss": 1.0267,
+      "step": 4485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640300960132957e-05,
+      "loss": 1.0221,
+      "step": 4486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640134326750516e-05,
+      "loss": 1.034,
+      "step": 4487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399676632481564e-05,
+      "loss": 1.0225,
+      "step": 4488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6398009696337193e-05,
+      "loss": 1.0704,
+      "step": 4489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639634245915049e-05,
+      "loss": 0.9836,
+      "step": 4490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63946749209999e-05,
+      "loss": 1.0907,
+      "step": 4491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6393007081963878e-05,
+      "loss": 0.9883,
+      "step": 4492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639133894212091e-05,
+      "loss": 1.0658,
+      "step": 4493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389670501549477e-05,
+      "loss": 1.0393,
+      "step": 4494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388001760328086e-05,
+      "loss": 1.0236,
+      "step": 4495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638633271853526e-05,
+      "loss": 0.9977,
+      "step": 4496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638466337624953e-05,
+      "loss": 0.9842,
+      "step": 4497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382993733549434e-05,
+      "loss": 1.0183,
+      "step": 4498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381323790513544e-05,
+      "loss": 1.0837,
+      "step": 4499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6379653547220435e-05,
+      "loss": 1.0725,
+      "step": 4500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637798300374868e-05,
+      "loss": 1.0074,
+      "step": 4501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6376312160176906e-05,
+      "loss": 1.0081,
+      "step": 4502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374641016583715e-05,
+      "loss": 1.0471,
+      "step": 4503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372969573047742e-05,
+      "loss": 0.9465,
+      "step": 4504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371297829647632e-05,
+      "loss": 1.03,
+      "step": 4505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369625786462046e-05,
+      "loss": 1.0547,
+      "step": 4506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367953443569658e-05,
+      "loss": 1.106,
+      "step": 4507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6366280801049153e-05,
+      "loss": 1.0704,
+      "step": 4508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364607858979237e-05,
+      "loss": 0.9907,
+      "step": 4509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6362934617438628e-05,
+      "loss": 1.0519,
+      "step": 4510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6361261076506044e-05,
+      "loss": 1.0148,
+      "step": 4511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6359587236260243e-05,
+      "loss": 0.9835,
+      "step": 4512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357913096779972e-05,
+      "loss": 1.0186,
+      "step": 4513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635623865814402e-05,
+      "loss": 1.0032,
+      "step": 4514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354563920431154e-05,
+      "loss": 0.9452,
+      "step": 4515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6352888883720188e-05,
+      "loss": 1.0505,
+      "step": 4516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6351213548089934e-05,
+      "loss": 0.9777,
+      "step": 4517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6349537913619215e-05,
+      "loss": 1.041,
+      "step": 4518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634786198038688e-05,
+      "loss": 1.0621,
+      "step": 4519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6346185748471782e-05,
+      "loss": 1.0542,
+      "step": 4520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344509217952792e-05,
+      "loss": 1.0022,
+      "step": 4521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63428323889088e-05,
+      "loss": 1.0216,
+      "step": 4522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63411552614187e-05,
+      "loss": 1.0592,
+      "step": 4523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6339477835561406e-05,
+      "loss": 1.0103,
+      "step": 4524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633780011141585e-05,
+      "loss": 1.0413,
+      "step": 4525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6336122089060965e-05,
+      "loss": 1.1047,
+      "step": 4526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633444376857571e-05,
+      "loss": 1.0579,
+      "step": 4527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633276515003905e-05,
+      "loss": 0.9626,
+      "step": 4528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331086233529977e-05,
+      "loss": 1.0543,
+      "step": 4529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329407019127487e-05,
+      "loss": 1.0526,
+      "step": 4530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632772750691058e-05,
+      "loss": 1.0173,
+      "step": 4531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632604769695829e-05,
+      "loss": 1.0417,
+      "step": 4532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632436758934966e-05,
+      "loss": 0.9958,
+      "step": 4533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322687184163736e-05,
+      "loss": 0.9859,
+      "step": 4534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321006481479587e-05,
+      "loss": 1.0646,
+      "step": 4535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319325481376295e-05,
+      "loss": 1.0711,
+      "step": 4536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6317644183932955e-05,
+      "loss": 1.063,
+      "step": 4537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631596258922868e-05,
+      "loss": 1.025,
+      "step": 4538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6314280697342587e-05,
+      "loss": 1.0208,
+      "step": 4539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312598508353815e-05,
+      "loss": 1.0455,
+      "step": 4540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310916022341514e-05,
+      "loss": 1.0472,
+      "step": 4541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309233239384853e-05,
+      "loss": 1.0878,
+      "step": 4542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307550159563012e-05,
+      "loss": 0.9697,
+      "step": 4543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630586678295518e-05,
+      "loss": 1.0426,
+      "step": 4544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304183109640567e-05,
+      "loss": 1.0404,
+      "step": 4545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630249913969839e-05,
+      "loss": 1.0709,
+      "step": 4546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6300814873207885e-05,
+      "loss": 1.0839,
+      "step": 4547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299130310248307e-05,
+      "loss": 1.0567,
+      "step": 4548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629744545089891e-05,
+      "loss": 1.0182,
+      "step": 4549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629576029523898e-05,
+      "loss": 0.9755,
+      "step": 4550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.62940748433478e-05,
+      "loss": 1.0031,
+      "step": 4551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6292389095304678e-05,
+      "loss": 0.94,
+      "step": 4552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290703051188933e-05,
+      "loss": 1.0016,
+      "step": 4553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.62890167110799e-05,
+      "loss": 0.9784,
+      "step": 4554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628733007505692e-05,
+      "loss": 1.1134,
+      "step": 4555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6285643143199356e-05,
+      "loss": 0.9775,
+      "step": 4556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628395591558658e-05,
+      "loss": 1.0283,
+      "step": 4557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6282268392297988e-05,
+      "loss": 1.0098,
+      "step": 4558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628058057341297e-05,
+      "loss": 0.903,
+      "step": 4559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6278892459010955e-05,
+      "loss": 0.996,
+      "step": 4560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6277204049171362e-05,
+      "loss": 1.1483,
+      "step": 4561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275515343973644e-05,
+      "loss": 1.026,
+      "step": 4562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273826343497252e-05,
+      "loss": 0.9904,
+      "step": 4563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6272137047821658e-05,
+      "loss": 1.0415,
+      "step": 4564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6270447457026355e-05,
+      "loss": 0.9903,
+      "step": 4565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268757571190835e-05,
+      "loss": 1.0316,
+      "step": 4566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267067390394614e-05,
+      "loss": 0.958,
+      "step": 4567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626537691471722e-05,
+      "loss": 1.0551,
+      "step": 4568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263686144238188e-05,
+      "loss": 1.0764,
+      "step": 4569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261995079037082e-05,
+      "loss": 0.9735,
+      "step": 4570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626030371919347e-05,
+      "loss": 1.0467,
+      "step": 4571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258612064786924e-05,
+      "loss": 1.0703,
+      "step": 4572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6256920115897053e-05,
+      "loss": 1.0538,
+      "step": 4573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6255227872603463e-05,
+      "loss": 0.9962,
+      "step": 4574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253535334985774e-05,
+      "loss": 0.949,
+      "step": 4575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251842503123633e-05,
+      "loss": 0.9841,
+      "step": 4576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250149377096686e-05,
+      "loss": 1.0274,
+      "step": 4577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248455956984596e-05,
+      "loss": 1.0147,
+      "step": 4578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6246762242867048e-05,
+      "loss": 1.0373,
+      "step": 4579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6245068234823734e-05,
+      "loss": 1.0194,
+      "step": 4580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624337393293436e-05,
+      "loss": 1.0361,
+      "step": 4581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624167933727865e-05,
+      "loss": 1.0185,
+      "step": 4582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623998444793634e-05,
+      "loss": 0.9577,
+      "step": 4583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623828926498717e-05,
+      "loss": 0.988,
+      "step": 4584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623659378851091e-05,
+      "loss": 1.0408,
+      "step": 4585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6234898018587336e-05,
+      "loss": 0.9862,
+      "step": 4586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233201955296238e-05,
+      "loss": 1.0203,
+      "step": 4587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231505598717418e-05,
+      "loss": 1.0396,
+      "step": 4588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6229808948930696e-05,
+      "loss": 1.0611,
+      "step": 4589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6228112006015903e-05,
+      "loss": 1.0519,
+      "step": 4590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226414770052885e-05,
+      "loss": 0.9962,
+      "step": 4591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62247172411215e-05,
+      "loss": 1.0992,
+      "step": 4592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223019419301616e-05,
+      "loss": 1.0285,
+      "step": 4593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221321304673127e-05,
+      "loss": 1.0409,
+      "step": 4594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219622897315933e-05,
+      "loss": 1.0441,
+      "step": 4595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621792419730995e-05,
+      "loss": 1.0678,
+      "step": 4596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216225204735098e-05,
+      "loss": 0.998,
+      "step": 4597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214525919671325e-05,
+      "loss": 1.0704,
+      "step": 4598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212826342198586e-05,
+      "loss": 1.0723,
+      "step": 4599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211126472396854e-05,
+      "loss": 1.0931,
+      "step": 4600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6209426310346098e-05,
+      "loss": 0.9943,
+      "step": 4601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6207725856126333e-05,
+      "loss": 1.0214,
+      "step": 4602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206025109817556e-05,
+      "loss": 1.0316,
+      "step": 4603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6204324071499802e-05,
+      "loss": 0.9916,
+      "step": 4604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62026227412531e-05,
+      "loss": 0.9966,
+      "step": 4605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6200921119157505e-05,
+      "loss": 1.0648,
+      "step": 4606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6199219205293086e-05,
+      "loss": 0.988,
+      "step": 4607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619751699973992e-05,
+      "loss": 1.0211,
+      "step": 4608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6195814502578095e-05,
+      "loss": 0.9627,
+      "step": 4609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194111713887725e-05,
+      "loss": 0.9536,
+      "step": 4610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619240863374893e-05,
+      "loss": 1.0619,
+      "step": 4611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619070526224184e-05,
+      "loss": 1.0375,
+      "step": 4612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189001599446602e-05,
+      "loss": 1.1348,
+      "step": 4613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187297645443383e-05,
+      "loss": 0.9958,
+      "step": 4614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185593400312356e-05,
+      "loss": 0.9771,
+      "step": 4615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618388886413371e-05,
+      "loss": 0.9498,
+      "step": 4616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182184036987644e-05,
+      "loss": 1.058,
+      "step": 4617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6180478918954378e-05,
+      "loss": 0.9712,
+      "step": 4618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617877351011415e-05,
+      "loss": 1.0101,
+      "step": 4619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177067810547184e-05,
+      "loss": 0.9979,
+      "step": 4620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175361820333755e-05,
+      "loss": 1.0401,
+      "step": 4621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173655539554127e-05,
+      "loss": 0.94,
+      "step": 4622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171948968288588e-05,
+      "loss": 1.0239,
+      "step": 4623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617024210661743e-05,
+      "loss": 1.0152,
+      "step": 4624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616853495462097e-05,
+      "loss": 0.9558,
+      "step": 4625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166827512379537e-05,
+      "loss": 0.9981,
+      "step": 4626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165119779973465e-05,
+      "loss": 1.02,
+      "step": 4627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6163411757483105e-05,
+      "loss": 1.0598,
+      "step": 4628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161703444988833e-05,
+      "loss": 1.0417,
+      "step": 4629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615999484257102e-05,
+      "loss": 1.0579,
+      "step": 4630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6158285950310065e-05,
+      "loss": 1.0038,
+      "step": 4631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156576768286373e-05,
+      "loss": 1.0919,
+      "step": 4632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154867296580367e-05,
+      "loss": 0.9601,
+      "step": 4633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615315753527248e-05,
+      "loss": 1.0857,
+      "step": 4634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6151447484443165e-05,
+      "loss": 1.0604,
+      "step": 4635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149737144172882e-05,
+      "loss": 1.0014,
+      "step": 4636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148026514542103e-05,
+      "loss": 0.9682,
+      "step": 4637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146315595631325e-05,
+      "loss": 0.9905,
+      "step": 4638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614460438752104e-05,
+      "loss": 1.0955,
+      "step": 4639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142892890291776e-05,
+      "loss": 1.041,
+      "step": 4640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6141181104024054e-05,
+      "loss": 0.9902,
+      "step": 4641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6139469028798428e-05,
+      "loss": 0.9466,
+      "step": 4642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6137756664695444e-05,
+      "loss": 0.9846,
+      "step": 4643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136044011795682e-05,
+      "loss": 1.0168,
+      "step": 4644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134331070179723e-05,
+      "loss": 1.0206,
+      "step": 4645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6132617839928166e-05,
+      "loss": 1.0045,
+      "step": 4646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130904321121618e-05,
+      "loss": 1.0139,
+      "step": 4647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612919051384071e-05,
+      "loss": 1.0888,
+      "step": 4648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612747641816608e-05,
+      "loss": 1.0547,
+      "step": 4649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612576203417838e-05,
+      "loss": 0.9759,
+      "step": 4650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612404736195828e-05,
+      "loss": 0.9848,
+      "step": 4651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612233240158645e-05,
+      "loss": 1.0289,
+      "step": 4652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120617153143587e-05,
+      "loss": 0.937,
+      "step": 4653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6118901616710406e-05,
+      "loss": 1.0669,
+      "step": 4654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117185792367617e-05,
+      "loss": 1.0458,
+      "step": 4655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6115469680195955e-05,
+      "loss": 0.9843,
+      "step": 4656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113753280276175e-05,
+      "loss": 1.0055,
+      "step": 4657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112036592689027e-05,
+      "loss": 1.0786,
+      "step": 4658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110319617515292e-05,
+      "loss": 1.0324,
+      "step": 4659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610860235483576e-05,
+      "loss": 0.9648,
+      "step": 4660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610688480473123e-05,
+      "loss": 1.0119,
+      "step": 4661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105166967282518e-05,
+      "loss": 1.0239,
+      "step": 4662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103448842570446e-05,
+      "loss": 1.0139,
+      "step": 4663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101730430675865e-05,
+      "loss": 0.9962,
+      "step": 4664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6100011731679625e-05,
+      "loss": 1.0231,
+      "step": 4665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098292745662598e-05,
+      "loss": 1.0365,
+      "step": 4666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096573472705666e-05,
+      "loss": 0.9964,
+      "step": 4667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6094853912889724e-05,
+      "loss": 1.0982,
+      "step": 4668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609313406629568e-05,
+      "loss": 1.0234,
+      "step": 4669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609141393300446e-05,
+      "loss": 0.9909,
+      "step": 4670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6089693513097e-05,
+      "loss": 0.9633,
+      "step": 4671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6087972806654252e-05,
+      "loss": 1.0479,
+      "step": 4672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086251813757178e-05,
+      "loss": 1.0209,
+      "step": 4673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608453053448675e-05,
+      "loss": 1.0771,
+      "step": 4674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6082808968923966e-05,
+      "loss": 1.0549,
+      "step": 4675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608108711714983e-05,
+      "loss": 1.0577,
+      "step": 4676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607936497924535e-05,
+      "loss": 1.0023,
+      "step": 4677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077642555291566e-05,
+      "loss": 1.0554,
+      "step": 4678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607591984536952e-05,
+      "loss": 1.0496,
+      "step": 4679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607419684956027e-05,
+      "loss": 1.0175,
+      "step": 4680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6072473567944883e-05,
+      "loss": 0.9783,
+      "step": 4681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070750000604448e-05,
+      "loss": 1.0464,
+      "step": 4682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6069026147620064e-05,
+      "loss": 1.0147,
+      "step": 4683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606730200907284e-05,
+      "loss": 1.0165,
+      "step": 4684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6065577585043906e-05,
+      "loss": 1.063,
+      "step": 4685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6063852875614393e-05,
+      "loss": 1.0321,
+      "step": 4686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6062127880865456e-05,
+      "loss": 1.0952,
+      "step": 4687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6060402600878263e-05,
+      "loss": 1.1613,
+      "step": 4688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605867703573399e-05,
+      "loss": 1.0085,
+      "step": 4689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605695118551383e-05,
+      "loss": 1.0176,
+      "step": 4690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055225050298985e-05,
+      "loss": 1.0843,
+      "step": 4691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053498630170675e-05,
+      "loss": 0.9711,
+      "step": 4692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051771925210143e-05,
+      "loss": 1.0304,
+      "step": 4693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050044935498618e-05,
+      "loss": 1.0082,
+      "step": 4694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048317661117374e-05,
+      "loss": 1.0017,
+      "step": 4695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046590102147673e-05,
+      "loss": 1.0002,
+      "step": 4696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044862258670806e-05,
+      "loss": 1.0355,
+      "step": 4697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043134130768072e-05,
+      "loss": 1.1383,
+      "step": 4698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041405718520786e-05,
+      "loss": 1.0435,
+      "step": 4699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603967702201027e-05,
+      "loss": 1.0835,
+      "step": 4700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603794804131786e-05,
+      "loss": 1.053,
+      "step": 4701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6036218776524915e-05,
+      "loss": 1.0099,
+      "step": 4702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60344892277128e-05,
+      "loss": 0.99,
+      "step": 4703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032759394962896e-05,
+      "loss": 0.9696,
+      "step": 4704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6031029278356594e-05,
+      "loss": 1.0632,
+      "step": 4705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60292988779753e-05,
+      "loss": 1.0096,
+      "step": 4706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027568193900432e-05,
+      "loss": 1.0427,
+      "step": 4707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602583722621343e-05,
+      "loss": 1.0225,
+      "step": 4708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024105974995734e-05,
+      "loss": 0.9418,
+      "step": 4709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6022374440328804e-05,
+      "loss": 1.0241,
+      "step": 4710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6020642622294115e-05,
+      "loss": 1.0217,
+      "step": 4711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601891052097315e-05,
+      "loss": 0.9809,
+      "step": 4712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601717813644742e-05,
+      "loss": 0.9643,
+      "step": 4713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015445468798424e-05,
+      "loss": 0.9746,
+      "step": 4714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601371251810769e-05,
+      "loss": 0.9376,
+      "step": 4715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601197928445677e-05,
+      "loss": 1.0407,
+      "step": 4716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60102457679272e-05,
+      "loss": 1.0724,
+      "step": 4717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600851196860056e-05,
+      "loss": 0.9759,
+      "step": 4718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6006777886558423e-05,
+      "loss": 0.9871,
+      "step": 4719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600504352188238e-05,
+      "loss": 1.037,
+      "step": 4720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6003308874654043e-05,
+      "loss": 1.0279,
+      "step": 4721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600157394495503e-05,
+      "loss": 1.0065,
+      "step": 4722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999838732866973e-05,
+      "loss": 1.1402,
+      "step": 4723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998103238471515e-05,
+      "loss": 1.0807,
+      "step": 4724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599636746185032e-05,
+      "loss": 1.0664,
+      "step": 4725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5994631403085055e-05,
+      "loss": 1.0301,
+      "step": 4726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5992895062257413e-05,
+      "loss": 0.9369,
+      "step": 4727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599115843944909e-05,
+      "loss": 1.0654,
+      "step": 4728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989421534741797e-05,
+      "loss": 1.0452,
+      "step": 4729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987684348217263e-05,
+      "loss": 1.0621,
+      "step": 4730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985946879957224e-05,
+      "loss": 1.1398,
+      "step": 4731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984209130043433e-05,
+      "loss": 1.0138,
+      "step": 4732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982471098557656e-05,
+      "loss": 1.1205,
+      "step": 4733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980732785581668e-05,
+      "loss": 1.0094,
+      "step": 4734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978994191197265e-05,
+      "loss": 1.0094,
+      "step": 4735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5977255315486255e-05,
+      "loss": 1.0993,
+      "step": 4736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597551615853045e-05,
+      "loss": 1.0027,
+      "step": 4737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973776720411686e-05,
+      "loss": 0.9775,
+      "step": 4738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972037001211806e-05,
+      "loss": 1.0507,
+      "step": 4739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597029700101267e-05,
+      "loss": 1.015,
+      "step": 4740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5968556719896146e-05,
+      "loss": 1.0822,
+      "step": 4741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5966816157944124e-05,
+      "loss": 1.0661,
+      "step": 4742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5965075315238492e-05,
+      "loss": 1.0818,
+      "step": 4743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5963334191861172e-05,
+      "loss": 0.9647,
+      "step": 4744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596159278789408e-05,
+      "loss": 0.9978,
+      "step": 4745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595985110341916e-05,
+      "loss": 1.0369,
+      "step": 4746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958109138518358e-05,
+      "loss": 1.0877,
+      "step": 4747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595636689327364e-05,
+      "loss": 0.9622,
+      "step": 4748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954624367766985e-05,
+      "loss": 1.1251,
+      "step": 4749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952881562080377e-05,
+      "loss": 1.0048,
+      "step": 4750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5951138476295823e-05,
+      "loss": 1.0317,
+      "step": 4751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594939511049534e-05,
+      "loss": 0.99,
+      "step": 4752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947651464760957e-05,
+      "loss": 1.0698,
+      "step": 4753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945907539174718e-05,
+      "loss": 0.9883,
+      "step": 4754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5944163333818677e-05,
+      "loss": 0.9882,
+      "step": 4755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942418848774905e-05,
+      "loss": 1.1004,
+      "step": 4756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940674084125483e-05,
+      "loss": 1.0459,
+      "step": 4757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5938929039952508e-05,
+      "loss": 1.0194,
+      "step": 4758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593718371633809e-05,
+      "loss": 1.0545,
+      "step": 4759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935438113364342e-05,
+      "loss": 1.0789,
+      "step": 4760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593369223111341e-05,
+      "loss": 1.023,
+      "step": 4761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931946069667438e-05,
+      "loss": 1.0361,
+      "step": 4762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930199629108588e-05,
+      "loss": 1.0287,
+      "step": 4763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928452909519032e-05,
+      "loss": 0.9769,
+      "step": 4764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926705910980958e-05,
+      "loss": 1.0827,
+      "step": 4765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592495863357657e-05,
+      "loss": 0.9866,
+      "step": 4766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923211077388078e-05,
+      "loss": 0.9715,
+      "step": 4767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5921463242497712e-05,
+      "loss": 1.1048,
+      "step": 4768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919715128987706e-05,
+      "loss": 1.0407,
+      "step": 4769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917966736940324e-05,
+      "loss": 1.0527,
+      "step": 4770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591621806643782e-05,
+      "loss": 1.0606,
+      "step": 4771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591446911756248e-05,
+      "loss": 0.9937,
+      "step": 4772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59127198903966e-05,
+      "loss": 0.9422,
+      "step": 4773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5910970385022474e-05,
+      "loss": 1.0394,
+      "step": 4774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590922060152243e-05,
+      "loss": 1.0214,
+      "step": 4775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59074705399788e-05,
+      "loss": 1.0497,
+      "step": 4776
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905720200473922e-05,
+      "loss": 0.9953,
+      "step": 4777
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590396958309016e-05,
+      "loss": 0.9654,
+      "step": 4778
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5902218687909877e-05,
+      "loss": 1.0364,
+      "step": 4779
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590046751501547e-05,
+      "loss": 1.0175,
+      "step": 4780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5898716064489326e-05,
+      "loss": 0.9405,
+      "step": 4781
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5896964336413862e-05,
+      "loss": 1.0627,
+      "step": 4782
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589521233087149e-05,
+      "loss": 1.0102,
+      "step": 4783
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893460047944657e-05,
+      "loss": 1.1178,
+      "step": 4784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5891707487715806e-05,
+      "loss": 1.0204,
+      "step": 4785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889954650267407e-05,
+      "loss": 1.0436,
+      "step": 4786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888201535681922e-05,
+      "loss": 1.0418,
+      "step": 4787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886448144041855e-05,
+      "loss": 1.0235,
+      "step": 4788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884694475429697e-05,
+      "loss": 0.9947,
+      "step": 4789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588294052992797e-05,
+      "loss": 0.9697,
+      "step": 4790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588118630761919e-05,
+      "loss": 0.9355,
+      "step": 4791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587943180858591e-05,
+      "loss": 0.9848,
+      "step": 4792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877677032910675e-05,
+      "loss": 1.1114,
+      "step": 4793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875921980676056e-05,
+      "loss": 0.9479,
+      "step": 4794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5874166651964632e-05,
+      "loss": 0.9577,
+      "step": 4795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587241104685899e-05,
+      "loss": 0.9989,
+      "step": 4796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587065516544175e-05,
+      "loss": 1.1335,
+      "step": 4797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868899007795516e-05,
+      "loss": 1.1399,
+      "step": 4798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5867142574002926e-05,
+      "loss": 1.0251,
+      "step": 4799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5865385864146622e-05,
+      "loss": 0.9994,
+      "step": 4800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863628878309264e-05,
+      "loss": 1.0535,
+      "step": 4801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586187161657352e-05,
+      "loss": 1.0687,
+      "step": 4802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5860114079022074e-05,
+      "loss": 1.0365,
+      "step": 4803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5858356265737627e-05,
+      "loss": 1.016,
+      "step": 4804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856598176802886e-05,
+      "loss": 1.0598,
+      "step": 4805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854839812300567e-05,
+      "loss": 0.9689,
+      "step": 4806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5853081172313417e-05,
+      "loss": 0.998,
+      "step": 4807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5851322256924177e-05,
+      "loss": 1.0925,
+      "step": 4808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584956306621561e-05,
+      "loss": 0.9986,
+      "step": 4809
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584780360027049e-05,
+      "loss": 1.0041,
+      "step": 4810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58460438591716e-05,
+      "loss": 1.0473,
+      "step": 4811
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5844283843001748e-05,
+      "loss": 0.9776,
+      "step": 4812
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842523551843746e-05,
+      "loss": 1.0001,
+      "step": 4813
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584076298578042e-05,
+      "loss": 1.0719,
+      "step": 4814
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58390021448946e-05,
+      "loss": 1.0051,
+      "step": 4815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583724102926915e-05,
+      "loss": 1.0143,
+      "step": 4816
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835479638986924e-05,
+      "loss": 0.9657,
+      "step": 4817
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583371797413081e-05,
+      "loss": 1.0303,
+      "step": 4818
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831956034783696e-05,
+      "loss": 1.0058,
+      "step": 4819
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830193821028477e-05,
+      "loss": 1.0261,
+      "step": 4820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828431332948083e-05,
+      "loss": 1.0439,
+      "step": 4821
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826668570625437e-05,
+      "loss": 1.0141,
+      "step": 4822
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582490553414348e-05,
+      "loss": 1.0647,
+      "step": 4823
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5823142223585167e-05,
+      "loss": 1.0159,
+      "step": 4824
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582137863903347e-05,
+      "loss": 1.0569,
+      "step": 4825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5819614780571366e-05,
+      "loss": 1.0658,
+      "step": 4826
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5817850648281853e-05,
+      "loss": 1.0499,
+      "step": 4827
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816086242247934e-05,
+      "loss": 1.1067,
+      "step": 4828
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814321562552632e-05,
+      "loss": 1.0572,
+      "step": 4829
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5812556609278976e-05,
+      "loss": 0.962,
+      "step": 4830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5810791382510013e-05,
+      "loss": 0.9409,
+      "step": 4831
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809025882328802e-05,
+      "loss": 1.0639,
+      "step": 4832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5807260108818415e-05,
+      "loss": 1.0201,
+      "step": 4833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580549406206193e-05,
+      "loss": 1.0034,
+      "step": 4834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803727742142458e-05,
+      "loss": 1.0195,
+      "step": 4835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5801961149143094e-05,
+      "loss": 0.9623,
+      "step": 4836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580019428314697e-05,
+      "loss": 0.9676,
+      "step": 4837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798427144237214e-05,
+      "loss": 0.9772,
+      "step": 4838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5796659732496978e-05,
+      "loss": 0.9697,
+      "step": 4839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5794892048009423e-05,
+      "loss": 1.012,
+      "step": 4840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5793124090857723e-05,
+      "loss": 1.0069,
+      "step": 4841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5791355861125068e-05,
+      "loss": 1.0616,
+      "step": 4842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789587358894653e-05,
+      "loss": 0.9941,
+      "step": 4843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787818584249694e-05,
+      "loss": 0.9958,
+      "step": 4844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578604953727341e-05,
+      "loss": 1.0435,
+      "step": 4845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5784280218049045e-05,
+      "loss": 0.9603,
+      "step": 4846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5782510626659847e-05,
+      "loss": 0.9672,
+      "step": 4847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780740763189085e-05,
+      "loss": 1.0115,
+      "step": 4848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577897062772003e-05,
+      "loss": 1.011,
+      "step": 4849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5777200220335972e-05,
+      "loss": 1.0035,
+      "step": 4850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775429541120213e-05,
+      "loss": 1.0594,
+      "step": 4851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5773658590156067e-05,
+      "loss": 1.0471,
+      "step": 4852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771887367526866e-05,
+      "loss": 1.0594,
+      "step": 4853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770115873315947e-05,
+      "loss": 0.9779,
+      "step": 4854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5768344107606662e-05,
+      "loss": 1.0284,
+      "step": 4855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5766572070482383e-05,
+      "loss": 1.0752,
+      "step": 4856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5764799762026485e-05,
+      "loss": 0.9582,
+      "step": 4857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763027182322356e-05,
+      "loss": 1.0011,
+      "step": 4858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761254331453406e-05,
+      "loss": 1.0164,
+      "step": 4859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575948120950305e-05,
+      "loss": 0.9939,
+      "step": 4860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757707816554715e-05,
+      "loss": 0.9947,
+      "step": 4861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575593415269185e-05,
+      "loss": 1.079,
+      "step": 4862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57541602179979e-05,
+      "loss": 0.9878,
+      "step": 4863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5752386012556345e-05,
+      "loss": 1.0537,
+      "step": 4864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575061153645066e-05,
+      "loss": 0.9737,
+      "step": 4865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5748836789764337e-05,
+      "loss": 1.0783,
+      "step": 4866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747061772580883e-05,
+      "loss": 0.982,
+      "step": 4867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5745286484983823e-05,
+      "loss": 1.0878,
+      "step": 4868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743510927056676e-05,
+      "loss": 0.9825,
+      "step": 4869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5741735098883004e-05,
+      "loss": 1.0599,
+      "step": 4870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739959000546346e-05,
+      "loss": 1.0184,
+      "step": 4871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5738182632130282e-05,
+      "loss": 0.9913,
+      "step": 4872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5736405993718394e-05,
+      "loss": 1.0071,
+      "step": 4873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5734629085394277e-05,
+      "loss": 0.9759,
+      "step": 4874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573285190724154e-05,
+      "loss": 1.0848,
+      "step": 4875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5731074459343797e-05,
+      "loss": 0.9346,
+      "step": 4876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729296741784686e-05,
+      "loss": 1.0635,
+      "step": 4877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5727518754647854e-05,
+      "loss": 1.067,
+      "step": 4878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725740498016962e-05,
+      "loss": 1.0523,
+      "step": 4879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723961971975674e-05,
+      "loss": 1.0356,
+      "step": 4880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572218317660768e-05,
+      "loss": 0.9431,
+      "step": 4881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720404111996675e-05,
+      "loss": 1.0943,
+      "step": 4882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718624778226372e-05,
+      "loss": 0.9989,
+      "step": 4883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5716845175380484e-05,
+      "loss": 1.0857,
+      "step": 4884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715065303542756e-05,
+      "loss": 1.0509,
+      "step": 4885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5713285162796927e-05,
+      "loss": 1.0007,
+      "step": 4886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711504753226763e-05,
+      "loss": 0.9544,
+      "step": 4887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5709724074916034e-05,
+      "loss": 1.091,
+      "step": 4888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707943127948528e-05,
+      "loss": 0.9574,
+      "step": 4889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570616191240804e-05,
+      "loss": 1.0178,
+      "step": 4890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5704380428378382e-05,
+      "loss": 1.0451,
+      "step": 4891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702598675943373e-05,
+      "loss": 0.9878,
+      "step": 4892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5700816655186857e-05,
+      "loss": 0.944,
+      "step": 4893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5699034366192678e-05,
+      "loss": 1.0442,
+      "step": 4894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5697251809044694e-05,
+      "loss": 1.067,
+      "step": 4895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5695468983826784e-05,
+      "loss": 1.0397,
+      "step": 4896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693685890622836e-05,
+      "loss": 1.0007,
+      "step": 4897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569190252951674e-05,
+      "loss": 0.9586,
+      "step": 4898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569011890059242e-05,
+      "loss": 1.0411,
+      "step": 4899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568833500393379e-05,
+      "loss": 1.0256,
+      "step": 4900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568655083962479e-05,
+      "loss": 0.988,
+      "step": 4901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568476640774937e-05,
+      "loss": 0.9475,
+      "step": 4902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682981708391486e-05,
+      "loss": 0.9761,
+      "step": 4903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5681196741635124e-05,
+      "loss": 1.0685,
+      "step": 4904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679411507564268e-05,
+      "loss": 1.0689,
+      "step": 4905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567762600626291e-05,
+      "loss": 1.068,
+      "step": 4906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5675840237815067e-05,
+      "loss": 1.0392,
+      "step": 4907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674054202304767e-05,
+      "loss": 1.0418,
+      "step": 4908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5672267899816045e-05,
+      "loss": 0.9842,
+      "step": 4909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567048133043295e-05,
+      "loss": 0.984,
+      "step": 4910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5668694494239542e-05,
+      "loss": 1.0686,
+      "step": 4911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56669073913199e-05,
+      "loss": 1.0504,
+      "step": 4912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665120021758116e-05,
+      "loss": 0.9818,
+      "step": 4913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566333238563828e-05,
+      "loss": 0.9893,
+      "step": 4914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5661544483044512e-05,
+      "loss": 1.0428,
+      "step": 4915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659756314060933e-05,
+      "loss": 1.0142,
+      "step": 4916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5657967878771683e-05,
+      "loss": 1.1271,
+      "step": 4917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5656179177260916e-05,
+      "loss": 0.9839,
+      "step": 4918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5654390209612787e-05,
+      "loss": 1.0932,
+      "step": 4919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565260097591148e-05,
+      "loss": 1.0201,
+      "step": 4920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5650811476241174e-05,
+      "loss": 1.0192,
+      "step": 4921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5649021710686074e-05,
+      "loss": 0.9921,
+      "step": 4922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647231679330395e-05,
+      "loss": 1.0719,
+      "step": 4923
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564544138225836e-05,
+      "loss": 0.9571,
+      "step": 4924
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564365081955421e-05,
+      "loss": 0.995,
+      "step": 4925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641859991302185e-05,
+      "loss": 1.0482,
+      "step": 4926
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5640068897586562e-05,
+      "loss": 0.9721,
+      "step": 4927
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638277538491607e-05,
+      "loss": 0.9173,
+      "step": 4928
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5636485914101616e-05,
+      "loss": 1.0351,
+      "step": 4929
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634694024500877e-05,
+      "loss": 0.9347,
+      "step": 4930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632901869773712e-05,
+      "loss": 1.0513,
+      "step": 4931
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5631109450004452e-05,
+      "loss": 1.0983,
+      "step": 4932
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562931676527742e-05,
+      "loss": 0.9518,
+      "step": 4933
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5627523815676975e-05,
+      "loss": 1.0198,
+      "step": 4934
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562573060128748e-05,
+      "loss": 1.0244,
+      "step": 4935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623937122193308e-05,
+      "loss": 1.0755,
+      "step": 4936
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562214337847885e-05,
+      "loss": 0.9825,
+      "step": 4937
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5620349370228497e-05,
+      "loss": 1.0232,
+      "step": 4938
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5618555097526672e-05,
+      "loss": 1.0486,
+      "step": 4939
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5616760560457798e-05,
+      "loss": 0.8722,
+      "step": 4940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561496575910631e-05,
+      "loss": 0.97,
+      "step": 4941
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5613170693556655e-05,
+      "loss": 0.989,
+      "step": 4942
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56113753638933e-05,
+      "loss": 0.9966,
+      "step": 4943
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560957977020072e-05,
+      "loss": 1.0544,
+      "step": 4944
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56077839125634e-05,
+      "loss": 1.0663,
+      "step": 4945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560598779106584e-05,
+      "loss": 1.0048,
+      "step": 4946
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604191405792554e-05,
+      "loss": 1.0,
+      "step": 4947
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560239475682806e-05,
+      "loss": 1.0187,
+      "step": 4948
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600597844256905e-05,
+      "loss": 0.9829,
+      "step": 4949
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5598800668163628e-05,
+      "loss": 0.9464,
+      "step": 4950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5597003228632798e-05,
+      "loss": 1.0625,
+      "step": 4951
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5595205525748986e-05,
+      "loss": 0.8621,
+      "step": 4952
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5593407559596777e-05,
+      "loss": 1.0467,
+      "step": 4953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559160933026077e-05,
+      "loss": 1.0539,
+      "step": 4954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589810837825586e-05,
+      "loss": 1.0636,
+      "step": 4955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588012082375838e-05,
+      "loss": 0.99,
+      "step": 4956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586213063996157e-05,
+      "loss": 1.0571,
+      "step": 4957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584413782771205e-05,
+      "loss": 0.973,
+      "step": 4958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582614238785632e-05,
+      "loss": 1.0671,
+      "step": 4959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580814432124117e-05,
+      "loss": 1.0657,
+      "step": 4960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5579014362871347e-05,
+      "loss": 0.975,
+      "step": 4961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577214031112014e-05,
+      "loss": 1.0321,
+      "step": 4962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575413436930828e-05,
+      "loss": 0.9898,
+      "step": 4963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5573612580412517e-05,
+      "loss": 1.0999,
+      "step": 4964
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5571811461641812e-05,
+      "loss": 0.9982,
+      "step": 4965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557001008070346e-05,
+      "loss": 1.0374,
+      "step": 4966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568208437682222e-05,
+      "loss": 0.9889,
+      "step": 4967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566406532662866e-05,
+      "loss": 0.9386,
+      "step": 4968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5564604365730186e-05,
+      "loss": 1.0796,
+      "step": 4969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556280193696897e-05,
+      "loss": 1.0438,
+      "step": 4970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5560999246464026e-05,
+      "loss": 1.0365,
+      "step": 4971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555919629430018e-05,
+      "loss": 1.0053,
+      "step": 4972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555739308056226e-05,
+      "loss": 0.9559,
+      "step": 4973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555589605335115e-05,
+      "loss": 1.0068,
+      "step": 4974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5553785868703607e-05,
+      "loss": 1.0765,
+      "step": 4975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5551981870752598e-05,
+      "loss": 1.0156,
+      "step": 4976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550177611566976e-05,
+      "loss": 1.035,
+      "step": 4977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5548373091231633e-05,
+      "loss": 1.0093,
+      "step": 4978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554656830983148e-05,
+      "loss": 1.0309,
+      "step": 4979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544763267451432e-05,
+      "loss": 0.9608,
+      "step": 4980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542957964176427e-05,
+      "loss": 0.9876,
+      "step": 4981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5541152400091403e-05,
+      "loss": 0.9951,
+      "step": 4982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539346575281316e-05,
+      "loss": 1.0236,
+      "step": 4983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5537540489831138e-05,
+      "loss": 1.0816,
+      "step": 4984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553573414382585e-05,
+      "loss": 0.9559,
+      "step": 4985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533927537350445e-05,
+      "loss": 1.0727,
+      "step": 4986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532120670489924e-05,
+      "loss": 1.0589,
+      "step": 4987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5530313543329306e-05,
+      "loss": 1.0219,
+      "step": 4988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5528506155953628e-05,
+      "loss": 0.9787,
+      "step": 4989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526698508447925e-05,
+      "loss": 0.9978,
+      "step": 4990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524890600897252e-05,
+      "loss": 1.0524,
+      "step": 4991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5523082433386675e-05,
+      "loss": 1.0237,
+      "step": 4992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5521274006001278e-05,
+      "loss": 1.0042,
+      "step": 4993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551946531882615e-05,
+      "loss": 1.0231,
+      "step": 4994
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551765637194639e-05,
+      "loss": 0.9451,
+      "step": 4995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515847165447114e-05,
+      "loss": 0.9908,
+      "step": 4996
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551403769941345e-05,
+      "loss": 1.041,
+      "step": 4997
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512227973930545e-05,
+      "loss": 0.9834,
+      "step": 4998
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510417989083547e-05,
+      "loss": 1.0449,
+      "step": 4999
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5508607744957615e-05,
+      "loss": 1.0813,
+      "step": 5000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550679724163793e-05,
+      "loss": 1.0574,
+      "step": 5001
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5504986479209683e-05,
+      "loss": 1.0218,
+      "step": 5002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550317545775807e-05,
+      "loss": 0.9665,
+      "step": 5003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501364177368305e-05,
+      "loss": 1.1191,
+      "step": 5004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549955263812562e-05,
+      "loss": 0.9772,
+      "step": 5005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549774084011524e-05,
+      "loss": 1.0263,
+      "step": 5006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5495928783422428e-05,
+      "loss": 1.0306,
+      "step": 5007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5494116468132435e-05,
+      "loss": 1.0492,
+      "step": 5008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549230389433054e-05,
+      "loss": 1.1174,
+      "step": 5009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490491062102032e-05,
+      "loss": 1.0295,
+      "step": 5010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54886779715322e-05,
+      "loss": 0.9914,
+      "step": 5011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486864622706365e-05,
+      "loss": 1.03,
+      "step": 5012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5485051015709843e-05,
+      "loss": 0.9954,
+      "step": 5013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548323715062797e-05,
+      "loss": 1.084,
+      "step": 5014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5481423027546094e-05,
+      "loss": 1.0112,
+      "step": 5015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479608646549575e-05,
+      "loss": 0.8949,
+      "step": 5016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477794007723785e-05,
+      "loss": 1.0117,
+      "step": 5017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54759791111541e-05,
+      "loss": 0.9657,
+      "step": 5018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547416395692592e-05,
+      "loss": 0.9415,
+      "step": 5019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472348545124657e-05,
+      "loss": 0.9897,
+      "step": 5020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547053287583572e-05,
+      "loss": 0.9843,
+      "step": 5021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5468716949144555e-05,
+      "loss": 1.09,
+      "step": 5022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5466900765136594e-05,
+      "loss": 1.0716,
+      "step": 5023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5465084323897296e-05,
+      "loss": 0.9468,
+      "step": 5024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463267625512135e-05,
+      "loss": 1.0219,
+      "step": 5025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5461450670066583e-05,
+      "loss": 1.0632,
+      "step": 5026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545963345764614e-05,
+      "loss": 1.0354,
+      "step": 5027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457815988336302e-05,
+      "loss": 0.9539,
+      "step": 5028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545599826222259e-05,
+      "loss": 1.0121,
+      "step": 5029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545418027939053e-05,
+      "loss": 1.0825,
+      "step": 5030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545236203992567e-05,
+      "loss": 1.0788,
+      "step": 5031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5450543543913555e-05,
+      "loss": 1.0504,
+      "step": 5032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544872479143975e-05,
+      "loss": 1.0734,
+      "step": 5033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544690578258984e-05,
+      "loss": 0.983,
+      "step": 5034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5445086517449402e-05,
+      "loss": 0.977,
+      "step": 5035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5443266996104046e-05,
+      "loss": 1.001,
+      "step": 5036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5441447218639386e-05,
+      "loss": 1.0456,
+      "step": 5037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439627185141037e-05,
+      "loss": 1.0653,
+      "step": 5038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437806895694644e-05,
+      "loss": 1.053,
+      "step": 5039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435986350385856e-05,
+      "loss": 0.9989,
+      "step": 5040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434165549300334e-05,
+      "loss": 0.9815,
+      "step": 5041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432344492523745e-05,
+      "loss": 1.0128,
+      "step": 5042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543052318014178e-05,
+      "loss": 1.0775,
+      "step": 5043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5428701612240137e-05,
+      "loss": 1.0023,
+      "step": 5044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5426879788904524e-05,
+      "loss": 1.0635,
+      "step": 5045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542505771022066e-05,
+      "loss": 1.0361,
+      "step": 5046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5423235376274286e-05,
+      "loss": 0.9758,
+      "step": 5047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5421412787151137e-05,
+      "loss": 0.9452,
+      "step": 5048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5419589942936976e-05,
+      "loss": 1.0284,
+      "step": 5049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541776684371757e-05,
+      "loss": 0.9768,
+      "step": 5050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415943489578708e-05,
+      "loss": 1.0291,
+      "step": 5051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541411988060617e-05,
+      "loss": 0.9886,
+      "step": 5052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412296016885772e-05,
+      "loss": 1.0455,
+      "step": 5053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5410471898503326e-05,
+      "loss": 1.0613,
+      "step": 5054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408647525544667e-05,
+      "loss": 0.9436,
+      "step": 5055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540682289809563e-05,
+      "loss": 1.0618,
+      "step": 5056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540499801624207e-05,
+      "loss": 0.9631,
+      "step": 5057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5403172880069857e-05,
+      "loss": 0.9734,
+      "step": 5058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540134748966486e-05,
+      "loss": 0.9723,
+      "step": 5059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399521845112976e-05,
+      "loss": 1.0661,
+      "step": 5060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5397695946500104e-05,
+      "loss": 0.9436,
+      "step": 5061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539586979391215e-05,
+      "loss": 1.0224,
+      "step": 5062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394043387435048e-05,
+      "loss": 1.0097,
+      "step": 5063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539221672715473e-05,
+      "loss": 1.1056,
+      "step": 5064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390389813157153e-05,
+      "loss": 1.0811,
+      "step": 5065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5388562645528264e-05,
+      "loss": 1.0467,
+      "step": 5066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538673522435405e-05,
+      "loss": 1.0286,
+      "step": 5067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384907549720485e-05,
+      "loss": 1.012,
+      "step": 5068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5383079621713572e-05,
+      "loss": 1.0052,
+      "step": 5069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381251440419315e-05,
+      "loss": 1.0087,
+      "step": 5070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537942300592374e-05,
+      "loss": 1.0068,
+      "step": 5071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537759431831287e-05,
+      "loss": 1.0498,
+      "step": 5072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537576537767276e-05,
+      "loss": 1.054,
+      "step": 5073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373936184089462e-05,
+      "loss": 0.9953,
+      "step": 5074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5372106737649045e-05,
+      "loss": 1.0005,
+      "step": 5075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5370277038437584e-05,
+      "loss": 1.0375,
+      "step": 5076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536844708654118e-05,
+      "loss": 0.9821,
+      "step": 5077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5366616882045925e-05,
+      "loss": 1.1073,
+      "step": 5078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536478642503795e-05,
+      "loss": 1.0552,
+      "step": 5079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536295571560337e-05,
+      "loss": 1.0789,
+      "step": 5080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361124753828325e-05,
+      "loss": 0.9656,
+      "step": 5081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5359293539798974e-05,
+      "loss": 0.9754,
+      "step": 5082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535746207360147e-05,
+      "loss": 1.0633,
+      "step": 5083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5355630355322e-05,
+      "loss": 0.9884,
+      "step": 5084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353798385046744e-05,
+      "loss": 1.0422,
+      "step": 5085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5351966162861903e-05,
+      "loss": 1.0209,
+      "step": 5086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350133688853682e-05,
+      "loss": 1.0567,
+      "step": 5087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5348300963108313e-05,
+      "loss": 1.0088,
+      "step": 5088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534646798571202e-05,
+      "loss": 0.9861,
+      "step": 5089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5344634756751058e-05,
+      "loss": 1.0711,
+      "step": 5090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534280127631168e-05,
+      "loss": 0.9787,
+      "step": 5091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5340967544480155e-05,
+      "loss": 1.0653,
+      "step": 5092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5339133561342768e-05,
+      "loss": 1.0062,
+      "step": 5093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337299326985814e-05,
+      "loss": 1.1124,
+      "step": 5094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5335464841495592e-05,
+      "loss": 0.9912,
+      "step": 5095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5333630104958425e-05,
+      "loss": 0.9896,
+      "step": 5096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5331795117460638e-05,
+      "loss": 1.0525,
+      "step": 5097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532995987908858e-05,
+      "loss": 0.9345,
+      "step": 5098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532812438992859e-05,
+      "loss": 1.0174,
+      "step": 5099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532628865006704e-05,
+      "loss": 1.0404,
+      "step": 5100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5324452659590306e-05,
+      "loss": 0.9915,
+      "step": 5101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5322616418584775e-05,
+      "loss": 1.0279,
+      "step": 5102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532077992713685e-05,
+      "loss": 1.0484,
+      "step": 5103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5318943185332934e-05,
+      "loss": 1.035,
+      "step": 5104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531710619325946e-05,
+      "loss": 0.9688,
+      "step": 5105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5315268951002858e-05,
+      "loss": 1.055,
+      "step": 5106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313431458649577e-05,
+      "loss": 1.0356,
+      "step": 5107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531159371628607e-05,
+      "loss": 1.007,
+      "step": 5108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309755723998817e-05,
+      "loss": 1.0074,
+      "step": 5109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5307917481874292e-05,
+      "loss": 1.0576,
+      "step": 5110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306078989998992e-05,
+      "loss": 0.9899,
+      "step": 5111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304240248459427e-05,
+      "loss": 0.98,
+      "step": 5112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302401257342105e-05,
+      "loss": 1.0343,
+      "step": 5113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530056201673356e-05,
+      "loss": 0.9042,
+      "step": 5114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298722526720333e-05,
+      "loss": 1.0487,
+      "step": 5115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296882787388977e-05,
+      "loss": 1.0475,
+      "step": 5116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529504279882606e-05,
+      "loss": 0.9217,
+      "step": 5117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5293202561118148e-05,
+      "loss": 1.0505,
+      "step": 5118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5291362074351837e-05,
+      "loss": 1.138,
+      "step": 5119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528952133861372e-05,
+      "loss": 1.031,
+      "step": 5120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287680353990415e-05,
+      "loss": 0.9902,
+      "step": 5121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5285839120568545e-05,
+      "loss": 1.0099,
+      "step": 5122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528399763843474e-05,
+      "loss": 1.0081,
+      "step": 5123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282155907675645e-05,
+      "loss": 0.9788,
+      "step": 5124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280313928377925e-05,
+      "loss": 1.0037,
+      "step": 5125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278471700628242e-05,
+      "loss": 0.9758,
+      "step": 5126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5276629224513287e-05,
+      "loss": 1.0192,
+      "step": 5127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5274786500119745e-05,
+      "loss": 1.0161,
+      "step": 5128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5272943527534324e-05,
+      "loss": 1.048,
+      "step": 5129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5271100306843737e-05,
+      "loss": 1.0599,
+      "step": 5130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526925683813472e-05,
+      "loss": 1.0106,
+      "step": 5131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5267413121494003e-05,
+      "loss": 1.0511,
+      "step": 5132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526556915700834e-05,
+      "loss": 0.9261,
+      "step": 5133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52637249447645e-05,
+      "loss": 0.9202,
+      "step": 5134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261880484849254e-05,
+      "loss": 0.9638,
+      "step": 5135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526003577734939e-05,
+      "loss": 1.0198,
+      "step": 5136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5258190822351701e-05,
+      "loss": 0.9364,
+      "step": 5137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256345619943002e-05,
+      "loss": 1.0088,
+      "step": 5138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254500170210111e-05,
+      "loss": 1.0384,
+      "step": 5139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252654473239865e-05,
+      "loss": 0.9726,
+      "step": 5140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250808529119105e-05,
+      "loss": 0.9883,
+      "step": 5141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248962337934692e-05,
+      "loss": 1.0808,
+      "step": 5142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247115899773485e-05,
+      "loss": 0.9784,
+      "step": 5143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524526921472237e-05,
+      "loss": 1.0312,
+      "step": 5144
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524342228286824e-05,
+      "loss": 1.0181,
+      "step": 5145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241575104297996e-05,
+      "loss": 0.9807,
+      "step": 5146
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523972767909855e-05,
+      "loss": 0.98,
+      "step": 5147
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237880007356829e-05,
+      "loss": 1.0578,
+      "step": 5148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523603208915977e-05,
+      "loss": 1.0322,
+      "step": 5149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234183924594325e-05,
+      "loss": 1.0087,
+      "step": 5150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523233551374745e-05,
+      "loss": 1.0415,
+      "step": 5151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230486856706122e-05,
+      "loss": 1.0357,
+      "step": 5152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228637953557322e-05,
+      "loss": 1.0794,
+      "step": 5153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226788804388049e-05,
+      "loss": 0.9698,
+      "step": 5154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224939409285306e-05,
+      "loss": 1.0402,
+      "step": 5155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5223089768336114e-05,
+      "loss": 0.9709,
+      "step": 5156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221239881627502e-05,
+      "loss": 0.9921,
+      "step": 5157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521938974924651e-05,
+      "loss": 0.9836,
+      "step": 5158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217539371280197e-05,
+      "loss": 0.9804,
+      "step": 5159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5215688747815627e-05,
+      "loss": 0.9617,
+      "step": 5160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213837878939868e-05,
+      "loss": 1.0626,
+      "step": 5161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211986764740018e-05,
+      "loss": 1.0664,
+      "step": 5162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210135405303172e-05,
+      "loss": 1.0514,
+      "step": 5163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208283800716439e-05,
+      "loss": 0.9055,
+      "step": 5164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206431951066949e-05,
+      "loss": 1.0404,
+      "step": 5165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204579856441827e-05,
+      "loss": 1.0993,
+      "step": 5166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202727516928229e-05,
+      "loss": 0.9818,
+      "step": 5167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200874932613302e-05,
+      "loss": 1.034,
+      "step": 5168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519902210358422e-05,
+      "loss": 0.9737,
+      "step": 5169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519716902992816e-05,
+      "loss": 1.0188,
+      "step": 5170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195315711732322e-05,
+      "loss": 0.9477,
+      "step": 5171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5193462149083894e-05,
+      "loss": 1.0402,
+      "step": 5172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519160834207011e-05,
+      "loss": 0.9926,
+      "step": 5173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518975429077818e-05,
+      "loss": 1.0654,
+      "step": 5174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187899995295353e-05,
+      "loss": 1.1021,
+      "step": 5175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5186045455708867e-05,
+      "loss": 1.0539,
+      "step": 5176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184190672105994e-05,
+      "loss": 1.0863,
+      "step": 5177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182335644573998e-05,
+      "loss": 0.9854,
+      "step": 5178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518048037320017e-05,
+      "loss": 1.0682,
+      "step": 5179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178624858071797e-05,
+      "loss": 1.0613,
+      "step": 5180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176769099276192e-05,
+      "loss": 1.0073,
+      "step": 5181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174913096900667e-05,
+      "loss": 1.0451,
+      "step": 5182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173056851032559e-05,
+      "loss": 1.0723,
+      "step": 5183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171200361759206e-05,
+      "loss": 0.9972,
+      "step": 5184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5169343629167959e-05,
+      "loss": 0.9469,
+      "step": 5185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5167486653346179e-05,
+      "loss": 1.0083,
+      "step": 5186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516562943438125e-05,
+      "loss": 1.0577,
+      "step": 5187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5163771972360553e-05,
+      "loss": 1.0224,
+      "step": 5188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161914267371487e-05,
+      "loss": 0.9553,
+      "step": 5189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516005631950146e-05,
+      "loss": 0.9451,
+      "step": 5190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158198128837893e-05,
+      "loss": 1.0515,
+      "step": 5191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156339695468225e-05,
+      "loss": 0.991,
+      "step": 5192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154481019479894e-05,
+      "loss": 0.9877,
+      "step": 5193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5152622100960352e-05,
+      "loss": 1.0401,
+      "step": 5194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150762939997075e-05,
+      "loss": 0.9889,
+      "step": 5195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148903536677537e-05,
+      "loss": 0.9872,
+      "step": 5196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5147043891089223e-05,
+      "loss": 1.0718,
+      "step": 5197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5145184003319642e-05,
+      "loss": 1.0547,
+      "step": 5198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143323873456302e-05,
+      "loss": 1.0101,
+      "step": 5199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141463501586726e-05,
+      "loss": 0.9684,
+      "step": 5200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513960288779845e-05,
+      "loss": 0.9891,
+      "step": 5201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5137742032179021e-05,
+      "loss": 0.9991,
+      "step": 5202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135880934815998e-05,
+      "loss": 1.0382,
+      "step": 5203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134019595796949e-05,
+      "loss": 1.0352,
+      "step": 5204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513215801520945e-05,
+      "loss": 0.9944,
+      "step": 5205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130296193141104e-05,
+      "loss": 1.0107,
+      "step": 5206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128434129679506e-05,
+      "loss": 1.0377,
+      "step": 5207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126571824912273e-05,
+      "loss": 0.9867,
+      "step": 5208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5124709278927027e-05,
+      "loss": 1.0019,
+      "step": 5209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122846491811417e-05,
+      "loss": 1.0436,
+      "step": 5210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5120983463653076e-05,
+      "loss": 1.0433,
+      "step": 5211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119120194539679e-05,
+      "loss": 1.0018,
+      "step": 5212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5117256684558885e-05,
+      "loss": 0.972,
+      "step": 5213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115392933798387e-05,
+      "loss": 1.0456,
+      "step": 5214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511352894234587e-05,
+      "loss": 1.009,
+      "step": 5215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111664710289047e-05,
+      "loss": 0.9349,
+      "step": 5216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109800237715633e-05,
+      "loss": 0.9924,
+      "step": 5217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107935524713355e-05,
+      "loss": 1.0035,
+      "step": 5218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5106070571369948e-05,
+      "loss": 0.9975,
+      "step": 5219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104205377773171e-05,
+      "loss": 1.0284,
+      "step": 5220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102339944010782e-05,
+      "loss": 1.0156,
+      "step": 5221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100474270170553e-05,
+      "loss": 1.0428,
+      "step": 5222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098608356340273e-05,
+      "loss": 1.0369,
+      "step": 5223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096742202607732e-05,
+      "loss": 0.874,
+      "step": 5224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509487580906074e-05,
+      "loss": 0.9546,
+      "step": 5225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093009175787118e-05,
+      "loss": 0.9804,
+      "step": 5226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091142302874691e-05,
+      "loss": 0.9828,
+      "step": 5227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5089275190411304e-05,
+      "loss": 1.0282,
+      "step": 5228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5087407838484807e-05,
+      "loss": 0.9983,
+      "step": 5229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085540247183063e-05,
+      "loss": 1.0349,
+      "step": 5230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083672416593949e-05,
+      "loss": 1.0602,
+      "step": 5231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081804346805351e-05,
+      "loss": 0.9759,
+      "step": 5232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5079936037905164e-05,
+      "loss": 1.0874,
+      "step": 5233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078067489981298e-05,
+      "loss": 1.0238,
+      "step": 5234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076198703121676e-05,
+      "loss": 1.0224,
+      "step": 5235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074329677414223e-05,
+      "loss": 1.0734,
+      "step": 5236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072460412946887e-05,
+      "loss": 1.1104,
+      "step": 5237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070590909807616e-05,
+      "loss": 1.1061,
+      "step": 5238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068721168084382e-05,
+      "loss": 1.0376,
+      "step": 5239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066851187865152e-05,
+      "loss": 1.0071,
+      "step": 5240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064980969237924e-05,
+      "loss": 1.0402,
+      "step": 5241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063110512290684e-05,
+      "loss": 1.024,
+      "step": 5242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061239817111453e-05,
+      "loss": 1.012,
+      "step": 5243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059368883788244e-05,
+      "loss": 1.0294,
+      "step": 5244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5057497712409094e-05,
+      "loss": 1.0709,
+      "step": 5245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055626303062042e-05,
+      "loss": 1.0633,
+      "step": 5246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053754655835148e-05,
+      "loss": 0.9771,
+      "step": 5247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505188277081647e-05,
+      "loss": 1.0445,
+      "step": 5248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050010648094096e-05,
+      "loss": 1.0419,
+      "step": 5249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5048138287756104e-05,
+      "loss": 0.9881,
+      "step": 5250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5046265689890596e-05,
+      "loss": 1.1011,
+      "step": 5251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044392854585684e-05,
+      "loss": 1.0553,
+      "step": 5252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042519781929487e-05,
+      "loss": 1.0421,
+      "step": 5253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040646472010144e-05,
+      "loss": 1.0544,
+      "step": 5254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038772924915793e-05,
+      "loss": 1.0349,
+      "step": 5255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036899140734584e-05,
+      "loss": 0.9465,
+      "step": 5256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035025119554697e-05,
+      "loss": 1.0487,
+      "step": 5257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033150861464298e-05,
+      "loss": 1.0193,
+      "step": 5258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.503127636655158e-05,
+      "loss": 0.9843,
+      "step": 5259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5029401634904748e-05,
+      "loss": 1.0002,
+      "step": 5260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027526666612e-05,
+      "loss": 1.0309,
+      "step": 5261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5025651461761568e-05,
+      "loss": 0.9986,
+      "step": 5262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023776020441681e-05,
+      "loss": 0.9892,
+      "step": 5263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5021900342740585e-05,
+      "loss": 1.0001,
+      "step": 5264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5020024428746534e-05,
+      "loss": 0.9889,
+      "step": 5265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5018148278547795e-05,
+      "loss": 0.9568,
+      "step": 5266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5016271892232646e-05,
+      "loss": 0.9519,
+      "step": 5267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5014395269889374e-05,
+      "loss": 1.0429,
+      "step": 5268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5012518411606282e-05,
+      "loss": 1.0509,
+      "step": 5269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501064131747168e-05,
+      "loss": 1.014,
+      "step": 5270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008763987573885e-05,
+      "loss": 0.9951,
+      "step": 5271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5006886422001237e-05,
+      "loss": 1.0543,
+      "step": 5272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005008620842073e-05,
+      "loss": 1.031,
+      "step": 5273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003130584184757e-05,
+      "loss": 0.9234,
+      "step": 5274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001252312117647e-05,
+      "loss": 1.0218,
+      "step": 5275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999373804729127e-05,
+      "loss": 0.9983,
+      "step": 5276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499749506210758e-05,
+      "loss": 0.9491,
+      "step": 5277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499561608434141e-05,
+      "loss": 0.9317,
+      "step": 5278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993736871519025e-05,
+      "loss": 1.0332,
+      "step": 5279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991857423728845e-05,
+      "loss": 0.9438,
+      "step": 5280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989977741059308e-05,
+      "loss": 1.0023,
+      "step": 5281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988097823598852e-05,
+      "loss": 0.9994,
+      "step": 5282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986217671435937e-05,
+      "loss": 1.0099,
+      "step": 5283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984337284659028e-05,
+      "loss": 0.9606,
+      "step": 5284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982456663356594e-05,
+      "loss": 0.9998,
+      "step": 5285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980575807617135e-05,
+      "loss": 0.9261,
+      "step": 5286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978694717529143e-05,
+      "loss": 1.0034,
+      "step": 5287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497681339318113e-05,
+      "loss": 1.0204,
+      "step": 5288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974931834661612e-05,
+      "loss": 1.0542,
+      "step": 5289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973050042059127e-05,
+      "loss": 0.9687,
+      "step": 5290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4971168015462216e-05,
+      "loss": 0.9742,
+      "step": 5291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969285754959436e-05,
+      "loss": 0.9111,
+      "step": 5292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967403260639348e-05,
+      "loss": 1.0987,
+      "step": 5293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965520532590526e-05,
+      "loss": 1.0228,
+      "step": 5294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963637570901567e-05,
+      "loss": 0.9798,
+      "step": 5295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4961754375661056e-05,
+      "loss": 1.0285,
+      "step": 5296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495987094695761e-05,
+      "loss": 0.9679,
+      "step": 5297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4957987284879851e-05,
+      "loss": 0.9988,
+      "step": 5298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4956103389516406e-05,
+      "loss": 1.0213,
+      "step": 5299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954219260955913e-05,
+      "loss": 1.1131,
+      "step": 5300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952334899287034e-05,
+      "loss": 0.9175,
+      "step": 5301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950450304598427e-05,
+      "loss": 0.9598,
+      "step": 5302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494856547697877e-05,
+      "loss": 0.9998,
+      "step": 5303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946680416516743e-05,
+      "loss": 0.9898,
+      "step": 5304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4944795123301052e-05,
+      "loss": 1.0494,
+      "step": 5305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942909597420397e-05,
+      "loss": 1.0073,
+      "step": 5306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49410238389635e-05,
+      "loss": 1.0142,
+      "step": 5307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493913784801909e-05,
+      "loss": 0.988,
+      "step": 5308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937251624675905e-05,
+      "loss": 0.9952,
+      "step": 5309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935365169022703e-05,
+      "loss": 1.0611,
+      "step": 5310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493347848114824e-05,
+      "loss": 1.0854,
+      "step": 5311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931591561141294e-05,
+      "loss": 0.987,
+      "step": 5312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929704409090648e-05,
+      "loss": 1.0257,
+      "step": 5313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927817025085094e-05,
+      "loss": 1.0196,
+      "step": 5314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925929409213443e-05,
+      "loss": 1.0234,
+      "step": 5315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924041561564508e-05,
+      "loss": 0.9534,
+      "step": 5316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492215348222712e-05,
+      "loss": 0.9813,
+      "step": 5317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920265171290116e-05,
+      "loss": 0.9958,
+      "step": 5318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918376628842346e-05,
+      "loss": 1.086,
+      "step": 5319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491648785497267e-05,
+      "loss": 1.0245,
+      "step": 5320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491459884976996e-05,
+      "loss": 1.0354,
+      "step": 5321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4912709613323105e-05,
+      "loss": 0.9369,
+      "step": 5322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4910820145720983e-05,
+      "loss": 1.0844,
+      "step": 5323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908930447052511e-05,
+      "loss": 1.0553,
+      "step": 5324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49070405174066e-05,
+      "loss": 0.9667,
+      "step": 5325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490515035687218e-05,
+      "loss": 1.0561,
+      "step": 5326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4903259965538182e-05,
+      "loss": 1.0372,
+      "step": 5327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4901369343493554e-05,
+      "loss": 0.9832,
+      "step": 5328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4899478490827255e-05,
+      "loss": 1.0646,
+      "step": 5329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897587407628257e-05,
+      "loss": 1.0128,
+      "step": 5330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489569609398554e-05,
+      "loss": 0.9798,
+      "step": 5331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4893804549988093e-05,
+      "loss": 1.0356,
+      "step": 5332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891912775724918e-05,
+      "loss": 0.9754,
+      "step": 5333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489002077128503e-05,
+      "loss": 0.9917,
+      "step": 5334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888128536757447e-05,
+      "loss": 1.063,
+      "step": 5335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4886236072231212e-05,
+      "loss": 1.036,
+      "step": 5336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884343377795363e-05,
+      "loss": 1.0092,
+      "step": 5337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882450453538957e-05,
+      "loss": 0.9633,
+      "step": 5338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4880557299551064e-05,
+      "loss": 1.0363,
+      "step": 5339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878663915920762e-05,
+      "loss": 1.0241,
+      "step": 5340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4876770302737132e-05,
+      "loss": 1.0619,
+      "step": 5341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874876460089281e-05,
+      "loss": 1.0092,
+      "step": 5342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872982388066317e-05,
+      "loss": 1.0733,
+      "step": 5343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487108808675736e-05,
+      "loss": 0.9932,
+      "step": 5344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486919355625154e-05,
+      "loss": 0.9715,
+      "step": 5345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867298796638004e-05,
+      "loss": 1.0231,
+      "step": 5346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48654038080059e-05,
+      "loss": 1.0657,
+      "step": 5347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4863508590444395e-05,
+      "loss": 0.9919,
+      "step": 5348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4861613144042666e-05,
+      "loss": 1.0263,
+      "step": 5349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859717468889893e-05,
+      "loss": 1.0191,
+      "step": 5350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857821565075277e-05,
+      "loss": 1.0372,
+      "step": 5351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855925432688019e-05,
+      "loss": 0.9494,
+      "step": 5352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4854029071817343e-05,
+      "loss": 1.0447,
+      "step": 5353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852132482552477e-05,
+      "loss": 0.9681,
+      "step": 5354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4850235664982658e-05,
+      "loss": 1.0029,
+      "step": 5355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848338619197131e-05,
+      "loss": 0.9775,
+      "step": 5356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4846441345285167e-05,
+      "loss": 1.0515,
+      "step": 5357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844543843336032e-05,
+      "loss": 1.0552,
+      "step": 5358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484264611343901e-05,
+      "loss": 0.9682,
+      "step": 5359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4840748155683392e-05,
+      "loss": 0.9875,
+      "step": 5360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483884997015848e-05,
+      "loss": 1.0688,
+      "step": 5361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836951556953592e-05,
+      "loss": 1.0361,
+      "step": 5362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835052916158052e-05,
+      "loss": 1.019,
+      "step": 5363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833154047861198e-05,
+      "loss": 1.0512,
+      "step": 5364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4831254952152373e-05,
+      "loss": 1.082,
+      "step": 5365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829355629120934e-05,
+      "loss": 1.0089,
+      "step": 5366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827456078856251e-05,
+      "loss": 0.9336,
+      "step": 5367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825556301447702e-05,
+      "loss": 1.0065,
+      "step": 5368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823656296984676e-05,
+      "loss": 0.9963,
+      "step": 5369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4821756065556571e-05,
+      "loss": 1.1397,
+      "step": 5370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4819855607252804e-05,
+      "loss": 1.017,
+      "step": 5371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817954922162791e-05,
+      "loss": 0.9672,
+      "step": 5372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4816054010375965e-05,
+      "loss": 1.0376,
+      "step": 5373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814152871981766e-05,
+      "loss": 0.9336,
+      "step": 5374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481225150706965e-05,
+      "loss": 0.9659,
+      "step": 5375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4810349915729085e-05,
+      "loss": 1.0167,
+      "step": 5376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808448098049537e-05,
+      "loss": 1.0944,
+      "step": 5377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806546054120498e-05,
+      "loss": 0.9727,
+      "step": 5378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804643784031463e-05,
+      "loss": 1.0304,
+      "step": 5379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802741287871935e-05,
+      "loss": 1.019,
+      "step": 5380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800838565731435e-05,
+      "loss": 1.0346,
+      "step": 5381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479893561769949e-05,
+      "loss": 1.0136,
+      "step": 5382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797032443865635e-05,
+      "loss": 1.0349,
+      "step": 5383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795129044319425e-05,
+      "loss": 1.0629,
+      "step": 5384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4793225419150414e-05,
+      "loss": 0.8824,
+      "step": 5385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791321568448176e-05,
+      "loss": 1.089,
+      "step": 5386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4789417492302289e-05,
+      "loss": 0.8691,
+      "step": 5387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4787513190802351e-05,
+      "loss": 1.0146,
+      "step": 5388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785608664037955e-05,
+      "loss": 0.9977,
+      "step": 5389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783703912098722e-05,
+      "loss": 0.994,
+      "step": 5390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781798935074268e-05,
+      "loss": 1.0482,
+      "step": 5391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4779893733054234e-05,
+      "loss": 1.0784,
+      "step": 5392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4777988306128256e-05,
+      "loss": 0.9957,
+      "step": 5393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776082654386e-05,
+      "loss": 1.0137,
+      "step": 5394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477417677791712e-05,
+      "loss": 1.0978,
+      "step": 5395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772270676811302e-05,
+      "loss": 1.0492,
+      "step": 5396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770364351158228e-05,
+      "loss": 1.0142,
+      "step": 5397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768457801047595e-05,
+      "loss": 1.0029,
+      "step": 5398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766551026569114e-05,
+      "loss": 1.0197,
+      "step": 5399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764644027812501e-05,
+      "loss": 0.9677,
+      "step": 5400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762736804867486e-05,
+      "loss": 0.9884,
+      "step": 5401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4760829357823809e-05,
+      "loss": 1.0231,
+      "step": 5402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4758921686771216e-05,
+      "loss": 0.9735,
+      "step": 5403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4757013791799476e-05,
+      "loss": 1.0622,
+      "step": 5404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755105672998353e-05,
+      "loss": 1.0227,
+      "step": 5405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4753197330457629e-05,
+      "loss": 1.0964,
+      "step": 5406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4751288764267103e-05,
+      "loss": 1.0067,
+      "step": 5407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4749379974516572e-05,
+      "loss": 1.0993,
+      "step": 5408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4747470961295849e-05,
+      "loss": 1.0804,
+      "step": 5409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4745561724694759e-05,
+      "loss": 1.059,
+      "step": 5410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743652264803136e-05,
+      "loss": 1.0297,
+      "step": 5411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474174258171083e-05,
+      "loss": 1.0,
+      "step": 5412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739832675507688e-05,
+      "loss": 0.9885,
+      "step": 5413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737922546283581e-05,
+      "loss": 0.985,
+      "step": 5414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736012194128386e-05,
+      "loss": 0.887,
+      "step": 5415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734101619131987e-05,
+      "loss": 1.023,
+      "step": 5416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473219082138428e-05,
+      "loss": 1.0528,
+      "step": 5417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4730279800975176e-05,
+      "loss": 0.9495,
+      "step": 5418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728368557994594e-05,
+      "loss": 1.0128,
+      "step": 5419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472645709253246e-05,
+      "loss": 1.0593,
+      "step": 5420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472454540467872e-05,
+      "loss": 0.9825,
+      "step": 5421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722633494523309e-05,
+      "loss": 0.9775,
+      "step": 5422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720721362156201e-05,
+      "loss": 0.9634,
+      "step": 5423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718809007667361e-05,
+      "loss": 0.9419,
+      "step": 5424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716896431146772e-05,
+      "loss": 0.95,
+      "step": 5425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714983632684423e-05,
+      "loss": 1.0764,
+      "step": 5426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713070612370319e-05,
+      "loss": 0.9756,
+      "step": 5427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471115737029447e-05,
+      "loss": 1.084,
+      "step": 5428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709243906546897e-05,
+      "loss": 1.0435,
+      "step": 5429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470733022121764e-05,
+      "loss": 1.0126,
+      "step": 5430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705416314396736e-05,
+      "loss": 1.0093,
+      "step": 5431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703502186174241e-05,
+      "loss": 0.9613,
+      "step": 5432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701587836640223e-05,
+      "loss": 1.0308,
+      "step": 5433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4699673265884752e-05,
+      "loss": 1.0295,
+      "step": 5434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697758473997916e-05,
+      "loss": 1.0799,
+      "step": 5435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469584346106981e-05,
+      "loss": 1.0008,
+      "step": 5436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693928227190538e-05,
+      "loss": 0.9468,
+      "step": 5437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4692012772450225e-05,
+      "loss": 1.0308,
+      "step": 5438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4690097096938984e-05,
+      "loss": 1.0398,
+      "step": 5439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688181200746967e-05,
+      "loss": 0.9846,
+      "step": 5440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468626508396431e-05,
+      "loss": 0.999,
+      "step": 5441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684348746681179e-05,
+      "loss": 1.0438,
+      "step": 5442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682432188987736e-05,
+      "loss": 0.9647,
+      "step": 5443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680515410974166e-05,
+      "loss": 1.0054,
+      "step": 5444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678598412730654e-05,
+      "loss": 1.0317,
+      "step": 5445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676681194347403e-05,
+      "loss": 1.0628,
+      "step": 5446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4674763755914616e-05,
+      "loss": 0.9834,
+      "step": 5447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672846097522519e-05,
+      "loss": 1.0528,
+      "step": 5448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670928219261345e-05,
+      "loss": 1.0182,
+      "step": 5449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466901012122133e-05,
+      "loss": 0.9472,
+      "step": 5450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667091803492726e-05,
+      "loss": 1.0432,
+      "step": 5451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4665173266165798e-05,
+      "loss": 1.0383,
+      "step": 5452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663254509330811e-05,
+      "loss": 1.0161,
+      "step": 5453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661335533078056e-05,
+      "loss": 1.0298,
+      "step": 5454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465941633749782e-05,
+      "loss": 1.0727,
+      "step": 5455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4657496922680408e-05,
+      "loss": 1.001,
+      "step": 5456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655577288716131e-05,
+      "loss": 1.0073,
+      "step": 5457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653657435695317e-05,
+      "loss": 1.0367,
+      "step": 5458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651737363708296e-05,
+      "loss": 0.9779,
+      "step": 5459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649817072845411e-05,
+      "loss": 1.0319,
+      "step": 5460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647896563197021e-05,
+      "loss": 0.9829,
+      "step": 5461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464597583485349e-05,
+      "loss": 1.0509,
+      "step": 5462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464405488790519e-05,
+      "loss": 0.9763,
+      "step": 5463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642133722442507e-05,
+      "loss": 1.0781,
+      "step": 5464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640212338555837e-05,
+      "loss": 0.9899,
+      "step": 5465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463829073633559e-05,
+      "loss": 1.0752,
+      "step": 5466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636368915872177e-05,
+      "loss": 1.0396,
+      "step": 5467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634446877256026e-05,
+      "loss": 0.957,
+      "step": 5468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4632524620577572e-05,
+      "loss": 1.082,
+      "step": 5469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630602145927267e-05,
+      "loss": 1.0152,
+      "step": 5470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628679453395563e-05,
+      "loss": 1.0128,
+      "step": 5471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462675654307293e-05,
+      "loss": 1.0292,
+      "step": 5472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624833415049845e-05,
+      "loss": 1.0221,
+      "step": 5473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622910069416799e-05,
+      "loss": 1.0102,
+      "step": 5474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620986506264286e-05,
+      "loss": 0.9919,
+      "step": 5475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619062725682812e-05,
+      "loss": 1.0492,
+      "step": 5476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617138727762905e-05,
+      "loss": 1.0072,
+      "step": 5477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615214512595087e-05,
+      "loss": 0.9724,
+      "step": 5478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613290080269898e-05,
+      "loss": 0.9888,
+      "step": 5479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611365430877886e-05,
+      "loss": 0.9604,
+      "step": 5480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609440564509616e-05,
+      "loss": 1.0008,
+      "step": 5481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607515481255652e-05,
+      "loss": 1.0721,
+      "step": 5482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605590181206579e-05,
+      "loss": 1.0423,
+      "step": 5483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4603664664452982e-05,
+      "loss": 1.0507,
+      "step": 5484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601738931085465e-05,
+      "loss": 1.0529,
+      "step": 5485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599812981194636e-05,
+      "loss": 1.0244,
+      "step": 5486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597886814871121e-05,
+      "loss": 1.035,
+      "step": 5487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595960432205542e-05,
+      "loss": 1.005,
+      "step": 5488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459403383328855e-05,
+      "loss": 0.9268,
+      "step": 5489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592107018210788e-05,
+      "loss": 0.9999,
+      "step": 5490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4590179987062922e-05,
+      "loss": 1.0973,
+      "step": 5491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588252739935622e-05,
+      "loss": 1.0127,
+      "step": 5492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586325276919573e-05,
+      "loss": 1.03,
+      "step": 5493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4584397598105462e-05,
+      "loss": 1.0657,
+      "step": 5494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582469703583994e-05,
+      "loss": 1.0059,
+      "step": 5495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458054159344588e-05,
+      "loss": 1.0855,
+      "step": 5496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578613267781847e-05,
+      "loss": 0.9785,
+      "step": 5497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457668472668262e-05,
+      "loss": 1.0157,
+      "step": 5498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574755970238945e-05,
+      "loss": 1.0043,
+      "step": 5499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572826998541576e-05,
+      "loss": 1.0129,
+      "step": 5500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4570897811681275e-05,
+      "loss": 0.9885,
+      "step": 5501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568968409748818e-05,
+      "loss": 1.0019,
+      "step": 5502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4567038792834982e-05,
+      "loss": 1.0806,
+      "step": 5503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4565108961030562e-05,
+      "loss": 0.9753,
+      "step": 5504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4563178914426366e-05,
+      "loss": 1.0149,
+      "step": 5505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4561248653113204e-05,
+      "loss": 1.0464,
+      "step": 5506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45593181771819e-05,
+      "loss": 0.9877,
+      "step": 5507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557387486723289e-05,
+      "loss": 0.9825,
+      "step": 5508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555456581828212e-05,
+      "loss": 1.0333,
+      "step": 5509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553525462587526e-05,
+      "loss": 0.9986,
+      "step": 5510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4551594129092092e-05,
+      "loss": 1.0162,
+      "step": 5511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549662581432787e-05,
+      "loss": 0.8971,
+      "step": 5512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547730819700492e-05,
+      "loss": 1.0205,
+      "step": 5513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545798843986104e-05,
+      "loss": 1.0089,
+      "step": 5514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543866654380525e-05,
+      "loss": 1.0313,
+      "step": 5515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4541934250974675e-05,
+      "loss": 1.0625,
+      "step": 5516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540001633859468e-05,
+      "loss": 1.0928,
+      "step": 5517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538068803125849e-05,
+      "loss": 1.0658,
+      "step": 5518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536135758864753e-05,
+      "loss": 1.088,
+      "step": 5519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4534202501167144e-05,
+      "loss": 1.0189,
+      "step": 5520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4532269030123981e-05,
+      "loss": 0.888,
+      "step": 5521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453033534582624e-05,
+      "loss": 0.9238,
+      "step": 5522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528401448364906e-05,
+      "loss": 1.0237,
+      "step": 5523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526467337830972e-05,
+      "loss": 0.9404,
+      "step": 5524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4524533014315445e-05,
+      "loss": 1.0628,
+      "step": 5525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452259847790934e-05,
+      "loss": 1.0525,
+      "step": 5526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452066372870368e-05,
+      "loss": 1.0288,
+      "step": 5527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4518728766789496e-05,
+      "loss": 0.9876,
+      "step": 5528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516793592257841e-05,
+      "loss": 1.0144,
+      "step": 5529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514858205199765e-05,
+      "loss": 1.0388,
+      "step": 5530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4512922605706332e-05,
+      "loss": 0.9902,
+      "step": 5531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451098679386862e-05,
+      "loss": 0.9902,
+      "step": 5532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4509050769777711e-05,
+      "loss": 1.0011,
+      "step": 5533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45071145335247e-05,
+      "loss": 0.9555,
+      "step": 5534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505178085200697e-05,
+      "loss": 1.1164,
+      "step": 5535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503241424896808e-05,
+      "loss": 1.0399,
+      "step": 5536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501304552704166e-05,
+      "loss": 1.0377,
+      "step": 5537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499367468713896e-05,
+      "loss": 0.997,
+      "step": 5538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497430173017154e-05,
+      "loss": 0.947,
+      "step": 5539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495492665705086e-05,
+      "loss": 0.9492,
+      "step": 5540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493554946868862e-05,
+      "loss": 0.9744,
+      "step": 5541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449161701659965e-05,
+      "loss": 1.0235,
+      "step": 5542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489678874988644e-05,
+      "loss": 1.0706,
+      "step": 5543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4487740522127029e-05,
+      "loss": 1.0225,
+      "step": 5544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4485801958106016e-05,
+      "loss": 0.9388,
+      "step": 5545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483863183016816e-05,
+      "loss": 0.9928,
+      "step": 5546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4481924196950652e-05,
+      "loss": 1.0475,
+      "step": 5547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447998499999876e-05,
+      "loss": 1.0413,
+      "step": 5548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478045592252388e-05,
+      "loss": 1.0483,
+      "step": 5549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447610597380278e-05,
+      "loss": 0.9793,
+      "step": 5550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447416614474121e-05,
+      "loss": 0.9884,
+      "step": 5551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472226105158949e-05,
+      "loss": 1.046,
+      "step": 5552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470285855147274e-05,
+      "loss": 0.9662,
+      "step": 5553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4468345394797488e-05,
+      "loss": 0.9836,
+      "step": 5554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466404724200888e-05,
+      "loss": 1.0254,
+      "step": 5555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464463843448792e-05,
+      "loss": 0.9769,
+      "step": 5556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462522752632519e-05,
+      "loss": 1.0043,
+      "step": 5557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460581451843407e-05,
+      "loss": 1.0215,
+      "step": 5558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458639941172796e-05,
+      "loss": 0.9761,
+      "step": 5559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445669822071204e-05,
+      "loss": 1.0424,
+      "step": 5560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454756290552498e-05,
+      "loss": 1.0302,
+      "step": 5561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452814150785547e-05,
+      "loss": 0.9989,
+      "step": 5562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450871801502571e-05,
+      "loss": 0.9442,
+      "step": 5563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448929242794958e-05,
+      "loss": 1.0063,
+      "step": 5564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444698647475411e-05,
+      "loss": 1.0353,
+      "step": 5565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445043497471443e-05,
+      "loss": 0.9945,
+      "step": 5566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443100311038376e-05,
+      "loss": 1.0198,
+      "step": 5567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441156915546343e-05,
+      "loss": 0.9689,
+      "step": 5568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4439213311086783e-05,
+      "loss": 0.9985,
+      "step": 5569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443726949775115e-05,
+      "loss": 1.0404,
+      "step": 5570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4435325475630901e-05,
+      "loss": 0.9147,
+      "step": 5571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433381244817515e-05,
+      "loss": 1.0097,
+      "step": 5572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431436805402466e-05,
+      "loss": 0.9531,
+      "step": 5573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429492157477248e-05,
+      "loss": 1.0176,
+      "step": 5574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427547301133356e-05,
+      "loss": 1.0625,
+      "step": 5575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4425602236462309e-05,
+      "loss": 0.996,
+      "step": 5576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442365696355562e-05,
+      "loss": 1.1011,
+      "step": 5577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421711482504825e-05,
+      "loss": 0.9649,
+      "step": 5578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419765793401457e-05,
+      "loss": 1.0523,
+      "step": 5579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417819896337073e-05,
+      "loss": 1.0396,
+      "step": 5580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415873791403225e-05,
+      "loss": 1.0197,
+      "step": 5581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441392747869149e-05,
+      "loss": 1.066,
+      "step": 5582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4411980958293436e-05,
+      "loss": 1.0264,
+      "step": 5583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410034230300663e-05,
+      "loss": 0.9574,
+      "step": 5584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4408087294804762e-05,
+      "loss": 1.0396,
+      "step": 5585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406140151897343e-05,
+      "loss": 1.0251,
+      "step": 5586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4404192801670026e-05,
+      "loss": 1.0605,
+      "step": 5587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440224524421444e-05,
+      "loss": 1.0235,
+      "step": 5588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400297479622217e-05,
+      "loss": 0.8873,
+      "step": 5589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398349507985003e-05,
+      "loss": 0.9301,
+      "step": 5590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4396401329394465e-05,
+      "loss": 1.0695,
+      "step": 5591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439445294394226e-05,
+      "loss": 1.0392,
+      "step": 5592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439250435172007e-05,
+      "loss": 1.0393,
+      "step": 5593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390555552819576e-05,
+      "loss": 1.0816,
+      "step": 5594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438860654733248e-05,
+      "loss": 1.0502,
+      "step": 5595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438665733535048e-05,
+      "loss": 0.9723,
+      "step": 5596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.43847079169653e-05,
+      "loss": 0.9756,
+      "step": 5597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438275829226866e-05,
+      "loss": 0.9306,
+      "step": 5598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4380808461352295e-05,
+      "loss": 0.9776,
+      "step": 5599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378858424307949e-05,
+      "loss": 1.0655,
+      "step": 5600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437690818122738e-05,
+      "loss": 1.0604,
+      "step": 5601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374957732202345e-05,
+      "loss": 0.9186,
+      "step": 5602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373007077324625e-05,
+      "loss": 0.9762,
+      "step": 5603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371056216685995e-05,
+      "loss": 1.0081,
+      "step": 5604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369105150378255e-05,
+      "loss": 1.0133,
+      "step": 5605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4367153878493204e-05,
+      "loss": 1.0125,
+      "step": 5606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365202401122657e-05,
+      "loss": 1.0593,
+      "step": 5607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363250718358433e-05,
+      "loss": 0.9817,
+      "step": 5608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361298830292363e-05,
+      "loss": 1.012,
+      "step": 5609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359346737016294e-05,
+      "loss": 0.8922,
+      "step": 5610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4357394438622067e-05,
+      "loss": 1.0139,
+      "step": 5611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435544193520155e-05,
+      "loss": 1.0532,
+      "step": 5612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4353489226846613e-05,
+      "loss": 1.0146,
+      "step": 5613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4351536313649134e-05,
+      "loss": 1.028,
+      "step": 5614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349583195701e-05,
+      "loss": 1.0029,
+      "step": 5615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347629873094114e-05,
+      "loss": 1.0498,
+      "step": 5616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345676345920385e-05,
+      "loss": 0.969,
+      "step": 5617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343722614271727e-05,
+      "loss": 1.0261,
+      "step": 5618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4341768678240072e-05,
+      "loss": 0.981,
+      "step": 5619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339814537917356e-05,
+      "loss": 1.0508,
+      "step": 5620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4337860193395526e-05,
+      "loss": 0.9686,
+      "step": 5621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335905644766541e-05,
+      "loss": 1.0716,
+      "step": 5622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333950892122361e-05,
+      "loss": 1.0928,
+      "step": 5623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433199593555497e-05,
+      "loss": 0.9717,
+      "step": 5624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330040775156351e-05,
+      "loss": 0.9354,
+      "step": 5625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328085411018497e-05,
+      "loss": 0.9791,
+      "step": 5626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326129843233413e-05,
+      "loss": 0.9592,
+      "step": 5627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324174071893115e-05,
+      "loss": 0.9734,
+      "step": 5628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4322218097089624e-05,
+      "loss": 1.0179,
+      "step": 5629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4320261918914981e-05,
+      "loss": 1.0695,
+      "step": 5630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318305537461221e-05,
+      "loss": 1.0133,
+      "step": 5631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.43163489528204e-05,
+      "loss": 0.9708,
+      "step": 5632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4314392165084578e-05,
+      "loss": 0.9913,
+      "step": 5633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312435174345831e-05,
+      "loss": 0.9898,
+      "step": 5634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4310477980696237e-05,
+      "loss": 0.9698,
+      "step": 5635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430852058422789e-05,
+      "loss": 0.9609,
+      "step": 5636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306562985032884e-05,
+      "loss": 1.0924,
+      "step": 5637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304605183203335e-05,
+      "loss": 1.0316,
+      "step": 5638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302647178831359e-05,
+      "loss": 1.0009,
+      "step": 5639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430068897200909e-05,
+      "loss": 1.0282,
+      "step": 5640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4298730562828658e-05,
+      "loss": 0.9528,
+      "step": 5641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4296771951382218e-05,
+      "loss": 1.0138,
+      "step": 5642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294813137761928e-05,
+      "loss": 0.9919,
+      "step": 5643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292854122059949e-05,
+      "loss": 0.9816,
+      "step": 5644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4290894904368463e-05,
+      "loss": 0.9965,
+      "step": 5645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4288935484779653e-05,
+      "loss": 0.9954,
+      "step": 5646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4286975863385718e-05,
+      "loss": 1.0421,
+      "step": 5647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4285016040278858e-05,
+      "loss": 1.0554,
+      "step": 5648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283056015551295e-05,
+      "loss": 0.953,
+      "step": 5649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281095789295245e-05,
+      "loss": 0.9888,
+      "step": 5650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279135361602947e-05,
+      "loss": 1.0862,
+      "step": 5651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277174732566641e-05,
+      "loss": 0.9927,
+      "step": 5652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4275213902278581e-05,
+      "loss": 0.9891,
+      "step": 5653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4273252870831029e-05,
+      "loss": 1.0204,
+      "step": 5654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271291638316255e-05,
+      "loss": 0.9457,
+      "step": 5655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4269330204826542e-05,
+      "loss": 0.9895,
+      "step": 5656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4267368570454178e-05,
+      "loss": 1.0093,
+      "step": 5657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265406735291465e-05,
+      "loss": 0.9809,
+      "step": 5658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263444699430712e-05,
+      "loss": 0.9786,
+      "step": 5659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261482462964236e-05,
+      "loss": 0.977,
+      "step": 5660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259520025984368e-05,
+      "loss": 1.0016,
+      "step": 5661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425755738858344e-05,
+      "loss": 1.0418,
+      "step": 5662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255594550853806e-05,
+      "loss": 1.0256,
+      "step": 5663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253631512887816e-05,
+      "loss": 0.9812,
+      "step": 5664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425166827477784e-05,
+      "loss": 1.0227,
+      "step": 5665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249704836616254e-05,
+      "loss": 1.0611,
+      "step": 5666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424774119849544e-05,
+      "loss": 0.9769,
+      "step": 5667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4245777360507794e-05,
+      "loss": 1.0262,
+      "step": 5668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243813322745716e-05,
+      "loss": 1.0453,
+      "step": 5669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241849085301622e-05,
+      "loss": 1.0087,
+      "step": 5670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4239884648267933e-05,
+      "loss": 1.0123,
+      "step": 5671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237920011737086e-05,
+      "loss": 0.9381,
+      "step": 5672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235955175801513e-05,
+      "loss": 1.0111,
+      "step": 5673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233990140553668e-05,
+      "loss": 1.0806,
+      "step": 5674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232024906086013e-05,
+      "loss": 1.0205,
+      "step": 5675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4230059472491018e-05,
+      "loss": 0.9555,
+      "step": 5676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228093839861156e-05,
+      "loss": 0.9574,
+      "step": 5677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422612800828892e-05,
+      "loss": 1.0218,
+      "step": 5678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224161977866803e-05,
+      "loss": 1.0408,
+      "step": 5679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222195748687317e-05,
+      "loss": 1.0151,
+      "step": 5680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4220229320842973e-05,
+      "loss": 0.9418,
+      "step": 5681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218262694426301e-05,
+      "loss": 0.9434,
+      "step": 5682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421629586952983e-05,
+      "loss": 1.0331,
+      "step": 5683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421432884624611e-05,
+      "loss": 0.9513,
+      "step": 5684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421236162466769e-05,
+      "loss": 1.034,
+      "step": 5685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4210394204887137e-05,
+      "loss": 1.0314,
+      "step": 5686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208426586997018e-05,
+      "loss": 1.021,
+      "step": 5687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4206458771089917e-05,
+      "loss": 1.0517,
+      "step": 5688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204490757258422e-05,
+      "loss": 0.9349,
+      "step": 5689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4202522545595136e-05,
+      "loss": 0.9485,
+      "step": 5690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200554136192671e-05,
+      "loss": 1.0194,
+      "step": 5691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198585529143641e-05,
+      "loss": 0.9185,
+      "step": 5692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4196616724540673e-05,
+      "loss": 0.982,
+      "step": 5693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194647722476407e-05,
+      "loss": 1.0508,
+      "step": 5694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4192678523043486e-05,
+      "loss": 1.044,
+      "step": 5695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190709126334574e-05,
+      "loss": 1.0352,
+      "step": 5696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188739532442326e-05,
+      "loss": 1.0181,
+      "step": 5697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4186769741459423e-05,
+      "loss": 1.0425,
+      "step": 5698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184799753478542e-05,
+      "loss": 1.0127,
+      "step": 5699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182829568592384e-05,
+      "loss": 0.9644,
+      "step": 5700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180859186893647e-05,
+      "loss": 1.0159,
+      "step": 5701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178888608475045e-05,
+      "loss": 0.9334,
+      "step": 5702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176917833429291e-05,
+      "loss": 1.0671,
+      "step": 5703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174946861849123e-05,
+      "loss": 0.9399,
+      "step": 5704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4172975693827277e-05,
+      "loss": 0.9573,
+      "step": 5705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171004329456504e-05,
+      "loss": 1.0442,
+      "step": 5706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169032768829558e-05,
+      "loss": 1.0096,
+      "step": 5707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167061012039205e-05,
+      "loss": 1.0027,
+      "step": 5708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4165089059178225e-05,
+      "loss": 0.9737,
+      "step": 5709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163116910339399e-05,
+      "loss": 0.981,
+      "step": 5710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161144565615528e-05,
+      "loss": 1.0841,
+      "step": 5711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159172025099407e-05,
+      "loss": 0.9822,
+      "step": 5712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157199288883859e-05,
+      "loss": 1.0286,
+      "step": 5713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4155226357061698e-05,
+      "loss": 1.0039,
+      "step": 5714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415325322972576e-05,
+      "loss": 1.1189,
+      "step": 5715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151279906968882e-05,
+      "loss": 1.0224,
+      "step": 5716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149306388883917e-05,
+      "loss": 1.0244,
+      "step": 5717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147332675563721e-05,
+      "loss": 0.9307,
+      "step": 5718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145358767101166e-05,
+      "loss": 1.0002,
+      "step": 5719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4143384663589123e-05,
+      "loss": 0.9584,
+      "step": 5720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141410365120486e-05,
+      "loss": 0.9654,
+      "step": 5721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139435871788147e-05,
+      "loss": 0.9944,
+      "step": 5722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413746118368501e-05,
+      "loss": 1.0403,
+      "step": 5723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135486300903991e-05,
+      "loss": 1.0048,
+      "step": 5724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4133511223538015e-05,
+      "loss": 0.9638,
+      "step": 5725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4131535951680006e-05,
+      "loss": 1.0162,
+      "step": 5726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129560485422915e-05,
+      "loss": 1.0679,
+      "step": 5727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4127584824859686e-05,
+      "loss": 1.0296,
+      "step": 5728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4125608970083283e-05,
+      "loss": 1.0147,
+      "step": 5729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123632921186674e-05,
+      "loss": 1.0606,
+      "step": 5730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121656678262836e-05,
+      "loss": 1.0156,
+      "step": 5731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4119680241404756e-05,
+      "loss": 0.9912,
+      "step": 5732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117703610705429e-05,
+      "loss": 1.0428,
+      "step": 5733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115726786257865e-05,
+      "loss": 0.9507,
+      "step": 5734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4113749768155077e-05,
+      "loss": 1.0315,
+      "step": 5735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4111772556490085e-05,
+      "loss": 0.9742,
+      "step": 5736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109795151355922e-05,
+      "loss": 1.1125,
+      "step": 5737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4107817552845635e-05,
+      "loss": 0.9327,
+      "step": 5738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4105839761052274e-05,
+      "loss": 0.98,
+      "step": 5739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4103861776068892e-05,
+      "loss": 0.9907,
+      "step": 5740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4101883597988566e-05,
+      "loss": 1.0596,
+      "step": 5741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409990522690437e-05,
+      "loss": 1.0471,
+      "step": 5742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4097926662909393e-05,
+      "loss": 0.9417,
+      "step": 5743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095947906096734e-05,
+      "loss": 0.9531,
+      "step": 5744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093968956559491e-05,
+      "loss": 0.9791,
+      "step": 5745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4091989814390788e-05,
+      "loss": 1.032,
+      "step": 5746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090010479683738e-05,
+      "loss": 1.0066,
+      "step": 5747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088030952531484e-05,
+      "loss": 1.0167,
+      "step": 5748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086051233027161e-05,
+      "loss": 1.1184,
+      "step": 5749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4084071321263922e-05,
+      "loss": 0.9755,
+      "step": 5750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4082091217334923e-05,
+      "loss": 1.0353,
+      "step": 5751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408011092133334e-05,
+      "loss": 0.9709,
+      "step": 5752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078130433352348e-05,
+      "loss": 1.0218,
+      "step": 5753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407614975348513e-05,
+      "loss": 0.9679,
+      "step": 5754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074168881824886e-05,
+      "loss": 1.0184,
+      "step": 5755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072187818464821e-05,
+      "loss": 1.0454,
+      "step": 5756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4070206563498147e-05,
+      "loss": 1.0052,
+      "step": 5757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406822511701809e-05,
+      "loss": 0.995,
+      "step": 5758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4066243479117878e-05,
+      "loss": 0.9968,
+      "step": 5759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4064261649890755e-05,
+      "loss": 0.9612,
+      "step": 5760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406227962942997e-05,
+      "loss": 0.9802,
+      "step": 5761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4060297417828781e-05,
+      "loss": 1.0574,
+      "step": 5762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058315015180457e-05,
+      "loss": 0.9719,
+      "step": 5763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056332421578276e-05,
+      "loss": 0.909,
+      "step": 5764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4054349637115522e-05,
+      "loss": 1.0722,
+      "step": 5765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405236666188549e-05,
+      "loss": 0.9395,
+      "step": 5766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4050383495981488e-05,
+      "loss": 0.9686,
+      "step": 5767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4048400139496825e-05,
+      "loss": 1.0474,
+      "step": 5768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4046416592524822e-05,
+      "loss": 0.9981,
+      "step": 5769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044432855158812e-05,
+      "loss": 1.0441,
+      "step": 5770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4042448927492133e-05,
+      "loss": 1.0438,
+      "step": 5771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040464809618137e-05,
+      "loss": 0.9532,
+      "step": 5772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403848050163018e-05,
+      "loss": 1.0322,
+      "step": 5773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036496003621628e-05,
+      "loss": 1.0239,
+      "step": 5774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4034511315685857e-05,
+      "loss": 1.0103,
+      "step": 5775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403252643791625e-05,
+      "loss": 1.0066,
+      "step": 5776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030541370406206e-05,
+      "loss": 0.9919,
+      "step": 5777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402855611324912e-05,
+      "loss": 1.0186,
+      "step": 5778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026570666538405e-05,
+      "loss": 1.0267,
+      "step": 5779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4024585030367487e-05,
+      "loss": 1.046,
+      "step": 5780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022599204829788e-05,
+      "loss": 0.9182,
+      "step": 5781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402061319001875e-05,
+      "loss": 0.9645,
+      "step": 5782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4018626986027821e-05,
+      "loss": 1.0119,
+      "step": 5783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016640592950453e-05,
+      "loss": 1.0225,
+      "step": 5784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4014654010880112e-05,
+      "loss": 0.9376,
+      "step": 5785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012667239910273e-05,
+      "loss": 1.0171,
+      "step": 5786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4010680280134419e-05,
+      "loss": 1.0549,
+      "step": 5787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008693131646038e-05,
+      "loss": 0.9869,
+      "step": 5788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006705794538636e-05,
+      "loss": 0.9674,
+      "step": 5789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004718268905714e-05,
+      "loss": 1.0534,
+      "step": 5790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40027305548408e-05,
+      "loss": 1.0367,
+      "step": 5791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4000742652437412e-05,
+      "loss": 1.0172,
+      "step": 5792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399875456178909e-05,
+      "loss": 0.9836,
+      "step": 5793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996766282989378e-05,
+      "loss": 1.0769,
+      "step": 5794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399477781613183e-05,
+      "loss": 1.0766,
+      "step": 5795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992789161310007e-05,
+      "loss": 0.9802,
+      "step": 5796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990800318617484e-05,
+      "loss": 1.013,
+      "step": 5797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3988811288147836e-05,
+      "loss": 1.0404,
+      "step": 5798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986822069994654e-05,
+      "loss": 0.9693,
+      "step": 5799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984832664251538e-05,
+      "loss": 1.0593,
+      "step": 5800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3982843071012088e-05,
+      "loss": 0.9604,
+      "step": 5801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980853290369926e-05,
+      "loss": 1.0528,
+      "step": 5802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3978863322418671e-05,
+      "loss": 1.0177,
+      "step": 5803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976873167251962e-05,
+      "loss": 0.9563,
+      "step": 5804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974882824963437e-05,
+      "loss": 0.9555,
+      "step": 5805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972892295646748e-05,
+      "loss": 1.0654,
+      "step": 5806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3970901579395549e-05,
+      "loss": 1.0014,
+      "step": 5807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968910676303516e-05,
+      "loss": 1.0479,
+      "step": 5808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396691958646432e-05,
+      "loss": 0.9789,
+      "step": 5809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3964928309971652e-05,
+      "loss": 0.997,
+      "step": 5810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962936846919202e-05,
+      "loss": 1.0249,
+      "step": 5811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960945197400675e-05,
+      "loss": 1.0577,
+      "step": 5812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958953361509782e-05,
+      "loss": 0.9188,
+      "step": 5813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3956961339340247e-05,
+      "loss": 1.0061,
+      "step": 5814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954969130985798e-05,
+      "loss": 0.9832,
+      "step": 5815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952976736540174e-05,
+      "loss": 0.974,
+      "step": 5816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950984156097115e-05,
+      "loss": 0.9805,
+      "step": 5817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948991389750388e-05,
+      "loss": 0.9747,
+      "step": 5818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3946998437593753e-05,
+      "loss": 1.1134,
+      "step": 5819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945005299720982e-05,
+      "loss": 0.9417,
+      "step": 5820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943011976225858e-05,
+      "loss": 0.9468,
+      "step": 5821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3941018467202173e-05,
+      "loss": 1.0704,
+      "step": 5822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939024772743723e-05,
+      "loss": 0.9951,
+      "step": 5823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3937030892944323e-05,
+      "loss": 0.976,
+      "step": 5824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935036827897782e-05,
+      "loss": 1.0082,
+      "step": 5825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393304257769793e-05,
+      "loss": 0.9613,
+      "step": 5826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931048142438601e-05,
+      "loss": 0.9686,
+      "step": 5827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392905352221364e-05,
+      "loss": 0.9486,
+      "step": 5828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39270587171169e-05,
+      "loss": 1.0305,
+      "step": 5829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925063727242231e-05,
+      "loss": 1.0553,
+      "step": 5830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3923068552683514e-05,
+      "loss": 1.0478,
+      "step": 5831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921073193534622e-05,
+      "loss": 1.0125,
+      "step": 5832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3919077649889445e-05,
+      "loss": 0.9983,
+      "step": 5833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3917081921841873e-05,
+      "loss": 1.0837,
+      "step": 5834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915086009485817e-05,
+      "loss": 1.0082,
+      "step": 5835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913089912915179e-05,
+      "loss": 1.0209,
+      "step": 5836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911093632223891e-05,
+      "loss": 1.0224,
+      "step": 5837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909097167505877e-05,
+      "loss": 1.019,
+      "step": 5838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390710051885508e-05,
+      "loss": 0.9403,
+      "step": 5839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905103686365442e-05,
+      "loss": 1.0096,
+      "step": 5840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903106670130923e-05,
+      "loss": 0.9558,
+      "step": 5841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3901109470245483e-05,
+      "loss": 1.0997,
+      "step": 5842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899112086803106e-05,
+      "loss": 0.9669,
+      "step": 5843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389711451989776e-05,
+      "loss": 1.0132,
+      "step": 5844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895116769623445e-05,
+      "loss": 0.9849,
+      "step": 5845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893118836074155e-05,
+      "loss": 1.0404,
+      "step": 5846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38911207193439e-05,
+      "loss": 0.934,
+      "step": 5847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38891224195267e-05,
+      "loss": 0.9743,
+      "step": 5848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3887123936716573e-05,
+      "loss": 1.0757,
+      "step": 5849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3885125271007557e-05,
+      "loss": 1.0189,
+      "step": 5850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3883126422493692e-05,
+      "loss": 1.002,
+      "step": 5851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881127391269029e-05,
+      "loss": 0.988,
+      "step": 5852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387912817742763e-05,
+      "loss": 0.9303,
+      "step": 5853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3877128781063558e-05,
+      "loss": 0.9858,
+      "step": 5854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875129202270898e-05,
+      "loss": 1.0212,
+      "step": 5855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3873129441143726e-05,
+      "loss": 1.0037,
+      "step": 5856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3871129497776141e-05,
+      "loss": 0.9994,
+      "step": 5857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386912937226225e-05,
+      "loss": 1.0639,
+      "step": 5858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867129064696147e-05,
+      "loss": 0.9399,
+      "step": 5859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3865128575171972e-05,
+      "loss": 0.9802,
+      "step": 5860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863127903783835e-05,
+      "loss": 0.9708,
+      "step": 5861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3861127050625891e-05,
+      "loss": 1.0651,
+      "step": 5862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859126015792267e-05,
+      "loss": 1.03,
+      "step": 5863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385712479937713e-05,
+      "loss": 1.0234,
+      "step": 5864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855123401474633e-05,
+      "loss": 0.9876,
+      "step": 5865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3853121822178953e-05,
+      "loss": 1.0072,
+      "step": 5866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3851120061584267e-05,
+      "loss": 1.0309,
+      "step": 5867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849118119784764e-05,
+      "loss": 0.939,
+      "step": 5868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3847115996874639e-05,
+      "loss": 1.0754,
+      "step": 5869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845113692948097e-05,
+      "loss": 0.9951,
+      "step": 5870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3843111208099349e-05,
+      "loss": 0.9796,
+      "step": 5871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3841108542422626e-05,
+      "loss": 0.9942,
+      "step": 5872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839105696012148e-05,
+      "loss": 1.0078,
+      "step": 5873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837102668962157e-05,
+      "loss": 0.9589,
+      "step": 5874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3835099461366903e-05,
+      "loss": 1.0296,
+      "step": 5875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383309607332064e-05,
+      "loss": 1.014,
+      "step": 5876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3831092504917634e-05,
+      "loss": 1.0487,
+      "step": 5877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829088756252156e-05,
+      "loss": 0.9752,
+      "step": 5878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827084827418486e-05,
+      "loss": 0.932,
+      "step": 5879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382508071851092e-05,
+      "loss": 1.0138,
+      "step": 5880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382307642962375e-05,
+      "loss": 1.0202,
+      "step": 5881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821071960851285e-05,
+      "loss": 0.9872,
+      "step": 5882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819067312287837e-05,
+      "loss": 1.0619,
+      "step": 5883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381706248402774e-05,
+      "loss": 0.8914,
+      "step": 5884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815057476165314e-05,
+      "loss": 1.0474,
+      "step": 5885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813052288794906e-05,
+      "loss": 1.0605,
+      "step": 5886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811046922010865e-05,
+      "loss": 1.0031,
+      "step": 5887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809041375907545e-05,
+      "loss": 1.0744,
+      "step": 5888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807035650579318e-05,
+      "loss": 1.0234,
+      "step": 5889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3805029746120549e-05,
+      "loss": 1.0132,
+      "step": 5890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803023662625632e-05,
+      "loss": 0.9839,
+      "step": 5891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380101740018895e-05,
+      "loss": 1.017,
+      "step": 5892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3799010958904905e-05,
+      "loss": 0.9837,
+      "step": 5893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797004338867907e-05,
+      "loss": 0.9928,
+      "step": 5894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794997540172373e-05,
+      "loss": 1.0123,
+      "step": 5895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792990562912722e-05,
+      "loss": 1.01,
+      "step": 5896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790983407183394e-05,
+      "loss": 0.9967,
+      "step": 5897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788976073078825e-05,
+      "loss": 0.9523,
+      "step": 5898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378696856069347e-05,
+      "loss": 0.9576,
+      "step": 5899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784960870121786e-05,
+      "loss": 1.0169,
+      "step": 5900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378295300145824e-05,
+      "loss": 0.953,
+      "step": 5901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780944954797301e-05,
+      "loss": 1.0282,
+      "step": 5902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778936730233463e-05,
+      "loss": 1.0774,
+      "step": 5903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776928327861212e-05,
+      "loss": 1.0278,
+      "step": 5904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774919747775051e-05,
+      "loss": 1.071,
+      "step": 5905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772910990069486e-05,
+      "loss": 1.0079,
+      "step": 5906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770902054839035e-05,
+      "loss": 0.985,
+      "step": 5907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768892942178223e-05,
+      "loss": 0.9452,
+      "step": 5908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766883652181583e-05,
+      "loss": 1.0321,
+      "step": 5909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3764874184943664e-05,
+      "loss": 1.0081,
+      "step": 5910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762864540559007e-05,
+      "loss": 0.9585,
+      "step": 5911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3760854719122175e-05,
+      "loss": 0.9539,
+      "step": 5912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758844720727736e-05,
+      "loss": 1.0409,
+      "step": 5913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756834545470263e-05,
+      "loss": 0.9457,
+      "step": 5914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3754824193444342e-05,
+      "loss": 1.0153,
+      "step": 5915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3752813664744564e-05,
+      "loss": 0.9945,
+      "step": 5916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3750802959465528e-05,
+      "loss": 1.0258,
+      "step": 5917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748792077701847e-05,
+      "loss": 1.1246,
+      "step": 5918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746781019548135e-05,
+      "loss": 1.0569,
+      "step": 5919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744769785099017e-05,
+      "loss": 1.025,
+      "step": 5920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742758374449126e-05,
+      "loss": 0.9658,
+      "step": 5921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3740746787693107e-05,
+      "loss": 1.0219,
+      "step": 5922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3738735024925607e-05,
+      "loss": 1.0399,
+      "step": 5923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736723086241288e-05,
+      "loss": 0.9882,
+      "step": 5924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734710971734813e-05,
+      "loss": 1.0365,
+      "step": 5925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732698681500859e-05,
+      "loss": 1.0361,
+      "step": 5926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3730686215634106e-05,
+      "loss": 0.9072,
+      "step": 5927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372867357422925e-05,
+      "loss": 1.0095,
+      "step": 5928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372666075738099e-05,
+      "loss": 0.9885,
+      "step": 5929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724647765184034e-05,
+      "loss": 1.0035,
+      "step": 5930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722634597733095e-05,
+      "loss": 0.9527,
+      "step": 5931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720621255122901e-05,
+      "loss": 1.0455,
+      "step": 5932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3718607737448184e-05,
+      "loss": 1.0546,
+      "step": 5933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716594044803686e-05,
+      "loss": 0.9928,
+      "step": 5934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714580177284153e-05,
+      "loss": 0.9606,
+      "step": 5935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712566134984346e-05,
+      "loss": 0.9644,
+      "step": 5936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371055191799903e-05,
+      "loss": 0.9296,
+      "step": 5937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370853752642298e-05,
+      "loss": 0.9946,
+      "step": 5938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706522960350972e-05,
+      "loss": 1.0805,
+      "step": 5939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704508219877804e-05,
+      "loss": 0.8881,
+      "step": 5940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370249330509827e-05,
+      "loss": 0.979,
+      "step": 5941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700478216107179e-05,
+      "loss": 0.9371,
+      "step": 5942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3698462952999344e-05,
+      "loss": 0.9739,
+      "step": 5943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369644751586959e-05,
+      "loss": 1.0406,
+      "step": 5944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694431904812751e-05,
+      "loss": 0.9802,
+      "step": 5945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369241611992366e-05,
+      "loss": 1.0222,
+      "step": 5946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690400161297167e-05,
+      "loss": 0.9684,
+      "step": 5947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368838402902813e-05,
+      "loss": 1.0691,
+      "step": 5948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686367723211414e-05,
+      "loss": 0.9701,
+      "step": 5949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3684351243941885e-05,
+      "loss": 0.9261,
+      "step": 5950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368233459131443e-05,
+      "loss": 1.0199,
+      "step": 5951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3680317765423932e-05,
+      "loss": 1.0127,
+      "step": 5952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3678300766365296e-05,
+      "loss": 0.9634,
+      "step": 5953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676283594233418e-05,
+      "loss": 0.9958,
+      "step": 5954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674266249123216e-05,
+      "loss": 0.9542,
+      "step": 5955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672248731129606e-05,
+      "loss": 1.015,
+      "step": 5956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670231040347528e-05,
+      "loss": 0.9741,
+      "step": 5957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668213176871907e-05,
+      "loss": 1.0377,
+      "step": 5958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666195140797695e-05,
+      "loss": 1.0328,
+      "step": 5959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3664176932219843e-05,
+      "loss": 1.0054,
+      "step": 5960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3662158551233317e-05,
+      "loss": 1.031,
+      "step": 5961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660139997933085e-05,
+      "loss": 1.0658,
+      "step": 5962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3658121272414122e-05,
+      "loss": 0.9629,
+      "step": 5963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3656102374771413e-05,
+      "loss": 0.8144,
+      "step": 5964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365408330509996e-05,
+      "loss": 1.0151,
+      "step": 5965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365206406349476e-05,
+      "loss": 1.0074,
+      "step": 5966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3650044650050824e-05,
+      "loss": 1.0383,
+      "step": 5967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648025064863168e-05,
+      "loss": 1.0003,
+      "step": 5968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3646005308026824e-05,
+      "loss": 0.95,
+      "step": 5969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3643985379636822e-05,
+      "loss": 1.0228,
+      "step": 5970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641965279788205e-05,
+      "loss": 0.9663,
+      "step": 5971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3639945008576028e-05,
+      "loss": 1.0908,
+      "step": 5972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3637924566095344e-05,
+      "loss": 0.9861,
+      "step": 5973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635903952441225e-05,
+      "loss": 1.0944,
+      "step": 5974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3633883167708741e-05,
+      "loss": 0.9819,
+      "step": 5975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3631862211992979e-05,
+      "loss": 0.9003,
+      "step": 5976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362984108538903e-05,
+      "loss": 1.062,
+      "step": 5977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3627819787991992e-05,
+      "loss": 0.9999,
+      "step": 5978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625798319896968e-05,
+      "loss": 1.0074,
+      "step": 5979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362377668119908e-05,
+      "loss": 1.0729,
+      "step": 5980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3621754871993446e-05,
+      "loss": 1.0392,
+      "step": 5981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619732892375199e-05,
+      "loss": 0.9342,
+      "step": 5982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3617710742439478e-05,
+      "loss": 0.9772,
+      "step": 5983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615688422281433e-05,
+      "loss": 0.9938,
+      "step": 5984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613665931996214e-05,
+      "loss": 1.1047,
+      "step": 5985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3611643271678992e-05,
+      "loss": 0.9972,
+      "step": 5986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609620441424926e-05,
+      "loss": 1.0108,
+      "step": 5987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607597441329207e-05,
+      "loss": 0.9878,
+      "step": 5988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3605574271487014e-05,
+      "loss": 0.9784,
+      "step": 5989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360355093199355e-05,
+      "loss": 0.9974,
+      "step": 5990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601527422944014e-05,
+      "loss": 0.9854,
+      "step": 5991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3599503744433616e-05,
+      "loss": 1.0625,
+      "step": 5992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597479896557575e-05,
+      "loss": 1.0078,
+      "step": 5993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359545587941112e-05,
+      "loss": 0.9646,
+      "step": 5994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3593431693089485e-05,
+      "loss": 0.9411,
+      "step": 5995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591407337687917e-05,
+      "loss": 1.0034,
+      "step": 5996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3589382813301657e-05,
+      "loss": 1.0099,
+      "step": 5997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587358120025975e-05,
+      "loss": 0.9828,
+      "step": 5998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3585333257956133e-05,
+      "loss": 1.0126,
+      "step": 5999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583308227187404e-05,
+      "loss": 0.9311,
+      "step": 6000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581283027815073e-05,
+      "loss": 0.976,
+      "step": 6001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3579257659934433e-05,
+      "loss": 1.0182,
+      "step": 6002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357723212364078e-05,
+      "loss": 1.0399,
+      "step": 6003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3575206419029416e-05,
+      "loss": 1.0259,
+      "step": 6004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573180546195665e-05,
+      "loss": 1.0336,
+      "step": 6005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3571154505234842e-05,
+      "loss": 0.9649,
+      "step": 6006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569128296242282e-05,
+      "loss": 1.0499,
+      "step": 6007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3567101919313318e-05,
+      "loss": 1.0508,
+      "step": 6008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565075374543302e-05,
+      "loss": 0.9532,
+      "step": 6009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3563048662027582e-05,
+      "loss": 0.9744,
+      "step": 6010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561021781861528e-05,
+      "loss": 1.0039,
+      "step": 6011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558994734140498e-05,
+      "loss": 0.9821,
+      "step": 6012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3556967518959882e-05,
+      "loss": 0.9934,
+      "step": 6013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554940136415058e-05,
+      "loss": 1.0063,
+      "step": 6014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355291258660142e-05,
+      "loss": 0.9516,
+      "step": 6015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550884869614372e-05,
+      "loss": 1.0016,
+      "step": 6016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354885698554932e-05,
+      "loss": 0.9581,
+      "step": 6017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546828934501686e-05,
+      "loss": 0.9993,
+      "step": 6018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3544800716566891e-05,
+      "loss": 1.0191,
+      "step": 6019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3542772331840367e-05,
+      "loss": 0.9571,
+      "step": 6020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540743780417555e-05,
+      "loss": 0.972,
+      "step": 6021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538715062393903e-05,
+      "loss": 1.0784,
+      "step": 6022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536686177864873e-05,
+      "loss": 1.0524,
+      "step": 6023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534657126925924e-05,
+      "loss": 1.0163,
+      "step": 6024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532627909672527e-05,
+      "loss": 0.9728,
+      "step": 6025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3530598526200162e-05,
+      "loss": 0.9949,
+      "step": 6026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528568976604321e-05,
+      "loss": 0.9865,
+      "step": 6027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526539260980495e-05,
+      "loss": 1.0771,
+      "step": 6028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352450937942419e-05,
+      "loss": 1.0992,
+      "step": 6029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352247933203091e-05,
+      "loss": 1.0715,
+      "step": 6030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520449118896186e-05,
+      "loss": 0.9604,
+      "step": 6031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3518418740115536e-05,
+      "loss": 1.0123,
+      "step": 6032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516388195784495e-05,
+      "loss": 0.9347,
+      "step": 6033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514357485998606e-05,
+      "loss": 0.9467,
+      "step": 6034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3512326610853421e-05,
+      "loss": 1.0195,
+      "step": 6035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510295570444494e-05,
+      "loss": 1.0144,
+      "step": 6036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508264364867395e-05,
+      "loss": 0.9861,
+      "step": 6037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506232994217694e-05,
+      "loss": 1.0164,
+      "step": 6038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3504201458590971e-05,
+      "loss": 1.0452,
+      "step": 6039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502169758082819e-05,
+      "loss": 1.0025,
+      "step": 6040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3500137892788831e-05,
+      "loss": 0.9496,
+      "step": 6041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498105862804615e-05,
+      "loss": 1.0027,
+      "step": 6042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349607366822578e-05,
+      "loss": 0.9968,
+      "step": 6043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349404130914795e-05,
+      "loss": 1.0694,
+      "step": 6044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3492008785666744e-05,
+      "loss": 1.0041,
+      "step": 6045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3489976097877804e-05,
+      "loss": 1.0512,
+      "step": 6046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.348794324587677e-05,
+      "loss": 1.0688,
+      "step": 6047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34859102297593e-05,
+      "loss": 1.0035,
+      "step": 6048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483877049621042e-05,
+      "loss": 0.9937,
+      "step": 6049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3481843705557669e-05,
+      "loss": 1.0243,
+      "step": 6050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479810197664849e-05,
+      "loss": 0.985,
+      "step": 6051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3477776526038273e-05,
+      "loss": 1.0846,
+      "step": 6052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347574269077362e-05,
+      "loss": 1.0567,
+      "step": 6053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473708691966598e-05,
+      "loss": 0.9766,
+      "step": 6054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3471674529712901e-05,
+      "loss": 0.9883,
+      "step": 6055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469640204108247e-05,
+      "loss": 0.988,
+      "step": 6056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467605715248358e-05,
+      "loss": 1.0184,
+      "step": 6057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465571063228957e-05,
+      "loss": 1.0485,
+      "step": 6058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463536248145779e-05,
+      "loss": 0.963,
+      "step": 6059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3461501270094572e-05,
+      "loss": 0.9543,
+      "step": 6060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3459466129171085e-05,
+      "loss": 1.0059,
+      "step": 6061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457430825471076e-05,
+      "loss": 1.0737,
+      "step": 6062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3455395359090309e-05,
+      "loss": 1.1409,
+      "step": 6063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345335973012456e-05,
+      "loss": 0.9948,
+      "step": 6064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451323938669614e-05,
+      "loss": 0.9721,
+      "step": 6065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3449287984821252e-05,
+      "loss": 0.9839,
+      "step": 6066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447251868675278e-05,
+      "loss": 0.944,
+      "step": 6067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3445215590327489e-05,
+      "loss": 0.9419,
+      "step": 6068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443179149873704e-05,
+      "loss": 1.0041,
+      "step": 6069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3441142547409737e-05,
+      "loss": 0.9643,
+      "step": 6070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439105783031426e-05,
+      "loss": 0.9361,
+      "step": 6071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343706885683459e-05,
+      "loss": 0.995,
+      "step": 6072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435031768915084e-05,
+      "loss": 1.0113,
+      "step": 6073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343299451936875e-05,
+      "loss": 0.9962,
+      "step": 6074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430957108291451e-05,
+      "loss": 1.0431,
+      "step": 6075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342891953577905e-05,
+      "loss": 0.9629,
+      "step": 6076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426881801927423e-05,
+      "loss": 0.9733,
+      "step": 6077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3424843906832443e-05,
+      "loss": 0.9855,
+      "step": 6078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3422805850590006e-05,
+      "loss": 0.988,
+      "step": 6079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3420767633296004e-05,
+      "loss": 1.03,
+      "step": 6080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418729255046344e-05,
+      "loss": 0.9733,
+      "step": 6081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416690715936929e-05,
+      "loss": 1.0973,
+      "step": 6082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3414652016063683e-05,
+      "loss": 0.9098,
+      "step": 6083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412613155522528e-05,
+      "loss": 1.0078,
+      "step": 6084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3410574134409404e-05,
+      "loss": 1.1116,
+      "step": 6085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3408534952820249e-05,
+      "loss": 1.0024,
+      "step": 6086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340649561085101e-05,
+      "loss": 1.0409,
+      "step": 6087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404456108597642e-05,
+      "loss": 0.9818,
+      "step": 6088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402416446156112e-05,
+      "loss": 0.9846,
+      "step": 6089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340037662362239e-05,
+      "loss": 0.9795,
+      "step": 6090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3398336641092455e-05,
+      "loss": 0.9772,
+      "step": 6091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3396296498662294e-05,
+      "loss": 1.0112,
+      "step": 6092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394256196427897e-05,
+      "loss": 0.9752,
+      "step": 6093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392215734485272e-05,
+      "loss": 0.9917,
+      "step": 6094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3390175112930421e-05,
+      "loss": 0.8816,
+      "step": 6095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388134331859363e-05,
+      "loss": 0.9814,
+      "step": 6096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386093391368122e-05,
+      "loss": 1.007,
+      "step": 6097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3384052291552732e-05,
+      "loss": 0.9303,
+      "step": 6098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3382011032509226e-05,
+      "loss": 1.0086,
+      "step": 6099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3379969614333656e-05,
+      "loss": 0.957,
+      "step": 6100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3377928037122072e-05,
+      "loss": 1.0014,
+      "step": 6101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375886300970537e-05,
+      "loss": 0.9647,
+      "step": 6102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3373844405975119e-05,
+      "loss": 0.9632,
+      "step": 6103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371802352231895e-05,
+      "loss": 0.9787,
+      "step": 6104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3369760139836951e-05,
+      "loss": 0.9951,
+      "step": 6105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3367717768886374e-05,
+      "loss": 0.9109,
+      "step": 6106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3365675239476263e-05,
+      "loss": 1.0123,
+      "step": 6107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3363632551702726e-05,
+      "loss": 0.9987,
+      "step": 6108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361589705661877e-05,
+      "loss": 0.9574,
+      "step": 6109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3359546701449838e-05,
+      "loss": 1.0045,
+      "step": 6110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3357503539162731e-05,
+      "loss": 1.0304,
+      "step": 6111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355460218896698e-05,
+      "loss": 1.0877,
+      "step": 6112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335341674074788e-05,
+      "loss": 0.9864,
+      "step": 6113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3351373104812433e-05,
+      "loss": 0.9756,
+      "step": 6114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349329311186505e-05,
+      "loss": 0.9933,
+      "step": 6115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334728535996627e-05,
+      "loss": 1.0317,
+      "step": 6116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3345241251247898e-05,
+      "loss": 0.9395,
+      "step": 6117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343196985127568e-05,
+      "loss": 1.0106,
+      "step": 6118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341152561701472e-05,
+      "loss": 0.9905,
+      "step": 6119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3339107981065803e-05,
+      "loss": 1.0086,
+      "step": 6120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337063243316762e-05,
+      "loss": 1.074,
+      "step": 6121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3335018348550559e-05,
+      "loss": 1.0525,
+      "step": 6122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3332973296863413e-05,
+      "loss": 1.0256,
+      "step": 6123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330928088351552e-05,
+      "loss": 1.0827,
+      "step": 6124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328882723111203e-05,
+      "loss": 1.0319,
+      "step": 6125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326837201238605e-05,
+      "loss": 0.9377,
+      "step": 6126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324791522830008e-05,
+      "loss": 1.0206,
+      "step": 6127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3322745687981668e-05,
+      "loss": 0.9717,
+      "step": 6128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320699696789843e-05,
+      "loss": 1.0384,
+      "step": 6129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33186535493508e-05,
+      "loss": 1.092,
+      "step": 6130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3316607245760822e-05,
+      "loss": 1.0149,
+      "step": 6131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3314560786116186e-05,
+      "loss": 0.9934,
+      "step": 6132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331251417051319e-05,
+      "loss": 1.0718,
+      "step": 6133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3310467399048127e-05,
+      "loss": 0.9506,
+      "step": 6134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3308420471817302e-05,
+      "loss": 1.0326,
+      "step": 6135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330637338891703e-05,
+      "loss": 1.0262,
+      "step": 6136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3304326150443636e-05,
+      "loss": 0.9419,
+      "step": 6137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3302278756493441e-05,
+      "loss": 1.0809,
+      "step": 6138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300231207162783e-05,
+      "loss": 1.0617,
+      "step": 6139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298183502548003e-05,
+      "loss": 1.0814,
+      "step": 6140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296135642745452e-05,
+      "loss": 1.0749,
+      "step": 6141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3294087627851486e-05,
+      "loss": 0.9817,
+      "step": 6142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292039457962469e-05,
+      "loss": 0.9382,
+      "step": 6143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289991133174774e-05,
+      "loss": 1.0183,
+      "step": 6144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287942653584778e-05,
+      "loss": 1.0512,
+      "step": 6145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3285894019288863e-05,
+      "loss": 0.9514,
+      "step": 6146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3283845230383437e-05,
+      "loss": 1.0142,
+      "step": 6147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281796286964884e-05,
+      "loss": 0.9861,
+      "step": 6148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3279747189129619e-05,
+      "loss": 0.9975,
+      "step": 6149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3277697936974058e-05,
+      "loss": 0.9944,
+      "step": 6150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327564853059462e-05,
+      "loss": 0.9613,
+      "step": 6151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327359897008774e-05,
+      "loss": 0.9781,
+      "step": 6152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3271549255549854e-05,
+      "loss": 0.9997,
+      "step": 6153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32694993870774e-05,
+      "loss": 1.0101,
+      "step": 6154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3267449364766833e-05,
+      "loss": 0.991,
+      "step": 6155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265399188714616e-05,
+      "loss": 1.0224,
+      "step": 6156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3263348859017208e-05,
+      "loss": 1.0491,
+      "step": 6157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261298375771087e-05,
+      "loss": 0.9614,
+      "step": 6158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259247739072732e-05,
+      "loss": 1.0186,
+      "step": 6159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325719694901863e-05,
+      "loss": 0.9744,
+      "step": 6160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255146005705272e-05,
+      "loss": 0.9423,
+      "step": 6161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325309490922917e-05,
+      "loss": 1.0113,
+      "step": 6162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3251043659686822e-05,
+      "loss": 1.0676,
+      "step": 6163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3248992257174754e-05,
+      "loss": 1.0277,
+      "step": 6164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246940701789482e-05,
+      "loss": 1.0039,
+      "step": 6165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244888993627543e-05,
+      "loss": 0.9846,
+      "step": 6166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242837132785468e-05,
+      "loss": 0.9548,
+      "step": 6167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324078511935981e-05,
+      "loss": 1.0076,
+      "step": 6168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3238732953447113e-05,
+      "loss": 1.0026,
+      "step": 6169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236680635143946e-05,
+      "loss": 1.0062,
+      "step": 6170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323462816454687e-05,
+      "loss": 1.003,
+      "step": 6171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232575541752457e-05,
+      "loss": 0.9729,
+      "step": 6172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230522766857293e-05,
+      "loss": 1.0233,
+      "step": 6173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3228469839957963e-05,
+      "loss": 0.9832,
+      "step": 6174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226416761151062e-05,
+      "loss": 1.1119,
+      "step": 6175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224363530533197e-05,
+      "loss": 0.9678,
+      "step": 6176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222310148200972e-05,
+      "loss": 1.003,
+      "step": 6177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3220256614251006e-05,
+      "loss": 1.0081,
+      "step": 6178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218202928779925e-05,
+      "loss": 1.0103,
+      "step": 6179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3216149091884356e-05,
+      "loss": 0.9544,
+      "step": 6180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321409510366094e-05,
+      "loss": 0.9817,
+      "step": 6181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3212040964206321e-05,
+      "loss": 0.9645,
+      "step": 6182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209986673617152e-05,
+      "loss": 1.0025,
+      "step": 6183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320793223199009e-05,
+      "loss": 0.9855,
+      "step": 6184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320587763942181e-05,
+      "loss": 0.9371,
+      "step": 6185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203822896008976e-05,
+      "loss": 0.9588,
+      "step": 6186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201768001848272e-05,
+      "loss": 0.9984,
+      "step": 6187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199712957036384e-05,
+      "loss": 0.9818,
+      "step": 6188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197657761670013e-05,
+      "loss": 0.939,
+      "step": 6189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195602415845858e-05,
+      "loss": 0.9635,
+      "step": 6190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193546919660628e-05,
+      "loss": 0.9802,
+      "step": 6191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3191491273211034e-05,
+      "loss": 0.9799,
+      "step": 6192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3189435476593808e-05,
+      "loss": 1.1095,
+      "step": 6193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187379529905675e-05,
+      "loss": 1.0448,
+      "step": 6194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3185323433243378e-05,
+      "loss": 1.0034,
+      "step": 6195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183267186703652e-05,
+      "loss": 1.0541,
+      "step": 6196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181210790383256e-05,
+      "loss": 1.0506,
+      "step": 6197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179154244378946e-05,
+      "loss": 1.0032,
+      "step": 6198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177097548787491e-05,
+      "loss": 1.0106,
+      "step": 6199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317504070370566e-05,
+      "loss": 1.0345,
+      "step": 6200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172983709230232e-05,
+      "loss": 0.9968,
+      "step": 6201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170926565457993e-05,
+      "loss": 1.0801,
+      "step": 6202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168869272485743e-05,
+      "loss": 1.0445,
+      "step": 6203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3166811830410277e-05,
+      "loss": 1.0644,
+      "step": 6204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164754239328406e-05,
+      "loss": 0.9972,
+      "step": 6205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.316269649933694e-05,
+      "loss": 1.0708,
+      "step": 6206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3160638610532708e-05,
+      "loss": 0.9981,
+      "step": 6207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3158580573012534e-05,
+      "loss": 1.0055,
+      "step": 6208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3156522386873254e-05,
+      "loss": 0.9843,
+      "step": 6209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154464052211711e-05,
+      "loss": 1.0028,
+      "step": 6210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3152405569124756e-05,
+      "loss": 0.9484,
+      "step": 6211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150346937709247e-05,
+      "loss": 1.0089,
+      "step": 6212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3148288158062041e-05,
+      "loss": 1.0033,
+      "step": 6213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146229230280017e-05,
+      "loss": 1.0204,
+      "step": 6214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314417015446005e-05,
+      "loss": 1.033,
+      "step": 6215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3142110930699023e-05,
+      "loss": 0.9842,
+      "step": 6216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3140051559093827e-05,
+      "loss": 1.0474,
+      "step": 6217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3137992039741368e-05,
+      "loss": 0.9797,
+      "step": 6218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313593237273854e-05,
+      "loss": 1.0249,
+      "step": 6219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133872558182264e-05,
+      "loss": 0.9605,
+      "step": 6220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3131812596169454e-05,
+      "loss": 0.9512,
+      "step": 6221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3129752486797041e-05,
+      "loss": 1.0486,
+      "step": 6222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3127692230161955e-05,
+      "loss": 1.0044,
+      "step": 6223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125631826361139e-05,
+      "loss": 1.0496,
+      "step": 6224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3123571275491535e-05,
+      "loss": 0.9224,
+      "step": 6225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3121510577650106e-05,
+      "loss": 0.9862,
+      "step": 6226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119449732933802e-05,
+      "loss": 0.9802,
+      "step": 6227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3117388741439602e-05,
+      "loss": 1.0239,
+      "step": 6228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3115327603264472e-05,
+      "loss": 1.0405,
+      "step": 6229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113266318505397e-05,
+      "loss": 0.9889,
+      "step": 6230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111204887259367e-05,
+      "loss": 1.1063,
+      "step": 6231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3109143309623375e-05,
+      "loss": 1.0551,
+      "step": 6232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107081585694428e-05,
+      "loss": 1.004,
+      "step": 6233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105019715569528e-05,
+      "loss": 0.9553,
+      "step": 6234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102957699345696e-05,
+      "loss": 0.9722,
+      "step": 6235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3100895537119953e-05,
+      "loss": 1.0124,
+      "step": 6236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3098833228989332e-05,
+      "loss": 1.0288,
+      "step": 6237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096770775050867e-05,
+      "loss": 1.0127,
+      "step": 6238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094708175401601e-05,
+      "loss": 0.9993,
+      "step": 6239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3092645430138586e-05,
+      "loss": 1.1182,
+      "step": 6240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090582539358878e-05,
+      "loss": 1.0285,
+      "step": 6241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088519503159545e-05,
+      "loss": 1.047,
+      "step": 6242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3086456321637653e-05,
+      "loss": 1.0278,
+      "step": 6243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3084392994890283e-05,
+      "loss": 1.0074,
+      "step": 6244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082329523014516e-05,
+      "loss": 1.0392,
+      "step": 6245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080265906107448e-05,
+      "loss": 0.9708,
+      "step": 6246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078202144266177e-05,
+      "loss": 1.076,
+      "step": 6247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076138237587802e-05,
+      "loss": 0.9856,
+      "step": 6248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3074074186169446e-05,
+      "loss": 1.0231,
+      "step": 6249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3072009990108216e-05,
+      "loss": 0.9999,
+      "step": 6250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3069945649501245e-05,
+      "loss": 0.9251,
+      "step": 6251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067881164445664e-05,
+      "loss": 1.0306,
+      "step": 6252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3065816535038612e-05,
+      "loss": 1.013,
+      "step": 6253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3063751761377233e-05,
+      "loss": 1.0462,
+      "step": 6254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3061686843558681e-05,
+      "loss": 0.9775,
+      "step": 6255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3059621781680117e-05,
+      "loss": 0.9503,
+      "step": 6256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3057556575838706e-05,
+      "loss": 0.969,
+      "step": 6257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055491226131618e-05,
+      "loss": 0.9811,
+      "step": 6258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305342573265604e-05,
+      "loss": 1.006,
+      "step": 6259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305136009550915e-05,
+      "loss": 1.117,
+      "step": 6260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049294314788154e-05,
+      "loss": 0.9767,
+      "step": 6261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3047228390590237e-05,
+      "loss": 0.9901,
+      "step": 6262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045162323012617e-05,
+      "loss": 1.0181,
+      "step": 6263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30430961121525e-05,
+      "loss": 0.9704,
+      "step": 6264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3041029758107114e-05,
+      "loss": 0.9222,
+      "step": 6265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303896326097368e-05,
+      "loss": 1.0167,
+      "step": 6266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3036896620849437e-05,
+      "loss": 0.9827,
+      "step": 6267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303482983783162e-05,
+      "loss": 0.9231,
+      "step": 6268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032762912017479e-05,
+      "loss": 0.9817,
+      "step": 6269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030695843504269e-05,
+      "loss": 0.9609,
+      "step": 6270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3028628632389251e-05,
+      "loss": 0.9641,
+      "step": 6271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026561278769686e-05,
+      "loss": 1.0602,
+      "step": 6272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024493782742858e-05,
+      "loss": 0.9786,
+      "step": 6273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3022426144406045e-05,
+      "loss": 1.0102,
+      "step": 6274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302035836385653e-05,
+      "loss": 1.0117,
+      "step": 6275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018290441191608e-05,
+      "loss": 1.0611,
+      "step": 6276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3016222376508582e-05,
+      "loss": 1.0011,
+      "step": 6277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3014154169904762e-05,
+      "loss": 1.0641,
+      "step": 6278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012085821477456e-05,
+      "loss": 0.9904,
+      "step": 6279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3010017331323997e-05,
+      "loss": 0.918,
+      "step": 6280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3007948699541695e-05,
+      "loss": 0.998,
+      "step": 6281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30058799262279e-05,
+      "loss": 1.012,
+      "step": 6282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003811011479943e-05,
+      "loss": 0.9691,
+      "step": 6283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3001741955395177e-05,
+      "loss": 0.9685,
+      "step": 6284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2999672758070954e-05,
+      "loss": 1.135,
+      "step": 6285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997603419604636e-05,
+      "loss": 1.0578,
+      "step": 6286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995533940093587e-05,
+      "loss": 1.0385,
+      "step": 6287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2993464319635188e-05,
+      "loss": 0.9577,
+      "step": 6288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2991394558326813e-05,
+      "loss": 1.031,
+      "step": 6289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989324656265856e-05,
+      "loss": 1.0564,
+      "step": 6290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987254613549704e-05,
+      "loss": 1.0291,
+      "step": 6291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985184430275761e-05,
+      "loss": 0.9624,
+      "step": 6292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983114106541437e-05,
+      "loss": 0.961,
+      "step": 6293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981043642444142e-05,
+      "loss": 1.0217,
+      "step": 6294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978973038081297e-05,
+      "loss": 1.0133,
+      "step": 6295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976902293550332e-05,
+      "loss": 0.9609,
+      "step": 6296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2974831408948675e-05,
+      "loss": 1.064,
+      "step": 6297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972760384373774e-05,
+      "loss": 1.0623,
+      "step": 6298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297068921992307e-05,
+      "loss": 0.9945,
+      "step": 6299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296861791569402e-05,
+      "loss": 0.8614,
+      "step": 6300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296654647178408e-05,
+      "loss": 0.9617,
+      "step": 6301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964474888290717e-05,
+      "loss": 1.0455,
+      "step": 6302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296240316531141e-05,
+      "loss": 1.0631,
+      "step": 6303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2960331302943634e-05,
+      "loss": 0.9874,
+      "step": 6304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2958259301284876e-05,
+      "loss": 0.9738,
+      "step": 6305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956187160432627e-05,
+      "loss": 0.969,
+      "step": 6306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954114880484393e-05,
+      "loss": 1.0521,
+      "step": 6307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2952042461537669e-05,
+      "loss": 1.1225,
+      "step": 6308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2949969903689978e-05,
+      "loss": 1.0397,
+      "step": 6309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947897207038834e-05,
+      "loss": 0.9289,
+      "step": 6310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945824371681763e-05,
+      "loss": 0.956,
+      "step": 6311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2943751397716296e-05,
+      "loss": 1.0211,
+      "step": 6312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941678285239978e-05,
+      "loss": 1.009,
+      "step": 6313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2939605034350345e-05,
+      "loss": 1.0266,
+      "step": 6314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937531645144954e-05,
+      "loss": 0.9601,
+      "step": 6315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2935458117721361e-05,
+      "loss": 1.0284,
+      "step": 6316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2933384452177131e-05,
+      "loss": 0.9354,
+      "step": 6317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931310648609838e-05,
+      "loss": 0.9126,
+      "step": 6318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929236707117057e-05,
+      "loss": 0.9823,
+      "step": 6319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292716262779637e-05,
+      "loss": 1.0072,
+      "step": 6320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925088410745375e-05,
+      "loss": 1.0118,
+      "step": 6321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923014056061658e-05,
+      "loss": 1.0305,
+      "step": 6322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2920939563842833e-05,
+      "loss": 1.0422,
+      "step": 6323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2918864934186506e-05,
+      "loss": 0.9693,
+      "step": 6324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916790167190291e-05,
+      "loss": 1.014,
+      "step": 6325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914715262951813e-05,
+      "loss": 1.0041,
+      "step": 6326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912640221568706e-05,
+      "loss": 0.9176,
+      "step": 6327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29105650431386e-05,
+      "loss": 0.9819,
+      "step": 6328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908489727759136e-05,
+      "loss": 1.0403,
+      "step": 6329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290641427552797e-05,
+      "loss": 0.9934,
+      "step": 6330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290433868654275e-05,
+      "loss": 1.023,
+      "step": 6331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290226296090114e-05,
+      "loss": 0.8726,
+      "step": 6332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900187098700812e-05,
+      "loss": 0.9768,
+      "step": 6333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898111100039434e-05,
+      "loss": 1.0085,
+      "step": 6334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289603496501469e-05,
+      "loss": 1.0283,
+      "step": 6335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893958693724269e-05,
+      "loss": 1.0267,
+      "step": 6336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891882286265862e-05,
+      "loss": 1.0669,
+      "step": 6337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288980574273717e-05,
+      "loss": 1.0604,
+      "step": 6338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887729063235902e-05,
+      "loss": 0.8775,
+      "step": 6339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885652247859768e-05,
+      "loss": 0.9717,
+      "step": 6340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2883575296706485e-05,
+      "loss": 0.992,
+      "step": 6341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881498209873788e-05,
+      "loss": 1.0047,
+      "step": 6342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.28794209874594e-05,
+      "loss": 0.9799,
+      "step": 6343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287734362956106e-05,
+      "loss": 0.98,
+      "step": 6344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287526613627652e-05,
+      "loss": 0.9715,
+      "step": 6345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2873188507703522e-05,
+      "loss": 1.1099,
+      "step": 6346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2871110743939832e-05,
+      "loss": 0.9328,
+      "step": 6347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2869032845083209e-05,
+      "loss": 1.0089,
+      "step": 6348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866954811231424e-05,
+      "loss": 1.054,
+      "step": 6349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2864876642482257e-05,
+      "loss": 1.0219,
+      "step": 6350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862798338933486e-05,
+      "loss": 0.9703,
+      "step": 6351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2860719900682904e-05,
+      "loss": 0.9934,
+      "step": 6352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858641327828301e-05,
+      "loss": 1.0113,
+      "step": 6353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2856562620467488e-05,
+      "loss": 0.9585,
+      "step": 6354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854483778698265e-05,
+      "loss": 1.029,
+      "step": 6355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852404802618456e-05,
+      "loss": 0.9228,
+      "step": 6356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285032569232587e-05,
+      "loss": 0.9844,
+      "step": 6357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848246447918347e-05,
+      "loss": 0.9735,
+      "step": 6358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2846167069493714e-05,
+      "loss": 1.0309,
+      "step": 6359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2844087557149806e-05,
+      "loss": 1.0739,
+      "step": 6360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284200791098448e-05,
+      "loss": 1.0034,
+      "step": 6361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839928131095581e-05,
+      "loss": 1.0019,
+      "step": 6362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283784821758097e-05,
+      "loss": 0.9814,
+      "step": 6363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835768170538513e-05,
+      "loss": 0.981,
+      "step": 6364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283368799006608e-05,
+      "loss": 0.962,
+      "step": 6365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283160767626155e-05,
+      "loss": 0.9468,
+      "step": 6366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829527229222806e-05,
+      "loss": 1.0051,
+      "step": 6367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827446649047737e-05,
+      "loss": 0.9324,
+      "step": 6368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825365935834243e-05,
+      "loss": 0.8948,
+      "step": 6369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823285089680227e-05,
+      "loss": 0.9857,
+      "step": 6370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821204110683592e-05,
+      "loss": 1.0457,
+      "step": 6371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819122998942258e-05,
+      "loss": 0.9593,
+      "step": 6372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817041754554144e-05,
+      "loss": 0.8816,
+      "step": 6373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281496037761718e-05,
+      "loss": 0.9735,
+      "step": 6374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812878868229302e-05,
+      "loss": 0.9112,
+      "step": 6375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2810797226488443e-05,
+      "loss": 1.0467,
+      "step": 6376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808715452492557e-05,
+      "loss": 1.0659,
+      "step": 6377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2806633546339592e-05,
+      "loss": 1.0316,
+      "step": 6378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280455150812751e-05,
+      "loss": 1.0838,
+      "step": 6379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802469337954272e-05,
+      "loss": 1.0215,
+      "step": 6380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800387035917856e-05,
+      "loss": 0.9387,
+      "step": 6381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798304602116232e-05,
+      "loss": 1.0278,
+      "step": 6382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796222036647388e-05,
+      "loss": 0.9151,
+      "step": 6383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2794139339609312e-05,
+      "loss": 0.9659,
+      "step": 6384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2792056511100006e-05,
+      "loss": 1.0761,
+      "step": 6385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789973551217462e-05,
+      "loss": 0.9751,
+      "step": 6386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787890460059695e-05,
+      "loss": 0.9946,
+      "step": 6387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2785807237724721e-05,
+      "loss": 1.0414,
+      "step": 6388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783723884310554e-05,
+      "loss": 0.9827,
+      "step": 6389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2781640399915228e-05,
+      "loss": 0.9053,
+      "step": 6390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2779556784636773e-05,
+      "loss": 0.9965,
+      "step": 6391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2777473038573229e-05,
+      "loss": 0.9727,
+      "step": 6392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277538916182264e-05,
+      "loss": 0.9963,
+      "step": 6393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277330515448306e-05,
+      "loss": 1.0008,
+      "step": 6394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2771221016652546e-05,
+      "loss": 1.0123,
+      "step": 6395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769136748429162e-05,
+      "loss": 0.9305,
+      "step": 6396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767052349910976e-05,
+      "loss": 0.9742,
+      "step": 6397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2764967821196064e-05,
+      "loss": 0.9613,
+      "step": 6398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762883162382514e-05,
+      "loss": 1.0878,
+      "step": 6399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760798373568405e-05,
+      "loss": 1.0249,
+      "step": 6400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2758713454851842e-05,
+      "loss": 1.061,
+      "step": 6401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756628406330918e-05,
+      "loss": 0.9631,
+      "step": 6402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754543228103739e-05,
+      "loss": 1.1012,
+      "step": 6403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752457920268428e-05,
+      "loss": 0.9812,
+      "step": 6404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750372482923093e-05,
+      "loss": 0.9835,
+      "step": 6405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748286916165863e-05,
+      "loss": 1.0767,
+      "step": 6406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274620122009487e-05,
+      "loss": 0.932,
+      "step": 6407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2744115394808252e-05,
+      "loss": 1.0558,
+      "step": 6408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742029440404146e-05,
+      "loss": 1.054,
+      "step": 6409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739943356980708e-05,
+      "loss": 1.0002,
+      "step": 6410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737857144636093e-05,
+      "loss": 1.0071,
+      "step": 6411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273577080346846e-05,
+      "loss": 0.9493,
+      "step": 6412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733684333575977e-05,
+      "loss": 0.977,
+      "step": 6413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273159773505682e-05,
+      "loss": 0.9156,
+      "step": 6414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2729511008009163e-05,
+      "loss": 0.9616,
+      "step": 6415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27274241525312e-05,
+      "loss": 0.9467,
+      "step": 6416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2725337168721117e-05,
+      "loss": 0.9903,
+      "step": 6417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723250056677113e-05,
+      "loss": 1.0522,
+      "step": 6418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721162816497388e-05,
+      "loss": 1.0428,
+      "step": 6419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2719075448280161e-05,
+      "loss": 1.0475,
+      "step": 6420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271698795212364e-05,
+      "loss": 1.0434,
+      "step": 6421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271490032812605e-05,
+      "loss": 1.0061,
+      "step": 6422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2712812576385618e-05,
+      "loss": 1.0148,
+      "step": 6423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271072469700058e-05,
+      "loss": 0.9637,
+      "step": 6424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2708636690069174e-05,
+      "loss": 1.012,
+      "step": 6425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2706548555689647e-05,
+      "loss": 1.0228,
+      "step": 6426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270446029396025e-05,
+      "loss": 1.0784,
+      "step": 6427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2702371904979238e-05,
+      "loss": 1.0332,
+      "step": 6428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270028338884488e-05,
+      "loss": 1.077,
+      "step": 6429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698194745655443e-05,
+      "loss": 1.0158,
+      "step": 6430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2696105975509208e-05,
+      "loss": 1.029,
+      "step": 6431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269401707850445e-05,
+      "loss": 1.0322,
+      "step": 6432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269192805473946e-05,
+      "loss": 0.9536,
+      "step": 6433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268983890431253e-05,
+      "loss": 1.0338,
+      "step": 6434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2687749627321963e-05,
+      "loss": 0.9972,
+      "step": 6435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685660223866058e-05,
+      "loss": 0.9038,
+      "step": 6436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2683570694043138e-05,
+      "loss": 1.0061,
+      "step": 6437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681481037951511e-05,
+      "loss": 0.9738,
+      "step": 6438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2679391255689505e-05,
+      "loss": 0.9286,
+      "step": 6439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677301347355444e-05,
+      "loss": 1.0127,
+      "step": 6440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675211313047674e-05,
+      "loss": 0.9634,
+      "step": 6441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673121152864525e-05,
+      "loss": 1.089,
+      "step": 6442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267103086690435e-05,
+      "loss": 0.9654,
+      "step": 6443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668940455265502e-05,
+      "loss": 0.9967,
+      "step": 6444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2666849918046337e-05,
+      "loss": 1.0047,
+      "step": 6445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664759255345226e-05,
+      "loss": 0.9668,
+      "step": 6446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662668467260535e-05,
+      "loss": 0.9067,
+      "step": 6447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660577553890639e-05,
+      "loss": 0.9995,
+      "step": 6448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658486515333928e-05,
+      "loss": 1.071,
+      "step": 6449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2656395351688784e-05,
+      "loss": 0.9297,
+      "step": 6450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654304063053605e-05,
+      "loss": 1.0384,
+      "step": 6451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652212649526787e-05,
+      "loss": 0.9308,
+      "step": 6452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265012111120674e-05,
+      "loss": 1.0869,
+      "step": 6453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2648029448191881e-05,
+      "loss": 0.9541,
+      "step": 6454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2645937660580616e-05,
+      "loss": 1.0043,
+      "step": 6455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643845748471381e-05,
+      "loss": 1.0476,
+      "step": 6456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2641753711962595e-05,
+      "loss": 1.0471,
+      "step": 6457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639661551152699e-05,
+      "loss": 0.985,
+      "step": 6458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637569266140135e-05,
+      "loss": 1.0181,
+      "step": 6459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635476857023349e-05,
+      "loss": 1.002,
+      "step": 6460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633384323900794e-05,
+      "loss": 1.0448,
+      "step": 6461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2631291666870933e-05,
+      "loss": 1.0407,
+      "step": 6462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2629198886032222e-05,
+      "loss": 1.0353,
+      "step": 6463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627105981483136e-05,
+      "loss": 1.046,
+      "step": 6464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625012953322155e-05,
+      "loss": 0.9873,
+      "step": 6465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2622919801647757e-05,
+      "loss": 0.978,
+      "step": 6466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2620826526558429e-05,
+      "loss": 0.999,
+      "step": 6467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261873312815267e-05,
+      "loss": 0.9943,
+      "step": 6468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616639606528972e-05,
+      "loss": 0.88,
+      "step": 6469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2614545961785851e-05,
+      "loss": 0.9928,
+      "step": 6470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2612452194021808e-05,
+      "loss": 1.0051,
+      "step": 6471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610358303335364e-05,
+      "loss": 1.063,
+      "step": 6472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608264289825041e-05,
+      "loss": 1.0126,
+      "step": 6473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606170153589372e-05,
+      "loss": 0.9586,
+      "step": 6474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604075894726886e-05,
+      "loss": 1.0359,
+      "step": 6475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2601981513336126e-05,
+      "loss": 1.0266,
+      "step": 6476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2599887009515635e-05,
+      "loss": 0.9915,
+      "step": 6477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2597792383363967e-05,
+      "loss": 1.042,
+      "step": 6478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259569763497968e-05,
+      "loss": 0.9423,
+      "step": 6479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593602764461334e-05,
+      "loss": 1.0669,
+      "step": 6480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.25915077719075e-05,
+      "loss": 0.9969,
+      "step": 6481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2589412657416753e-05,
+      "loss": 0.9917,
+      "step": 6482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2587317421087675e-05,
+      "loss": 0.9651,
+      "step": 6483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585222063018848e-05,
+      "loss": 0.9805,
+      "step": 6484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583126583308867e-05,
+      "loss": 0.9618,
+      "step": 6485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2581030982056324e-05,
+      "loss": 0.9303,
+      "step": 6486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257893525935983e-05,
+      "loss": 0.9956,
+      "step": 6487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576839415317989e-05,
+      "loss": 0.9325,
+      "step": 6488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257474345002942e-05,
+      "loss": 0.9654,
+      "step": 6489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572647363592739e-05,
+      "loss": 1.0226,
+      "step": 6490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2570551156106574e-05,
+      "loss": 0.9892,
+      "step": 6491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2568454827669556e-05,
+      "loss": 1.0215,
+      "step": 6492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566358378380326e-05,
+      "loss": 0.9645,
+      "step": 6493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564261808337522e-05,
+      "loss": 1.0556,
+      "step": 6494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2562165117639796e-05,
+      "loss": 0.989,
+      "step": 6495
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560068306385798e-05,
+      "loss": 0.9666,
+      "step": 6496
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557971374674195e-05,
+      "loss": 1.0412,
+      "step": 6497
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555874322603647e-05,
+      "loss": 0.9761,
+      "step": 6498
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553777150272834e-05,
+      "loss": 0.9973,
+      "step": 6499
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551679857780421e-05,
+      "loss": 1.0689,
+      "step": 6500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549582445225101e-05,
+      "loss": 1.1374,
+      "step": 6501
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2547484912705556e-05,
+      "loss": 1.0182,
+      "step": 6502
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254538726032049e-05,
+      "loss": 0.9523,
+      "step": 6503
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2543289488168588e-05,
+      "loss": 0.9948,
+      "step": 6504
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2541191596348568e-05,
+      "loss": 0.9674,
+      "step": 6505
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539093584959131e-05,
+      "loss": 1.0962,
+      "step": 6506
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536995454099003e-05,
+      "loss": 1.0156,
+      "step": 6507
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534897203866903e-05,
+      "loss": 1.0105,
+      "step": 6508
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2532798834361557e-05,
+      "loss": 1.058,
+      "step": 6509
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2530700345681698e-05,
+      "loss": 0.9596,
+      "step": 6510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528601737926068e-05,
+      "loss": 1.0031,
+      "step": 6511
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252650301119341e-05,
+      "loss": 0.9169,
+      "step": 6512
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2524404165582475e-05,
+      "loss": 1.0057,
+      "step": 6513
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252230520119202e-05,
+      "loss": 1.0296,
+      "step": 6514
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2520206118120802e-05,
+      "loss": 0.9237,
+      "step": 6515
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518106916467593e-05,
+      "loss": 0.9881,
+      "step": 6516
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2516007596331167e-05,
+      "loss": 1.0262,
+      "step": 6517
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2513908157810298e-05,
+      "loss": 1.007,
+      "step": 6518
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511808601003767e-05,
+      "loss": 1.0106,
+      "step": 6519
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250970892601037e-05,
+      "loss": 0.9805,
+      "step": 6520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25076091329289e-05,
+      "loss": 0.9999,
+      "step": 6521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2505509221858158e-05,
+      "loss": 0.8962,
+      "step": 6522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2503409192896947e-05,
+      "loss": 1.0139,
+      "step": 6523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2501309046144083e-05,
+      "loss": 0.9892,
+      "step": 6524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2499208781698378e-05,
+      "loss": 0.9113,
+      "step": 6525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249710839965866e-05,
+      "loss": 1.0236,
+      "step": 6526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2495007900123755e-05,
+      "loss": 1.0311,
+      "step": 6527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2492907283192494e-05,
+      "loss": 0.9013,
+      "step": 6528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490806548963719e-05,
+      "loss": 1.0511,
+      "step": 6529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488705697536277e-05,
+      "loss": 0.9312,
+      "step": 6530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248660472900901e-05,
+      "loss": 0.9841,
+      "step": 6531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2484503643480788e-05,
+      "loss": 0.9746,
+      "step": 6532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482402441050459e-05,
+      "loss": 1.0506,
+      "step": 6533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2480301121816896e-05,
+      "loss": 1.0321,
+      "step": 6534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478199685878968e-05,
+      "loss": 0.9655,
+      "step": 6535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247609813333556e-05,
+      "loss": 0.9776,
+      "step": 6536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2473996464285545e-05,
+      "loss": 0.9268,
+      "step": 6537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247189467882782e-05,
+      "loss": 0.9776,
+      "step": 6538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2469792777061275e-05,
+      "loss": 1.0052,
+      "step": 6539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246769075908481e-05,
+      "loss": 1.0769,
+      "step": 6540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465588624997333e-05,
+      "loss": 0.9849,
+      "step": 6541
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463486374897753e-05,
+      "loss": 1.0033,
+      "step": 6542
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461384008884985e-05,
+      "loss": 1.0418,
+      "step": 6543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459281527057951e-05,
+      "loss": 0.98,
+      "step": 6544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245717892951558e-05,
+      "loss": 1.0717,
+      "step": 6545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2455076216356802e-05,
+      "loss": 0.9364,
+      "step": 6546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2452973387680558e-05,
+      "loss": 1.0023,
+      "step": 6547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2450870443585785e-05,
+      "loss": 0.9526,
+      "step": 6548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244876738417144e-05,
+      "loss": 1.0272,
+      "step": 6549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244666420953647e-05,
+      "loss": 1.027,
+      "step": 6550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444560919779842e-05,
+      "loss": 1.0372,
+      "step": 6551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2442457515000515e-05,
+      "loss": 0.9524,
+      "step": 6552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244035399529746e-05,
+      "loss": 0.9159,
+      "step": 6553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438250360769653e-05,
+      "loss": 1.0074,
+      "step": 6554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436146611516082e-05,
+      "loss": 0.9343,
+      "step": 6555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434042747635725e-05,
+      "loss": 1.0508,
+      "step": 6556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2431938769227576e-05,
+      "loss": 1.0076,
+      "step": 6557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2429834676390634e-05,
+      "loss": 0.9293,
+      "step": 6558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24277304692239e-05,
+      "loss": 0.9882,
+      "step": 6559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425626147826387e-05,
+      "loss": 1.0402,
+      "step": 6560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423521712297103e-05,
+      "loss": 0.9535,
+      "step": 6561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242141716273507e-05,
+      "loss": 0.9962,
+      "step": 6562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2419312499239311e-05,
+      "loss": 1.0181,
+      "step": 6563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417207721908855e-05,
+      "loss": 1.0274,
+      "step": 6564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415102830842742e-05,
+      "loss": 0.9651,
+      "step": 6565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2412997826140005e-05,
+      "loss": 1.0277,
+      "step": 6566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410892707899695e-05,
+      "loss": 1.038,
+      "step": 6567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408787476220863e-05,
+      "loss": 1.0177,
+      "step": 6568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240668213120256e-05,
+      "loss": 0.9358,
+      "step": 6569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404576672943854e-05,
+      "loss": 0.9676,
+      "step": 6570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240247110154381e-05,
+      "loss": 0.9146,
+      "step": 6571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400365417101497e-05,
+      "loss": 1.053,
+      "step": 6572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2398259619715995e-05,
+      "loss": 0.9752,
+      "step": 6573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2396153709486392e-05,
+      "loss": 0.9548,
+      "step": 6574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239404768651177e-05,
+      "loss": 0.8911,
+      "step": 6575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391941550891226e-05,
+      "loss": 0.9557,
+      "step": 6576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389835302723856e-05,
+      "loss": 0.9164,
+      "step": 6577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387728942108765e-05,
+      "loss": 0.9708,
+      "step": 6578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385622469145067e-05,
+      "loss": 1.0464,
+      "step": 6579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2383515883931871e-05,
+      "loss": 0.9786,
+      "step": 6580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381409186568298e-05,
+      "loss": 1.0584,
+      "step": 6581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379302377153477e-05,
+      "loss": 0.9348,
+      "step": 6582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2377195455786532e-05,
+      "loss": 1.0204,
+      "step": 6583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375088422566611e-05,
+      "loss": 0.9245,
+      "step": 6584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2372981277592842e-05,
+      "loss": 1.0408,
+      "step": 6585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370874020964376e-05,
+      "loss": 1.0357,
+      "step": 6586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368766652780367e-05,
+      "loss": 1.0138,
+      "step": 6587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366659173139973e-05,
+      "loss": 0.9538,
+      "step": 6588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236455158214235e-05,
+      "loss": 0.9718,
+      "step": 6589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362443879886672e-05,
+      "loss": 0.9013,
+      "step": 6590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360336066472104e-05,
+      "loss": 0.9781,
+      "step": 6591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2358228141997834e-05,
+      "loss": 0.9892,
+      "step": 6592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356120106563036e-05,
+      "loss": 0.9411,
+      "step": 6593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2354011960266904e-05,
+      "loss": 0.993,
+      "step": 6594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351903703208629e-05,
+      "loss": 0.9757,
+      "step": 6595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2349795335487412e-05,
+      "loss": 1.0202,
+      "step": 6596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2347686857202454e-05,
+      "loss": 0.96,
+      "step": 6597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345578268452966e-05,
+      "loss": 0.9989,
+      "step": 6598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234346956933816e-05,
+      "loss": 0.8959,
+      "step": 6599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234136075995726e-05,
+      "loss": 0.9907,
+      "step": 6600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2339251840409487e-05,
+      "loss": 0.9905,
+      "step": 6601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233714281079407e-05,
+      "loss": 0.8948,
+      "step": 6602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335033671210251e-05,
+      "loss": 0.9203,
+      "step": 6603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332924421757262e-05,
+      "loss": 1.0373,
+      "step": 6604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330815062534354e-05,
+      "loss": 0.972,
+      "step": 6605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232870559364077e-05,
+      "loss": 0.9957,
+      "step": 6606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326596015175778e-05,
+      "loss": 0.8631,
+      "step": 6607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232448632723863e-05,
+      "loss": 0.9937,
+      "step": 6608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322376529928594e-05,
+      "loss": 1.0308,
+      "step": 6609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2320266623344942e-05,
+      "loss": 0.9937,
+      "step": 6610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318156607586951e-05,
+      "loss": 0.95,
+      "step": 6611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.23160464827539e-05,
+      "loss": 1.0382,
+      "step": 6612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231393624894508e-05,
+      "loss": 0.9961,
+      "step": 6613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311825906259776e-05,
+      "loss": 1.0729,
+      "step": 6614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2309715454797288e-05,
+      "loss": 1.0143,
+      "step": 6615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307604894656922e-05,
+      "loss": 0.9905,
+      "step": 6616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305494225937985e-05,
+      "loss": 0.9893,
+      "step": 6617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2303383448739778e-05,
+      "loss": 0.9874,
+      "step": 6618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301272563161632e-05,
+      "loss": 0.9951,
+      "step": 6619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299161569302862e-05,
+      "loss": 1.0137,
+      "step": 6620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297050467262799e-05,
+      "loss": 0.9969,
+      "step": 6621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294939257140775e-05,
+      "loss": 1.0116,
+      "step": 6622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292827939036125e-05,
+      "loss": 1.0035,
+      "step": 6623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2290716513048193e-05,
+      "loss": 1.0231,
+      "step": 6624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228860497927633e-05,
+      "loss": 1.0043,
+      "step": 6625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286493337819885e-05,
+      "loss": 1.0204,
+      "step": 6626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284381588778216e-05,
+      "loss": 0.9996,
+      "step": 6627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2282269732250689e-05,
+      "loss": 0.9968,
+      "step": 6628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2280157768336668e-05,
+      "loss": 0.9954,
+      "step": 6629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278045697135534e-05,
+      "loss": 1.0469,
+      "step": 6630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275933518746657e-05,
+      "loss": 0.9412,
+      "step": 6631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273821233269423e-05,
+      "loss": 1.0051,
+      "step": 6632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227170884080322e-05,
+      "loss": 0.9621,
+      "step": 6633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2269596341447446e-05,
+      "loss": 0.9365,
+      "step": 6634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267483735301491e-05,
+      "loss": 0.9799,
+      "step": 6635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2265371022464767e-05,
+      "loss": 0.9921,
+      "step": 6636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263258203036672e-05,
+      "loss": 0.9873,
+      "step": 6637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226114527711663e-05,
+      "loss": 0.9759,
+      "step": 6638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259032244804051e-05,
+      "loss": 0.9581,
+      "step": 6639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256919106198368e-05,
+      "loss": 1.0483,
+      "step": 6640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254805861399002e-05,
+      "loss": 0.9568,
+      "step": 6641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252692510505387e-05,
+      "loss": 0.9588,
+      "step": 6642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2250579053616962e-05,
+      "loss": 0.9606,
+      "step": 6643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248465490833174e-05,
+      "loss": 0.9899,
+      "step": 6644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2246351822253465e-05,
+      "loss": 1.0204,
+      "step": 6645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244238047977297e-05,
+      "loss": 1.0206,
+      "step": 6646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2242124168104119e-05,
+      "loss": 0.9827,
+      "step": 6647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2240010182733401e-05,
+      "loss": 1.0094,
+      "step": 6648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237896091964608e-05,
+      "loss": 0.9605,
+      "step": 6649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235781895897216e-05,
+      "loss": 0.9533,
+      "step": 6650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2233667594630698e-05,
+      "loss": 1.0239,
+      "step": 6651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231553188264543e-05,
+      "loss": 1.0059,
+      "step": 6652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229438676898236e-05,
+      "loss": 1.0121,
+      "step": 6653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227324060631269e-05,
+      "loss": 1.0114,
+      "step": 6654
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225209339563144e-05,
+      "loss": 1.0212,
+      "step": 6655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223094513793364e-05,
+      "loss": 1.0179,
+      "step": 6656
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2220979583421428e-05,
+      "loss": 0.9991,
+      "step": 6657
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221886454854686e-05,
+      "loss": 0.9238,
+      "step": 6658
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216749409269173e-05,
+      "loss": 0.9644,
+      "step": 6659
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214634165687889e-05,
+      "loss": 1.0006,
+      "step": 6660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2212518817902535e-05,
+      "loss": 1.0557,
+      "step": 6661
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2210403366012644e-05,
+      "loss": 0.9714,
+      "step": 6662
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208287810117756e-05,
+      "loss": 0.9663,
+      "step": 6663
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2206172150317408e-05,
+      "loss": 0.9731,
+      "step": 6664
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2204056386711152e-05,
+      "loss": 0.9473,
+      "step": 6665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2201940519398537e-05,
+      "loss": 0.9414,
+      "step": 6666
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219982454847912e-05,
+      "loss": 0.9636,
+      "step": 6667
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197708474052464e-05,
+      "loss": 0.9707,
+      "step": 6668
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2195592296218137e-05,
+      "loss": 0.9627,
+      "step": 6669
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2193476015075705e-05,
+      "loss": 1.033,
+      "step": 6670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219135963072475e-05,
+      "loss": 0.9996,
+      "step": 6671
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189243143264848e-05,
+      "loss": 1.0104,
+      "step": 6672
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187126552795592e-05,
+      "loss": 0.9835,
+      "step": 6673
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185009859416568e-05,
+      "loss": 1.0254,
+      "step": 6674
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182893063227373e-05,
+      "loss": 0.9779,
+      "step": 6675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2180776164327603e-05,
+      "loss": 1.0237,
+      "step": 6676
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2178659162816873e-05,
+      "loss": 0.9848,
+      "step": 6677
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176542058794782e-05,
+      "loss": 0.9456,
+      "step": 6678
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174424852360958e-05,
+      "loss": 1.0334,
+      "step": 6679
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172307543615008e-05,
+      "loss": 1.0451,
+      "step": 6680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170190132656563e-05,
+      "loss": 0.9956,
+      "step": 6681
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2168072619585251e-05,
+      "loss": 0.9409,
+      "step": 6682
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216595500450071e-05,
+      "loss": 1.0722,
+      "step": 6683
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2163837287502576e-05,
+      "loss": 1.0066,
+      "step": 6684
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161719468690492e-05,
+      "loss": 0.9997,
+      "step": 6685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159601548164106e-05,
+      "loss": 1.0604,
+      "step": 6686
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157483526023075e-05,
+      "loss": 1.0184,
+      "step": 6687
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2155365402367053e-05,
+      "loss": 0.9178,
+      "step": 6688
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153247177295708e-05,
+      "loss": 1.0258,
+      "step": 6689
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151128850908705e-05,
+      "loss": 1.0207,
+      "step": 6690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149010423305714e-05,
+      "loss": 0.9135,
+      "step": 6691
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146891894586416e-05,
+      "loss": 0.9756,
+      "step": 6692
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144773264850496e-05,
+      "loss": 0.9615,
+      "step": 6693
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2142654534197633e-05,
+      "loss": 0.9596,
+      "step": 6694
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140535702727521e-05,
+      "loss": 0.9708,
+      "step": 6695
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138416770539861e-05,
+      "loss": 0.9641,
+      "step": 6696
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2136297737734348e-05,
+      "loss": 0.9817,
+      "step": 6697
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134178604410693e-05,
+      "loss": 0.9348,
+      "step": 6698
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132059370668602e-05,
+      "loss": 0.952,
+      "step": 6699
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2129940036607795e-05,
+      "loss": 1.0378,
+      "step": 6700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127820602327986e-05,
+      "loss": 1.0106,
+      "step": 6701
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2125701067928906e-05,
+      "loss": 1.0897,
+      "step": 6702
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2123581433510282e-05,
+      "loss": 1.0047,
+      "step": 6703
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121461699171844e-05,
+      "loss": 0.9612,
+      "step": 6704
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2119341865013334e-05,
+      "loss": 1.0281,
+      "step": 6705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117221931134497e-05,
+      "loss": 1.0083,
+      "step": 6706
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115101897635081e-05,
+      "loss": 1.0473,
+      "step": 6707
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211298176461484e-05,
+      "loss": 0.9636,
+      "step": 6708
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2110861532173524e-05,
+      "loss": 0.8982,
+      "step": 6709
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108741200410903e-05,
+      "loss": 1.0028,
+      "step": 6710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210662076942674e-05,
+      "loss": 1.0464,
+      "step": 6711
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2104500239320813e-05,
+      "loss": 1.0407,
+      "step": 6712
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2102379610192886e-05,
+      "loss": 0.9674,
+      "step": 6713
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100258882142754e-05,
+      "loss": 0.9616,
+      "step": 6714
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209813805527019e-05,
+      "loss": 0.9614,
+      "step": 6715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096017129674995e-05,
+      "loss": 0.8804,
+      "step": 6716
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2093896105456957e-05,
+      "loss": 1.0015,
+      "step": 6717
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2091774982715876e-05,
+      "loss": 0.9588,
+      "step": 6718
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089653761551558e-05,
+      "loss": 1.053,
+      "step": 6719
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087532442063812e-05,
+      "loss": 0.997,
+      "step": 6720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208541102435245e-05,
+      "loss": 1.0061,
+      "step": 6721
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208328950851729e-05,
+      "loss": 0.9591,
+      "step": 6722
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2081167894658158e-05,
+      "loss": 1.0252,
+      "step": 6723
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079046182874872e-05,
+      "loss": 0.9875,
+      "step": 6724
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2076924373267276e-05,
+      "loss": 0.943,
+      "step": 6725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074802465935198e-05,
+      "loss": 1.051,
+      "step": 6726
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2072680460978483e-05,
+      "loss": 1.0352,
+      "step": 6727
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070558358496974e-05,
+      "loss": 1.0921,
+      "step": 6728
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068436158590524e-05,
+      "loss": 0.938,
+      "step": 6729
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066313861358985e-05,
+      "loss": 0.9487,
+      "step": 6730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206419146690222e-05,
+      "loss": 0.9889,
+      "step": 6731
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062068975320087e-05,
+      "loss": 0.9746,
+      "step": 6732
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2059946386712462e-05,
+      "loss": 1.0589,
+      "step": 6733
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057823701179211e-05,
+      "loss": 0.9318,
+      "step": 6734
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055700918820217e-05,
+      "loss": 1.017,
+      "step": 6735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053578039735362e-05,
+      "loss": 0.9583,
+      "step": 6736
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2051455064024529e-05,
+      "loss": 0.9334,
+      "step": 6737
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204933199178761e-05,
+      "loss": 0.949,
+      "step": 6738
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047208823124503e-05,
+      "loss": 1.0021,
+      "step": 6739
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2045085558135106e-05,
+      "loss": 0.9753,
+      "step": 6740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204296219691933e-05,
+      "loss": 0.945,
+      "step": 6741
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2040838739577076e-05,
+      "loss": 1.0335,
+      "step": 6742
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2038715186208264e-05,
+      "loss": 1.0122,
+      "step": 6743
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036591536912807e-05,
+      "loss": 0.9822,
+      "step": 6744
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034467791790635e-05,
+      "loss": 0.9601,
+      "step": 6745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2032343950941671e-05,
+      "loss": 0.9647,
+      "step": 6746
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2030220014465847e-05,
+      "loss": 0.9548,
+      "step": 6747
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2028095982463104e-05,
+      "loss": 1.0453,
+      "step": 6748
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025971855033374e-05,
+      "loss": 0.9864,
+      "step": 6749
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202384763227661e-05,
+      "loss": 1.0357,
+      "step": 6750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2021723314292761e-05,
+      "loss": 0.9783,
+      "step": 6751
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019598901181779e-05,
+      "loss": 0.9628,
+      "step": 6752
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2017474393043625e-05,
+      "loss": 0.9254,
+      "step": 6753
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201534978997826e-05,
+      "loss": 0.9207,
+      "step": 6754
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2013225092085657e-05,
+      "loss": 0.927,
+      "step": 6755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2011100299465784e-05,
+      "loss": 1.0412,
+      "step": 6756
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008975412218616e-05,
+      "loss": 0.9643,
+      "step": 6757
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200685043044414e-05,
+      "loss": 1.0094,
+      "step": 6758
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004725354242335e-05,
+      "loss": 0.9332,
+      "step": 6759
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20026001837132e-05,
+      "loss": 1.0466,
+      "step": 6760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200047491895672e-05,
+      "loss": 0.9444,
+      "step": 6761
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998349560072902e-05,
+      "loss": 1.0049,
+      "step": 6762
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996224107161743e-05,
+      "loss": 0.9531,
+      "step": 6763
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1994098560323257e-05,
+      "loss": 1.0036,
+      "step": 6764
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991972919657449e-05,
+      "loss": 0.9523,
+      "step": 6765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989847185264342e-05,
+      "loss": 0.9808,
+      "step": 6766
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987721357243953e-05,
+      "loss": 1.0017,
+      "step": 6767
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985595435696312e-05,
+      "loss": 1.0333,
+      "step": 6768
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983469420721446e-05,
+      "loss": 1.0308,
+      "step": 6769
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198134331241939e-05,
+      "loss": 0.9432,
+      "step": 6770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979217110890177e-05,
+      "loss": 0.9567,
+      "step": 6771
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197709081623386e-05,
+      "loss": 0.9596,
+      "step": 6772
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197496442855048e-05,
+      "loss": 1.0692,
+      "step": 6773
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972837947940095e-05,
+      "loss": 1.0558,
+      "step": 6774
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970711374502752e-05,
+      "loss": 1.007,
+      "step": 6775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196858470833852e-05,
+      "loss": 0.9741,
+      "step": 6776
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966457949547458e-05,
+      "loss": 1.0259,
+      "step": 6777
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196433109822964e-05,
+      "loss": 0.9979,
+      "step": 6778
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962204154485135e-05,
+      "loss": 1.0346,
+      "step": 6779
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1960077118414027e-05,
+      "loss": 0.9791,
+      "step": 6780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957949990116394e-05,
+      "loss": 0.9775,
+      "step": 6781
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.195582276969232e-05,
+      "loss": 1.0081,
+      "step": 6782
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953695457241904e-05,
+      "loss": 0.9557,
+      "step": 6783
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1951568052865232e-05,
+      "loss": 0.9546,
+      "step": 6784
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949440556662413e-05,
+      "loss": 1.0655,
+      "step": 6785
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947312968733544e-05,
+      "loss": 1.0523,
+      "step": 6786
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945185289178738e-05,
+      "loss": 1.0135,
+      "step": 6787
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1943057518098105e-05,
+      "loss": 0.9351,
+      "step": 6788
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940929655591762e-05,
+      "loss": 1.0855,
+      "step": 6789
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.193880170175983e-05,
+      "loss": 0.9766,
+      "step": 6790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1936673656702438e-05,
+      "loss": 1.0123,
+      "step": 6791
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934545520519709e-05,
+      "loss": 0.9785,
+      "step": 6792
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932417293311786e-05,
+      "loss": 0.9838,
+      "step": 6793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930288975178799e-05,
+      "loss": 0.9756,
+      "step": 6794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928160566220896e-05,
+      "loss": 0.9693,
+      "step": 6795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926032066538222e-05,
+      "loss": 0.9429,
+      "step": 6796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192390347623093e-05,
+      "loss": 0.9796,
+      "step": 6797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921774795399173e-05,
+      "loss": 1.0052,
+      "step": 6798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919646024143112e-05,
+      "loss": 1.0447,
+      "step": 6799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917517162562908e-05,
+      "loss": 1.022,
+      "step": 6800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915388210758736e-05,
+      "loss": 0.9384,
+      "step": 6801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913259168830764e-05,
+      "loss": 1.0091,
+      "step": 6802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191113003687917e-05,
+      "loss": 0.9384,
+      "step": 6803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190900081500413e-05,
+      "loss": 0.9322,
+      "step": 6804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906871503305839e-05,
+      "loss": 1.0663,
+      "step": 6805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1904742101884478e-05,
+      "loss": 0.9565,
+      "step": 6806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902612610840246e-05,
+      "loss": 1.0159,
+      "step": 6807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1900483030273333e-05,
+      "loss": 1.0243,
+      "step": 6808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1898353360283952e-05,
+      "loss": 0.9707,
+      "step": 6809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1896223600972301e-05,
+      "loss": 1.0391,
+      "step": 6810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894093752438594e-05,
+      "loss": 0.9311,
+      "step": 6811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891963814783045e-05,
+      "loss": 0.9292,
+      "step": 6812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889833788105873e-05,
+      "loss": 1.0032,
+      "step": 6813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1887703672507302e-05,
+      "loss": 0.9682,
+      "step": 6814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1885573468087557e-05,
+      "loss": 0.9529,
+      "step": 6815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188344317494687e-05,
+      "loss": 0.9569,
+      "step": 6816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188131279318548e-05,
+      "loss": 1.0062,
+      "step": 6817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879182322903622e-05,
+      "loss": 1.0261,
+      "step": 6818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877051764201542e-05,
+      "loss": 0.9739,
+      "step": 6819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187492111717949e-05,
+      "loss": 1.0011,
+      "step": 6820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872790381937715e-05,
+      "loss": 0.9511,
+      "step": 6821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1870659558576474e-05,
+      "loss": 0.8868,
+      "step": 6822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186852864719603e-05,
+      "loss": 1.0459,
+      "step": 6823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1866397647896644e-05,
+      "loss": 1.0654,
+      "step": 6824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1864266560778587e-05,
+      "loss": 1.0813,
+      "step": 6825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862135385942138e-05,
+      "loss": 1.0269,
+      "step": 6826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186000412348756e-05,
+      "loss": 0.9928,
+      "step": 6827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857872773515147e-05,
+      "loss": 0.9471,
+      "step": 6828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855741336125175e-05,
+      "loss": 0.9214,
+      "step": 6829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1853609811417942e-05,
+      "loss": 1.0487,
+      "step": 6830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851478199493735e-05,
+      "loss": 0.9741,
+      "step": 6831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1849346500452858e-05,
+      "loss": 0.9571,
+      "step": 6832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1847214714395605e-05,
+      "loss": 0.9666,
+      "step": 6833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1845082841422288e-05,
+      "loss": 0.9546,
+      "step": 6834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1842950881633215e-05,
+      "loss": 0.9516,
+      "step": 6835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18408188351287e-05,
+      "loss": 1.0652,
+      "step": 6836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838686702009059e-05,
+      "loss": 0.971,
+      "step": 6837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836554482374618e-05,
+      "loss": 0.9031,
+      "step": 6838
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834422176325702e-05,
+      "loss": 0.9839,
+      "step": 6839
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183228978396264e-05,
+      "loss": 0.9724,
+      "step": 6840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830157305385765e-05,
+      "loss": 0.9286,
+      "step": 6841
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828024740695418e-05,
+      "loss": 1.0571,
+      "step": 6842
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825892089991944e-05,
+      "loss": 0.9782,
+      "step": 6843
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1823759353375684e-05,
+      "loss": 0.9378,
+      "step": 6844
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1821626530946995e-05,
+      "loss": 0.9403,
+      "step": 6845
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819493622806221e-05,
+      "loss": 1.0488,
+      "step": 6846
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1817360629053733e-05,
+      "loss": 0.9291,
+      "step": 6847
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1815227549789884e-05,
+      "loss": 0.9775,
+      "step": 6848
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181309438511505e-05,
+      "loss": 1.0221,
+      "step": 6849
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810961135129593e-05,
+      "loss": 0.9564,
+      "step": 6850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808827799933894e-05,
+      "loss": 1.0475,
+      "step": 6851
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806694379628326e-05,
+      "loss": 0.8611,
+      "step": 6852
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804560874313276e-05,
+      "loss": 1.0337,
+      "step": 6853
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180242728408913e-05,
+      "loss": 1.0121,
+      "step": 6854
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180029360905628e-05,
+      "loss": 1.0003,
+      "step": 6855
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1798159849315114e-05,
+      "loss": 0.9755,
+      "step": 6856
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796026004966039e-05,
+      "loss": 0.973,
+      "step": 6857
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793892076109455e-05,
+      "loss": 0.9865,
+      "step": 6858
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791758062845768e-05,
+      "loss": 0.9696,
+      "step": 6859
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1789623965275387e-05,
+      "loss": 1.0842,
+      "step": 6860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178748978349873e-05,
+      "loss": 1.0609,
+      "step": 6861
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785355517616212e-05,
+      "loss": 1.0224,
+      "step": 6862
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783221167728263e-05,
+      "loss": 0.9626,
+      "step": 6863
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781086733935299e-05,
+      "loss": 1.0316,
+      "step": 6864
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1778952216337758e-05,
+      "loss": 0.9426,
+      "step": 6865
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177681761503607e-05,
+      "loss": 0.9495,
+      "step": 6866
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1774682930130676e-05,
+      "loss": 0.9893,
+      "step": 6867
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177254816172202e-05,
+      "loss": 0.9614,
+      "step": 6868
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770413309910542e-05,
+      "loss": 0.9967,
+      "step": 6869
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1768278374796696e-05,
+      "loss": 0.9645,
+      "step": 6870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766143356480936e-05,
+      "loss": 0.9825,
+      "step": 6871
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764008255063724e-05,
+      "loss": 1.0169,
+      "step": 6872
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176187307064551e-05,
+      "loss": 1.0008,
+      "step": 6873
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759737803326775e-05,
+      "loss": 0.9498,
+      "step": 6874
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1757602453207976e-05,
+      "loss": 1.0079,
+      "step": 6875
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1755467020389595e-05,
+      "loss": 0.9252,
+      "step": 6876
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1753331504972101e-05,
+      "loss": 1.0451,
+      "step": 6877
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751195907055987e-05,
+      "loss": 1.0096,
+      "step": 6878
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749060226741726e-05,
+      "loss": 0.9659,
+      "step": 6879
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746924464129816e-05,
+      "loss": 1.0134,
+      "step": 6880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744788619320744e-05,
+      "loss": 0.9679,
+      "step": 6881
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742652692415012e-05,
+      "loss": 0.9493,
+      "step": 6882
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1740516683513117e-05,
+      "loss": 0.9592,
+      "step": 6883
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738380592715567e-05,
+      "loss": 0.992,
+      "step": 6884
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736244420122861e-05,
+      "loss": 0.9816,
+      "step": 6885
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1734108165835524e-05,
+      "loss": 1.0209,
+      "step": 6886
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1731971829954066e-05,
+      "loss": 1.0538,
+      "step": 6887
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172983541257901e-05,
+      "loss": 1.0417,
+      "step": 6888
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1727698913810872e-05,
+      "loss": 0.9627,
+      "step": 6889
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172556233375019e-05,
+      "loss": 1.0881,
+      "step": 6890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723425672497484e-05,
+      "loss": 1.0028,
+      "step": 6891
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721288930153301e-05,
+      "loss": 0.9687,
+      "step": 6892
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719152106818173e-05,
+      "loss": 0.9202,
+      "step": 6893
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1717015202592643e-05,
+      "loss": 0.9662,
+      "step": 6894
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714878217577258e-05,
+      "loss": 0.9612,
+      "step": 6895
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1712741151872573e-05,
+      "loss": 0.9569,
+      "step": 6896
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1710604005579139e-05,
+      "loss": 1.0162,
+      "step": 6897
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708466778797511e-05,
+      "loss": 0.9831,
+      "step": 6898
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1706329471628254e-05,
+      "loss": 1.099,
+      "step": 6899
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704192084171934e-05,
+      "loss": 1.0332,
+      "step": 6900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702054616529121e-05,
+      "loss": 0.958,
+      "step": 6901
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1699917068800388e-05,
+      "loss": 0.9534,
+      "step": 6902
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169777944108631e-05,
+      "loss": 1.0129,
+      "step": 6903
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695641733487467e-05,
+      "loss": 0.9925,
+      "step": 6904
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693503946104446e-05,
+      "loss": 0.9663,
+      "step": 6905
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691366079037833e-05,
+      "loss": 0.9145,
+      "step": 6906
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689228132388225e-05,
+      "loss": 0.8947,
+      "step": 6907
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687090106256212e-05,
+      "loss": 1.0074,
+      "step": 6908
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684952000742395e-05,
+      "loss": 1.0879,
+      "step": 6909
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682813815947378e-05,
+      "loss": 1.0086,
+      "step": 6910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680675551971771e-05,
+      "loss": 0.9902,
+      "step": 6911
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678537208916178e-05,
+      "loss": 1.0236,
+      "step": 6912
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1676398786881218e-05,
+      "loss": 0.9785,
+      "step": 6913
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1674260285967506e-05,
+      "loss": 0.9562,
+      "step": 6914
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167212170627567e-05,
+      "loss": 0.8948,
+      "step": 6915
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1669983047906329e-05,
+      "loss": 0.9772,
+      "step": 6916
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667844310960112e-05,
+      "loss": 1.0083,
+      "step": 6917
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1665705495537655e-05,
+      "loss": 1.0684,
+      "step": 6918
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663566601739597e-05,
+      "loss": 0.9776,
+      "step": 6919
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661427629666573e-05,
+      "loss": 1.0238,
+      "step": 6920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1659288579419233e-05,
+      "loss": 1.0116,
+      "step": 6921
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657149451098216e-05,
+      "loss": 1.0017,
+      "step": 6922
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655010244804179e-05,
+      "loss": 0.9899,
+      "step": 6923
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652870960637776e-05,
+      "loss": 1.0012,
+      "step": 6924
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650731598699667e-05,
+      "loss": 0.9877,
+      "step": 6925
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648592159090515e-05,
+      "loss": 1.0301,
+      "step": 6926
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1646452641910983e-05,
+      "loss": 0.9046,
+      "step": 6927
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644313047261742e-05,
+      "loss": 0.9965,
+      "step": 6928
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1642173375243464e-05,
+      "loss": 0.9614,
+      "step": 6929
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1640033625956829e-05,
+      "loss": 1.0048,
+      "step": 6930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1637893799502513e-05,
+      "loss": 1.0092,
+      "step": 6931
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635753895981204e-05,
+      "loss": 0.8927,
+      "step": 6932
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1633613915493586e-05,
+      "loss": 0.9932,
+      "step": 6933
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1631473858140356e-05,
+      "loss": 0.9783,
+      "step": 6934
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629333724022207e-05,
+      "loss": 1.0301,
+      "step": 6935
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627193513239836e-05,
+      "loss": 1.0049,
+      "step": 6936
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1625053225893942e-05,
+      "loss": 1.0033,
+      "step": 6937
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1622912862085239e-05,
+      "loss": 0.9417,
+      "step": 6938
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162077242191443e-05,
+      "loss": 1.0435,
+      "step": 6939
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618631905482235e-05,
+      "loss": 0.9973,
+      "step": 6940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161649131288936e-05,
+      "loss": 0.9264,
+      "step": 6941
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614350644236532e-05,
+      "loss": 0.864,
+      "step": 6942
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1612209899624474e-05,
+      "loss": 0.9884,
+      "step": 6943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610069079153915e-05,
+      "loss": 0.9933,
+      "step": 6944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607928182925585e-05,
+      "loss": 1.0888,
+      "step": 6945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605787211040218e-05,
+      "loss": 1.0236,
+      "step": 6946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160364616359855e-05,
+      "loss": 1.0428,
+      "step": 6947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1601505040701327e-05,
+      "loss": 1.0405,
+      "step": 6948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159936384244929e-05,
+      "loss": 0.9894,
+      "step": 6949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159722256894319e-05,
+      "loss": 0.9663,
+      "step": 6950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595081220283774e-05,
+      "loss": 0.9877,
+      "step": 6951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592939796571809e-05,
+      "loss": 0.9561,
+      "step": 6952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1590798297908044e-05,
+      "loss": 0.8892,
+      "step": 6953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158865672439325e-05,
+      "loss": 0.9578,
+      "step": 6954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586515076128185e-05,
+      "loss": 0.9574,
+      "step": 6955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584373353213625e-05,
+      "loss": 0.9761,
+      "step": 6956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158223155575034e-05,
+      "loss": 1.0017,
+      "step": 6957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1580089683839108e-05,
+      "loss": 1.0565,
+      "step": 6958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1577947737580712e-05,
+      "loss": 1.0636,
+      "step": 6959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1575805717075934e-05,
+      "loss": 0.9101,
+      "step": 6960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1573663622425559e-05,
+      "loss": 1.1488,
+      "step": 6961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571521453730381e-05,
+      "loss": 1.0122,
+      "step": 6962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569379211091194e-05,
+      "loss": 0.9971,
+      "step": 6963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1567236894608792e-05,
+      "loss": 0.9367,
+      "step": 6964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565094504383985e-05,
+      "loss": 1.0752,
+      "step": 6965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562952040517568e-05,
+      "loss": 0.9352,
+      "step": 6966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560809503110355e-05,
+      "loss": 1.006,
+      "step": 6967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1558666892263157e-05,
+      "loss": 0.9382,
+      "step": 6968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556524208076789e-05,
+      "loss": 0.9806,
+      "step": 6969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1554381450652066e-05,
+      "loss": 0.9578,
+      "step": 6970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552238620089815e-05,
+      "loss": 0.9552,
+      "step": 6971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155009571649086e-05,
+      "loss": 1.0038,
+      "step": 6972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547952739956032e-05,
+      "loss": 0.9917,
+      "step": 6973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1545809690586159e-05,
+      "loss": 0.9433,
+      "step": 6974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154366656848208e-05,
+      "loss": 1.0467,
+      "step": 6975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541523373744632e-05,
+      "loss": 0.9114,
+      "step": 6976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153938010647466e-05,
+      "loss": 1.0054,
+      "step": 6977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153723676677301e-05,
+      "loss": 0.954,
+      "step": 6978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153509335474053e-05,
+      "loss": 0.9134,
+      "step": 6979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532949870478069e-05,
+      "loss": 1.0709,
+      "step": 6980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530806314086493e-05,
+      "loss": 0.9171,
+      "step": 6981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528662685666657e-05,
+      "loss": 1.0002,
+      "step": 6982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1526518985319422e-05,
+      "loss": 1.0047,
+      "step": 6983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524375213145654e-05,
+      "loss": 1.0569,
+      "step": 6984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1522231369246227e-05,
+      "loss": 0.9779,
+      "step": 6985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152008745372201e-05,
+      "loss": 1.0457,
+      "step": 6986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1517943466673884e-05,
+      "loss": 1.023,
+      "step": 6987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1515799408202724e-05,
+      "loss": 1.026,
+      "step": 6988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513655278409416e-05,
+      "loss": 1.0127,
+      "step": 6989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151151107739485e-05,
+      "loss": 1.0546,
+      "step": 6990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509366805259907e-05,
+      "loss": 0.981,
+      "step": 6991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150722246210549e-05,
+      "loss": 1.0448,
+      "step": 6992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150507804803249e-05,
+      "loss": 1.0359,
+      "step": 6993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150293356314181e-05,
+      "loss": 0.9751,
+      "step": 6994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150078900753435e-05,
+      "loss": 0.9447,
+      "step": 6995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149864438131102e-05,
+      "loss": 1.0298,
+      "step": 6996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1496499684572729e-05,
+      "loss": 0.8489,
+      "step": 6997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149435491742039e-05,
+      "loss": 0.9932,
+      "step": 6998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492210079954915e-05,
+      "loss": 0.9812,
+      "step": 6999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1490065172277235e-05,
+      "loss": 0.9813,
+      "step": 7000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487920194488262e-05,
+      "loss": 0.9791,
+      "step": 7001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1485775146688931e-05,
+      "loss": 0.975,
+      "step": 7002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1483630028980167e-05,
+      "loss": 0.9461,
+      "step": 7003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481484841462904e-05,
+      "loss": 0.9971,
+      "step": 7004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479339584238079e-05,
+      "loss": 1.0248,
+      "step": 7005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1477194257406634e-05,
+      "loss": 1.0522,
+      "step": 7006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475048861069507e-05,
+      "loss": 1.0278,
+      "step": 7007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1472903395327645e-05,
+      "loss": 0.9311,
+      "step": 7008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470757860282002e-05,
+      "loss": 0.9958,
+      "step": 7009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1468612256033528e-05,
+      "loss": 0.994,
+      "step": 7010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146646658268318e-05,
+      "loss": 1.011,
+      "step": 7011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1464320840331915e-05,
+      "loss": 0.9596,
+      "step": 7012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462175029080696e-05,
+      "loss": 1.0023,
+      "step": 7013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460029149030492e-05,
+      "loss": 0.9875,
+      "step": 7014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145788320028227e-05,
+      "loss": 0.9683,
+      "step": 7015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455737182937001e-05,
+      "loss": 0.9756,
+      "step": 7016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145359109709566e-05,
+      "loss": 0.9802,
+      "step": 7017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1451444942859232e-05,
+      "loss": 1.0278,
+      "step": 7018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144929872032869e-05,
+      "loss": 0.9845,
+      "step": 7019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447152429605022e-05,
+      "loss": 0.998,
+      "step": 7020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144500607078922e-05,
+      "loss": 0.9379,
+      "step": 7021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442859643982273e-05,
+      "loss": 0.9399,
+      "step": 7022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1440713149285179e-05,
+      "loss": 1.0459,
+      "step": 7023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1438566586798926e-05,
+      "loss": 1.001,
+      "step": 7024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143641995662453e-05,
+      "loss": 1.014,
+      "step": 7025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143427325886298e-05,
+      "loss": 1.0327,
+      "step": 7026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432126493615294e-05,
+      "loss": 1.0086,
+      "step": 7027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429979660982477e-05,
+      "loss": 0.977,
+      "step": 7028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142783276106555e-05,
+      "loss": 1.0285,
+      "step": 7029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1425685793965522e-05,
+      "loss": 1.01,
+      "step": 7030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423538759783416e-05,
+      "loss": 1.0326,
+      "step": 7031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421391658620257e-05,
+      "loss": 1.019,
+      "step": 7032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1419244490577071e-05,
+      "loss": 1.0016,
+      "step": 7033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1417097255754883e-05,
+      "loss": 1.0333,
+      "step": 7034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1414949954254735e-05,
+      "loss": 1.029,
+      "step": 7035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412802586177654e-05,
+      "loss": 0.9988,
+      "step": 7036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410655151624684e-05,
+      "loss": 0.9943,
+      "step": 7037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1408507650696863e-05,
+      "loss": 0.9256,
+      "step": 7038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1406360083495243e-05,
+      "loss": 0.9921,
+      "step": 7039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404212450120867e-05,
+      "loss": 0.9613,
+      "step": 7040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402064750674788e-05,
+      "loss": 1.0092,
+      "step": 7041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399916985258058e-05,
+      "loss": 0.9812,
+      "step": 7042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397769153971741e-05,
+      "loss": 0.9942,
+      "step": 7043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395621256916891e-05,
+      "loss": 0.9881,
+      "step": 7044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393473294194576e-05,
+      "loss": 1.018,
+      "step": 7045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1391325265905862e-05,
+      "loss": 1.0079,
+      "step": 7046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138917717215182e-05,
+      "loss": 0.9294,
+      "step": 7047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387029013033522e-05,
+      "loss": 0.9673,
+      "step": 7048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1384880788652044e-05,
+      "loss": 0.9653,
+      "step": 7049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382732499108467e-05,
+      "loss": 0.9889,
+      "step": 7050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138058414450387e-05,
+      "loss": 1.0326,
+      "step": 7051
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1378435724939343e-05,
+      "loss": 0.9541,
+      "step": 7052
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1376287240515969e-05,
+      "loss": 0.9512,
+      "step": 7053
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1374138691334848e-05,
+      "loss": 0.9656,
+      "step": 7054
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1371990077497064e-05,
+      "loss": 0.9795,
+      "step": 7055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1369841399103723e-05,
+      "loss": 0.9515,
+      "step": 7056
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136769265625592e-05,
+      "loss": 0.9419,
+      "step": 7057
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365543849054765e-05,
+      "loss": 0.9982,
+      "step": 7058
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1363394977601359e-05,
+      "loss": 0.977,
+      "step": 7059
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361246041996817e-05,
+      "loss": 1.0107,
+      "step": 7060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359097042342245e-05,
+      "loss": 0.9716,
+      "step": 7061
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1356947978738764e-05,
+      "loss": 1.0125,
+      "step": 7062
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135479885128749e-05,
+      "loss": 1.0172,
+      "step": 7063
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352649660089548e-05,
+      "loss": 0.9657,
+      "step": 7064
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350500405246059e-05,
+      "loss": 1.0029,
+      "step": 7065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1348351086858154e-05,
+      "loss": 1.0089,
+      "step": 7066
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134620170502696e-05,
+      "loss": 0.9622,
+      "step": 7067
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134405225985362e-05,
+      "loss": 1.0444,
+      "step": 7068
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1341902751439258e-05,
+      "loss": 1.0619,
+      "step": 7069
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339753179885023e-05,
+      "loss": 0.9362,
+      "step": 7070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133760354529205e-05,
+      "loss": 0.9858,
+      "step": 7071
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1335453847761492e-05,
+      "loss": 1.0057,
+      "step": 7072
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333304087394496e-05,
+      "loss": 0.9489,
+      "step": 7073
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331154264292215e-05,
+      "loss": 1.0457,
+      "step": 7074
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1329004378555794e-05,
+      "loss": 0.946,
+      "step": 7075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326854430286402e-05,
+      "loss": 1.0466,
+      "step": 7076
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324704419585195e-05,
+      "loss": 0.9726,
+      "step": 7077
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1322554346553335e-05,
+      "loss": 0.9826,
+      "step": 7078
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1320404211291993e-05,
+      "loss": 0.9887,
+      "step": 7079
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131825401390233e-05,
+      "loss": 0.9785,
+      "step": 7080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316103754485526e-05,
+      "loss": 0.9914,
+      "step": 7081
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313953433142753e-05,
+      "loss": 0.9755,
+      "step": 7082
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131180304997519e-05,
+      "loss": 0.9624,
+      "step": 7083
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1309652605084012e-05,
+      "loss": 1.0211,
+      "step": 7084
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130750209857041e-05,
+      "loss": 0.9318,
+      "step": 7085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130535153053557e-05,
+      "loss": 0.9787,
+      "step": 7086
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303200901080682e-05,
+      "loss": 0.98,
+      "step": 7087
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1301050210306931e-05,
+      "loss": 0.9874,
+      "step": 7088
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1298899458315525e-05,
+      "loss": 0.9903,
+      "step": 7089
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296748645207648e-05,
+      "loss": 0.9852,
+      "step": 7090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294597771084513e-05,
+      "loss": 0.9933,
+      "step": 7091
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129244683604732e-05,
+      "loss": 0.9592,
+      "step": 7092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290295840197272e-05,
+      "loss": 0.9284,
+      "step": 7093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288144783635582e-05,
+      "loss": 0.9236,
+      "step": 7094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285993666463464e-05,
+      "loss": 1.0404,
+      "step": 7095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128384248878213e-05,
+      "loss": 0.9161,
+      "step": 7096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281691250692804e-05,
+      "loss": 1.0088,
+      "step": 7097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12795399522967e-05,
+      "loss": 1.0169,
+      "step": 7098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277388593695046e-05,
+      "loss": 0.9023,
+      "step": 7099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275237174989066e-05,
+      "loss": 0.9385,
+      "step": 7100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127308569628e-05,
+      "loss": 0.9308,
+      "step": 7101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1270934157669065e-05,
+      "loss": 0.9438,
+      "step": 7102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268782559257505e-05,
+      "loss": 0.9626,
+      "step": 7103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1266630901146554e-05,
+      "loss": 0.9876,
+      "step": 7104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1264479183437458e-05,
+      "loss": 1.0014,
+      "step": 7105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126232740623146e-05,
+      "loss": 0.965,
+      "step": 7106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260175569629804e-05,
+      "loss": 0.9783,
+      "step": 7107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1258023673733737e-05,
+      "loss": 0.9446,
+      "step": 7108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1255871718644514e-05,
+      "loss": 0.9723,
+      "step": 7109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253719704463393e-05,
+      "loss": 0.9711,
+      "step": 7110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251567631291627e-05,
+      "loss": 0.9773,
+      "step": 7111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249415499230477e-05,
+      "loss": 1.0257,
+      "step": 7112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1247263308381207e-05,
+      "loss": 0.9304,
+      "step": 7113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245111058845083e-05,
+      "loss": 0.982,
+      "step": 7114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1242958750723374e-05,
+      "loss": 0.9492,
+      "step": 7115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240806384117353e-05,
+      "loss": 0.9569,
+      "step": 7116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1238653959128289e-05,
+      "loss": 1.0223,
+      "step": 7117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236501475857466e-05,
+      "loss": 0.9794,
+      "step": 7118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234348934406156e-05,
+      "loss": 1.0175,
+      "step": 7119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232196334875651e-05,
+      "loss": 1.0095,
+      "step": 7120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230043677367229e-05,
+      "loss": 0.9865,
+      "step": 7121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1227890961982178e-05,
+      "loss": 1.0367,
+      "step": 7122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1225738188821788e-05,
+      "loss": 1.0308,
+      "step": 7123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223585357987358e-05,
+      "loss": 0.9384,
+      "step": 7124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1221432469580182e-05,
+      "loss": 0.9853,
+      "step": 7125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219279523701556e-05,
+      "loss": 0.9765,
+      "step": 7126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1217126520452782e-05,
+      "loss": 1.0211,
+      "step": 7127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214973459935166e-05,
+      "loss": 0.9291,
+      "step": 7128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212820342250012e-05,
+      "loss": 0.9958,
+      "step": 7129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1210667167498637e-05,
+      "loss": 1.0274,
+      "step": 7130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1208513935782344e-05,
+      "loss": 1.0381,
+      "step": 7131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206360647202453e-05,
+      "loss": 1.018,
+      "step": 7132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1204207301860277e-05,
+      "loss": 1.0051,
+      "step": 7133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202053899857144e-05,
+      "loss": 1.0544,
+      "step": 7134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1199900441294371e-05,
+      "loss": 1.0318,
+      "step": 7135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197746926273285e-05,
+      "loss": 0.9922,
+      "step": 7136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195593354895215e-05,
+      "loss": 0.9156,
+      "step": 7137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193439727261491e-05,
+      "loss": 0.947,
+      "step": 7138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191286043473447e-05,
+      "loss": 1.0272,
+      "step": 7139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118913230363242e-05,
+      "loss": 0.9384,
+      "step": 7140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1186978507839748e-05,
+      "loss": 0.9757,
+      "step": 7141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1184824656196769e-05,
+      "loss": 0.9618,
+      "step": 7142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1182670748804834e-05,
+      "loss": 0.9513,
+      "step": 7143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180516785765287e-05,
+      "loss": 0.9612,
+      "step": 7144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1178362767179476e-05,
+      "loss": 0.9545,
+      "step": 7145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1176208693148752e-05,
+      "loss": 0.9664,
+      "step": 7146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117405456377447e-05,
+      "loss": 1.0495,
+      "step": 7147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1171900379157989e-05,
+      "loss": 0.9452,
+      "step": 7148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169746139400673e-05,
+      "loss": 0.8944,
+      "step": 7149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1167591844603874e-05,
+      "loss": 1.0647,
+      "step": 7150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1165437494868965e-05,
+      "loss": 0.9953,
+      "step": 7151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163283090297307e-05,
+      "loss": 0.93,
+      "step": 7152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116112863099028e-05,
+      "loss": 1.0596,
+      "step": 7153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158974117049247e-05,
+      "loss": 0.9799,
+      "step": 7154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156819548575588e-05,
+      "loss": 0.9708,
+      "step": 7155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154664925670678e-05,
+      "loss": 1.0393,
+      "step": 7156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152510248435901e-05,
+      "loss": 0.9588,
+      "step": 7157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115035551697264e-05,
+      "loss": 0.9342,
+      "step": 7158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1148200731382277e-05,
+      "loss": 1.0099,
+      "step": 7159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11460458917662e-05,
+      "loss": 1.03,
+      "step": 7160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143890998225804e-05,
+      "loss": 1.0336,
+      "step": 7161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141736050862476e-05,
+      "loss": 1.0256,
+      "step": 7162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113958104977762e-05,
+      "loss": 0.9969,
+      "step": 7163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137425995072624e-05,
+      "loss": 1.0005,
+      "step": 7164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135270886848896e-05,
+      "loss": 1.0145,
+      "step": 7165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133115725207839e-05,
+      "loss": 0.9205,
+      "step": 7166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130960510250855e-05,
+      "loss": 1.0228,
+      "step": 7167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1128805242079355e-05,
+      "loss": 1.0398,
+      "step": 7168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112664992079475e-05,
+      "loss": 1.0124,
+      "step": 7169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112449454649845e-05,
+      "loss": 0.9665,
+      "step": 7170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122339119291872e-05,
+      "loss": 1.0569,
+      "step": 7171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120183639276439e-05,
+      "loss": 0.9878,
+      "step": 7172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1118028106553564e-05,
+      "loss": 0.9767,
+      "step": 7173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1115872521224678e-05,
+      "loss": 1.0156,
+      "step": 7174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113716883391198e-05,
+      "loss": 0.9673,
+      "step": 7175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111561193154561e-05,
+      "loss": 0.9996,
+      "step": 7176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109405450616191e-05,
+      "loss": 1.0071,
+      "step": 7177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1107249655877525e-05,
+      "loss": 1.0105,
+      "step": 7178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1105093809039995e-05,
+      "loss": 0.9763,
+      "step": 7179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102937910205044e-05,
+      "loss": 0.9881,
+      "step": 7180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1100781959474108e-05,
+      "loss": 1.0545,
+      "step": 7181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098625956948632e-05,
+      "loss": 0.972,
+      "step": 7182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096469902730058e-05,
+      "loss": 1.0074,
+      "step": 7183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109431379691984e-05,
+      "loss": 0.9434,
+      "step": 7184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1092157639619421e-05,
+      "loss": 0.9878,
+      "step": 7185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109000143093026e-05,
+      "loss": 1.0772,
+      "step": 7186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087845170953811e-05,
+      "loss": 0.954,
+      "step": 7187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085688859791525e-05,
+      "loss": 1.0835,
+      "step": 7188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1083532497544869e-05,
+      "loss": 1.0157,
+      "step": 7189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10813760843153e-05,
+      "loss": 0.9579,
+      "step": 7190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107921962020429e-05,
+      "loss": 1.0001,
+      "step": 7191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077063105313299e-05,
+      "loss": 0.9816,
+      "step": 7192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074906539743796e-05,
+      "loss": 0.9673,
+      "step": 7193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1072749923597259e-05,
+      "loss": 0.9374,
+      "step": 7194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1070593256975157e-05,
+      "loss": 1.0014,
+      "step": 7195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1068436539978971e-05,
+      "loss": 0.9618,
+      "step": 7196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106627977271017e-05,
+      "loss": 0.9407,
+      "step": 7197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106412295527025e-05,
+      "loss": 1.0154,
+      "step": 7198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1061966087760684e-05,
+      "loss": 1.0284,
+      "step": 7199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105980917028296e-05,
+      "loss": 1.0155,
+      "step": 7200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057652202938568e-05,
+      "loss": 1.0316,
+      "step": 7201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1055495185829002e-05,
+      "loss": 0.9444,
+      "step": 7202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053338119055748e-05,
+      "loss": 0.9777,
+      "step": 7203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1051181002720304e-05,
+      "loss": 0.9534,
+      "step": 7204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1049023836924168e-05,
+      "loss": 1.064,
+      "step": 7205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104686662176884e-05,
+      "loss": 0.9601,
+      "step": 7206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044709357355827e-05,
+      "loss": 1.0349,
+      "step": 7207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042552043786625e-05,
+      "loss": 1.0381,
+      "step": 7208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1040394681162748e-05,
+      "loss": 0.9755,
+      "step": 7209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1038237269585698e-05,
+      "loss": 0.9955,
+      "step": 7210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036079809156997e-05,
+      "loss": 1.0246,
+      "step": 7211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1033922299978148e-05,
+      "loss": 1.0525,
+      "step": 7212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1031764742150676e-05,
+      "loss": 1.0126,
+      "step": 7213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029607135776094e-05,
+      "loss": 0.9517,
+      "step": 7214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027449480955927e-05,
+      "loss": 0.9128,
+      "step": 7215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1025291777791693e-05,
+      "loss": 0.9948,
+      "step": 7216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1023134026384921e-05,
+      "loss": 1.0167,
+      "step": 7217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1020976226837137e-05,
+      "loss": 1.0196,
+      "step": 7218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018818379249872e-05,
+      "loss": 0.8974,
+      "step": 7219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1016660483724658e-05,
+      "loss": 0.9252,
+      "step": 7220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1014502540363029e-05,
+      "loss": 1.0319,
+      "step": 7221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1012344549266517e-05,
+      "loss": 0.9731,
+      "step": 7222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010186510536671e-05,
+      "loss": 0.9777,
+      "step": 7223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1008028424275026e-05,
+      "loss": 1.0337,
+      "step": 7224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005870290583125e-05,
+      "loss": 0.9982,
+      "step": 7225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003712109562513e-05,
+      "loss": 0.9787,
+      "step": 7226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1001553881314742e-05,
+      "loss": 1.0018,
+      "step": 7227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999395605941356e-05,
+      "loss": 0.9343,
+      "step": 7228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997237283543911e-05,
+      "loss": 0.9202,
+      "step": 7229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0995078914223965e-05,
+      "loss": 0.9631,
+      "step": 7230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992920498083067e-05,
+      "loss": 0.9912,
+      "step": 7231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990762035222782e-05,
+      "loss": 1.0945,
+      "step": 7232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0988603525744663e-05,
+      "loss": 1.015,
+      "step": 7233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0986444969750287e-05,
+      "loss": 1.0205,
+      "step": 7234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0984286367341205e-05,
+      "loss": 0.995,
+      "step": 7235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0982127718618993e-05,
+      "loss": 0.9236,
+      "step": 7236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979969023685218e-05,
+      "loss": 1.0126,
+      "step": 7237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0977810282641455e-05,
+      "loss": 1.0007,
+      "step": 7238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975651495589274e-05,
+      "loss": 1.0303,
+      "step": 7239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973492662630253e-05,
+      "loss": 1.0254,
+      "step": 7240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971333783865971e-05,
+      "loss": 0.9991,
+      "step": 7241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0969174859398011e-05,
+      "loss": 0.9953,
+      "step": 7242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0967015889327948e-05,
+      "loss": 0.9873,
+      "step": 7243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0964856873757377e-05,
+      "loss": 0.9662,
+      "step": 7244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962697812787877e-05,
+      "loss": 0.9068,
+      "step": 7245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960538706521042e-05,
+      "loss": 1.0133,
+      "step": 7246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095837955505846e-05,
+      "loss": 0.9765,
+      "step": 7247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0956220358501728e-05,
+      "loss": 0.9149,
+      "step": 7248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954061116952438e-05,
+      "loss": 1.0123,
+      "step": 7249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951901830512193e-05,
+      "loss": 0.8967,
+      "step": 7250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0949742499282586e-05,
+      "loss": 0.9925,
+      "step": 7251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947583123365223e-05,
+      "loss": 0.945,
+      "step": 7252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0945423702861709e-05,
+      "loss": 0.9523,
+      "step": 7253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0943264237873646e-05,
+      "loss": 0.953,
+      "step": 7254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941104728502643e-05,
+      "loss": 0.9586,
+      "step": 7255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938945174850315e-05,
+      "loss": 0.9222,
+      "step": 7256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0936785577018272e-05,
+      "loss": 0.9776,
+      "step": 7257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934625935108128e-05,
+      "loss": 0.9761,
+      "step": 7258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932466249221496e-05,
+      "loss": 1.0345,
+      "step": 7259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093030651946e-05,
+      "loss": 0.9641,
+      "step": 7260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0928146745925257e-05,
+      "loss": 0.9773,
+      "step": 7261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092598692871889e-05,
+      "loss": 0.9687,
+      "step": 7262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923827067942531e-05,
+      "loss": 0.8659,
+      "step": 7263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921667163697798e-05,
+      "loss": 1.034,
+      "step": 7264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0919507216086323e-05,
+      "loss": 0.8898,
+      "step": 7265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0917347225209733e-05,
+      "loss": 1.0011,
+      "step": 7266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915187191169669e-05,
+      "loss": 0.9891,
+      "step": 7267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913027114067763e-05,
+      "loss": 0.961,
+      "step": 7268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910866994005648e-05,
+      "loss": 1.0628,
+      "step": 7269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0908706831084966e-05,
+      "loss": 0.9663,
+      "step": 7270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906546625407356e-05,
+      "loss": 0.9581,
+      "step": 7271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904386377074467e-05,
+      "loss": 0.92,
+      "step": 7272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0902226086187938e-05,
+      "loss": 0.9688,
+      "step": 7273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900065752849418e-05,
+      "loss": 0.9668,
+      "step": 7274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897905377160557e-05,
+      "loss": 0.9927,
+      "step": 7275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895744959223005e-05,
+      "loss": 1.007,
+      "step": 7276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089358449913842e-05,
+      "loss": 1.0515,
+      "step": 7277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0891423997008443e-05,
+      "loss": 0.8953,
+      "step": 7278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889263452934746e-05,
+      "loss": 0.9778,
+      "step": 7279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0887102867018981e-05,
+      "loss": 0.9852,
+      "step": 7280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884942239362813e-05,
+      "loss": 0.9285,
+      "step": 7281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08827815700679e-05,
+      "loss": 1.008,
+      "step": 7282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088062085923591e-05,
+      "loss": 0.9262,
+      "step": 7283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878460106968508e-05,
+      "loss": 0.9724,
+      "step": 7284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876299313367364e-05,
+      "loss": 0.9458,
+      "step": 7285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0874138478534149e-05,
+      "loss": 1.0493,
+      "step": 7286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871977602570535e-05,
+      "loss": 1.0155,
+      "step": 7287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0869816685578199e-05,
+      "loss": 0.9755,
+      "step": 7288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867655727658813e-05,
+      "loss": 0.925,
+      "step": 7289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865494728914057e-05,
+      "loss": 1.0164,
+      "step": 7290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0863333689445615e-05,
+      "loss": 0.9354,
+      "step": 7291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861172609355166e-05,
+      "loss": 0.9548,
+      "step": 7292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859011488744391e-05,
+      "loss": 1.0315,
+      "step": 7293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0856850327714983e-05,
+      "loss": 0.9313,
+      "step": 7294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854689126368625e-05,
+      "loss": 0.9775,
+      "step": 7295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852527884807014e-05,
+      "loss": 1.0113,
+      "step": 7296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850366603131832e-05,
+      "loss": 0.997,
+      "step": 7297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848205281444778e-05,
+      "loss": 0.9199,
+      "step": 7298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846043919847547e-05,
+      "loss": 0.922,
+      "step": 7299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0843882518441837e-05,
+      "loss": 1.0274,
+      "step": 7300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841721077329345e-05,
+      "loss": 1.0293,
+      "step": 7301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0839559596611778e-05,
+      "loss": 0.9897,
+      "step": 7302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.083739807639083e-05,
+      "loss": 1.0435,
+      "step": 7303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0835236516768216e-05,
+      "loss": 0.9697,
+      "step": 7304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833074917845634e-05,
+      "loss": 0.9557,
+      "step": 7305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830913279724798e-05,
+      "loss": 0.9907,
+      "step": 7306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828751602507415e-05,
+      "loss": 1.0253,
+      "step": 7307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08265898862952e-05,
+      "loss": 1.056,
+      "step": 7308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0824428131189866e-05,
+      "loss": 0.917,
+      "step": 7309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0822266337293133e-05,
+      "loss": 1.0746,
+      "step": 7310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820104504706707e-05,
+      "loss": 0.966,
+      "step": 7311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817942633532323e-05,
+      "loss": 1.0092,
+      "step": 7312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0815780723871691e-05,
+      "loss": 1.0191,
+      "step": 7313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081361877582654e-05,
+      "loss": 0.958,
+      "step": 7314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0811456789498593e-05,
+      "loss": 0.9473,
+      "step": 7315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080929476498958e-05,
+      "loss": 0.9995,
+      "step": 7316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807132702401224e-05,
+      "loss": 1.0411,
+      "step": 7317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804970601835258e-05,
+      "loss": 0.9986,
+      "step": 7318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0802808463393415e-05,
+      "loss": 0.9618,
+      "step": 7319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0800646287177429e-05,
+      "loss": 0.9521,
+      "step": 7320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0798484073289037e-05,
+      "loss": 1.0055,
+      "step": 7321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796321821829971e-05,
+      "loss": 0.9119,
+      "step": 7322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0794159532901978e-05,
+      "loss": 0.9309,
+      "step": 7323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791997206606795e-05,
+      "loss": 0.9359,
+      "step": 7324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789834843046166e-05,
+      "loss": 1.019,
+      "step": 7325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078767244232183e-05,
+      "loss": 1.0679,
+      "step": 7326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785510004535542e-05,
+      "loss": 0.9063,
+      "step": 7327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783347529789045e-05,
+      "loss": 0.9229,
+      "step": 7328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0781185018184093e-05,
+      "loss": 0.9591,
+      "step": 7329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0779022469822431e-05,
+      "loss": 0.9937,
+      "step": 7330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776859884805817e-05,
+      "loss": 0.9503,
+      "step": 7331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0774697263236004e-05,
+      "loss": 1.0376,
+      "step": 7332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0772534605214753e-05,
+      "loss": 0.9917,
+      "step": 7333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770371910843817e-05,
+      "loss": 0.9323,
+      "step": 7334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076820918022496e-05,
+      "loss": 0.968,
+      "step": 7335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076604641345994e-05,
+      "loss": 0.9007,
+      "step": 7336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763883610650525e-05,
+      "loss": 0.9415,
+      "step": 7337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0761720771898476e-05,
+      "loss": 0.9251,
+      "step": 7338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759557897305566e-05,
+      "loss": 0.9645,
+      "step": 7339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0757394986973554e-05,
+      "loss": 0.9211,
+      "step": 7340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755232041004222e-05,
+      "loss": 1.0115,
+      "step": 7341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0753069059499332e-05,
+      "loss": 0.9791,
+      "step": 7342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750906042560665e-05,
+      "loss": 0.9982,
+      "step": 7343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074874299028999e-05,
+      "loss": 0.9126,
+      "step": 7344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074657990278909e-05,
+      "loss": 0.9749,
+      "step": 7345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0744416780159739e-05,
+      "loss": 0.9535,
+      "step": 7346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0742253622503717e-05,
+      "loss": 0.9809,
+      "step": 7347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740090429922814e-05,
+      "loss": 0.9493,
+      "step": 7348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737927202518803e-05,
+      "loss": 0.9859,
+      "step": 7349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735763940393478e-05,
+      "loss": 1.0181,
+      "step": 7350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0733600643648615e-05,
+      "loss": 1.0417,
+      "step": 7351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731437312386015e-05,
+      "loss": 1.0085,
+      "step": 7352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0729273946707463e-05,
+      "loss": 1.0177,
+      "step": 7353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072711054671475e-05,
+      "loss": 0.9866,
+      "step": 7354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724947112509665e-05,
+      "loss": 0.9772,
+      "step": 7355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0722783644194013e-05,
+      "loss": 1.0285,
+      "step": 7356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072062014186958e-05,
+      "loss": 1.0601,
+      "step": 7357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718456605638175e-05,
+      "loss": 1.0307,
+      "step": 7358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716293035601588e-05,
+      "loss": 1.0003,
+      "step": 7359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714129431861626e-05,
+      "loss": 1.0042,
+      "step": 7360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0711965794520089e-05,
+      "loss": 0.9822,
+      "step": 7361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070980212367879e-05,
+      "loss": 0.9008,
+      "step": 7362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707638419439519e-05,
+      "loss": 1.0222,
+      "step": 7363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705474681904098e-05,
+      "loss": 0.93,
+      "step": 7364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070331091117433e-05,
+      "loss": 1.0334,
+      "step": 7365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701147107352026e-05,
+      "loss": 1.0139,
+      "step": 7366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0698983270539e-05,
+      "loss": 0.8838,
+      "step": 7367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696819400837066e-05,
+      "loss": 0.9958,
+      "step": 7368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069465549834804e-05,
+      "loss": 0.9609,
+      "step": 7369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0692491563173735e-05,
+      "loss": 0.9927,
+      "step": 7370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690327595415975e-05,
+      "loss": 0.9606,
+      "step": 7371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688163595176575e-05,
+      "loss": 1.0137,
+      "step": 7372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685999562557358e-05,
+      "loss": 0.9875,
+      "step": 7373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.068383549766015e-05,
+      "loss": 0.9661,
+      "step": 7374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681671400586775e-05,
+      "loss": 1.0166,
+      "step": 7375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679507271439056e-05,
+      "loss": 1.0211,
+      "step": 7376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0677343110318822e-05,
+      "loss": 0.9889,
+      "step": 7377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0675178917327907e-05,
+      "loss": 0.9713,
+      "step": 7378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0673014692568133e-05,
+      "loss": 0.9991,
+      "step": 7379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0670850436141338e-05,
+      "loss": 0.9325,
+      "step": 7380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0668686148149354e-05,
+      "loss": 1.007,
+      "step": 7381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066652182869402e-05,
+      "loss": 0.9303,
+      "step": 7382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0664357477877166e-05,
+      "loss": 0.9885,
+      "step": 7383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0662193095800633e-05,
+      "loss": 0.9628,
+      "step": 7384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0660028682566262e-05,
+      "loss": 1.0258,
+      "step": 7385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657864238275892e-05,
+      "loss": 1.014,
+      "step": 7386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0655699763031368e-05,
+      "loss": 1.0504,
+      "step": 7387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065353525693453e-05,
+      "loss": 1.0191,
+      "step": 7388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651370720087229e-05,
+      "loss": 1.0621,
+      "step": 7389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0649206152591304e-05,
+      "loss": 0.9946,
+      "step": 7390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0647041554548614e-05,
+      "loss": 0.9338,
+      "step": 7391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0644876926060999e-05,
+      "loss": 1.0315,
+      "step": 7392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0642712267230313e-05,
+      "loss": 0.9488,
+      "step": 7393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640547578158411e-05,
+      "loss": 0.98,
+      "step": 7394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0638382858947145e-05,
+      "loss": 1.0172,
+      "step": 7395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0636218109698372e-05,
+      "loss": 0.9875,
+      "step": 7396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634053330513949e-05,
+      "loss": 1.0104,
+      "step": 7397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063188852149573e-05,
+      "loss": 0.9531,
+      "step": 7398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629723682745578e-05,
+      "loss": 1.0345,
+      "step": 7399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627558814365358e-05,
+      "loss": 1.037,
+      "step": 7400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625393916456925e-05,
+      "loss": 0.9826,
+      "step": 7401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0623228989122145e-05,
+      "loss": 1.031,
+      "step": 7402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621064032462887e-05,
+      "loss": 0.9912,
+      "step": 7403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618899046581015e-05,
+      "loss": 1.0769,
+      "step": 7404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0616734031578397e-05,
+      "loss": 0.9852,
+      "step": 7405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06145689875569e-05,
+      "loss": 0.9505,
+      "step": 7406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0612403914618399e-05,
+      "loss": 1.028,
+      "step": 7407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0610238812864764e-05,
+      "loss": 0.9017,
+      "step": 7408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0608073682397866e-05,
+      "loss": 1.0037,
+      "step": 7409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0605908523319586e-05,
+      "loss": 0.9329,
+      "step": 7410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0603743335731796e-05,
+      "loss": 0.9722,
+      "step": 7411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0601578119736376e-05,
+      "loss": 1.0177,
+      "step": 7412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599412875435197e-05,
+      "loss": 0.9843,
+      "step": 7413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059724760293015e-05,
+      "loss": 0.9715,
+      "step": 7414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059508230232311e-05,
+      "loss": 0.9408,
+      "step": 7415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059291697371596e-05,
+      "loss": 0.9889,
+      "step": 7416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590751617210586e-05,
+      "loss": 0.9443,
+      "step": 7417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0588586232908875e-05,
+      "loss": 1.0425,
+      "step": 7418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058642082091271e-05,
+      "loss": 0.9698,
+      "step": 7419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584255381323982e-05,
+      "loss": 0.9286,
+      "step": 7420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0582089914244578e-05,
+      "loss": 0.9339,
+      "step": 7421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579924419776389e-05,
+      "loss": 0.9334,
+      "step": 7422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577758898021306e-05,
+      "loss": 0.977,
+      "step": 7423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057559334908123e-05,
+      "loss": 0.9961,
+      "step": 7424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573427773058043e-05,
+      "loss": 1.0419,
+      "step": 7425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571262170053649e-05,
+      "loss": 1.0248,
+      "step": 7426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569096540169944e-05,
+      "loss": 0.9663,
+      "step": 7427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0566930883508823e-05,
+      "loss": 1.0258,
+      "step": 7428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0564765200172192e-05,
+      "loss": 0.9783,
+      "step": 7429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562599490261945e-05,
+      "loss": 0.913,
+      "step": 7430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560433753879984e-05,
+      "loss": 0.947,
+      "step": 7431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558267991128219e-05,
+      "loss": 0.9501,
+      "step": 7432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0556102202108552e-05,
+      "loss": 1.004,
+      "step": 7433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553936386922884e-05,
+      "loss": 1.0128,
+      "step": 7434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551770545673125e-05,
+      "loss": 0.9479,
+      "step": 7435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0549604678461186e-05,
+      "loss": 0.9322,
+      "step": 7436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0547438785388975e-05,
+      "loss": 0.9585,
+      "step": 7437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05452728665584e-05,
+      "loss": 1.0417,
+      "step": 7438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0543106922071373e-05,
+      "loss": 0.9745,
+      "step": 7439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054094095202981e-05,
+      "loss": 1.0759,
+      "step": 7440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538774956535627e-05,
+      "loss": 0.9212,
+      "step": 7441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053660893569073e-05,
+      "loss": 1.0093,
+      "step": 7442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053444288959705e-05,
+      "loss": 0.9498,
+      "step": 7443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0532276818356493e-05,
+      "loss": 0.955,
+      "step": 7444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530110722070983e-05,
+      "loss": 0.9542,
+      "step": 7445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527944600842437e-05,
+      "loss": 0.9369,
+      "step": 7446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525778454772784e-05,
+      "loss": 1.1151,
+      "step": 7447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523612283963938e-05,
+      "loss": 0.944,
+      "step": 7448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0521446088517828e-05,
+      "loss": 1.0161,
+      "step": 7449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519279868536375e-05,
+      "loss": 0.9727,
+      "step": 7450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517113624121506e-05,
+      "loss": 1.0145,
+      "step": 7451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514947355375152e-05,
+      "loss": 0.9444,
+      "step": 7452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.051278106239924e-05,
+      "loss": 0.998,
+      "step": 7453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510614745295695e-05,
+      "loss": 1.0152,
+      "step": 7454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508448404166454e-05,
+      "loss": 1.0596,
+      "step": 7455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0506282039113441e-05,
+      "loss": 1.0007,
+      "step": 7456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05041156502386e-05,
+      "loss": 0.9143,
+      "step": 7457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501949237643854e-05,
+      "loss": 0.9881,
+      "step": 7458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0499782801431143e-05,
+      "loss": 0.982,
+      "step": 7459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497616341702401e-05,
+      "loss": 0.982,
+      "step": 7460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495449858559572e-05,
+      "loss": 0.8816,
+      "step": 7461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0493283352104586e-05,
+      "loss": 1.0234,
+      "step": 7462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491116822439389e-05,
+      "loss": 0.9535,
+      "step": 7463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488950269665916e-05,
+      "loss": 0.9361,
+      "step": 7464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0486783693886114e-05,
+      "loss": 0.9856,
+      "step": 7465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0484617095201924e-05,
+      "loss": 1.0282,
+      "step": 7466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482450473715287e-05,
+      "loss": 1.0153,
+      "step": 7467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480283829528151e-05,
+      "loss": 0.9809,
+      "step": 7468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0478117162742458e-05,
+      "loss": 0.986,
+      "step": 7469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475950473460163e-05,
+      "loss": 0.9142,
+      "step": 7470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0473783761783204e-05,
+      "loss": 1.079,
+      "step": 7471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0471617027813543e-05,
+      "loss": 1.0503,
+      "step": 7472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046945027165312e-05,
+      "loss": 1.0618,
+      "step": 7473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467283493403889e-05,
+      "loss": 0.9448,
+      "step": 7474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0465116693167798e-05,
+      "loss": 1.0136,
+      "step": 7475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462949871046813e-05,
+      "loss": 1.0083,
+      "step": 7476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460783027142873e-05,
+      "loss": 0.9541,
+      "step": 7477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0458616161557944e-05,
+      "loss": 1.0493,
+      "step": 7478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456449274393978e-05,
+      "loss": 0.9977,
+      "step": 7479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0454282365752935e-05,
+      "loss": 1.0306,
+      "step": 7480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0452115435736771e-05,
+      "loss": 0.9555,
+      "step": 7481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449948484447447e-05,
+      "loss": 0.9593,
+      "step": 7482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044778151198692e-05,
+      "loss": 1.0206,
+      "step": 7483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044561451845716e-05,
+      "loss": 0.922,
+      "step": 7484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443447503960118e-05,
+      "loss": 0.993,
+      "step": 7485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0441280468597769e-05,
+      "loss": 1.0464,
+      "step": 7486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439113412472069e-05,
+      "loss": 1.0103,
+      "step": 7487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0436946335684985e-05,
+      "loss": 1.0746,
+      "step": 7488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0434779238338486e-05,
+      "loss": 0.9423,
+      "step": 7489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0432612120534537e-05,
+      "loss": 0.9843,
+      "step": 7490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043044498237511e-05,
+      "loss": 0.9798,
+      "step": 7491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428277823962169e-05,
+      "loss": 1.0244,
+      "step": 7492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0426110645397687e-05,
+      "loss": 0.9643,
+      "step": 7493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0423943446783633e-05,
+      "loss": 1.0413,
+      "step": 7494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421776228221984e-05,
+      "loss": 0.9318,
+      "step": 7495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0419608989814709e-05,
+      "loss": 1.008,
+      "step": 7496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0417441731663784e-05,
+      "loss": 1.0324,
+      "step": 7497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041527445387118e-05,
+      "loss": 0.877,
+      "step": 7498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413107156538878e-05,
+      "loss": 1.017,
+      "step": 7499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0410939839768853e-05,
+      "loss": 0.974,
+      "step": 7500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0408772503663081e-05,
+      "loss": 0.9655,
+      "step": 7501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0406605148323538e-05,
+      "loss": 1.0037,
+      "step": 7502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404437773852214e-05,
+      "loss": 0.9375,
+      "step": 7503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402270380351077e-05,
+      "loss": 0.9611,
+      "step": 7504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0400102967922119e-05,
+      "loss": 1.0355,
+      "step": 7505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397935536667312e-05,
+      "loss": 0.9733,
+      "step": 7506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395768086688647e-05,
+      "loss": 0.9987,
+      "step": 7507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0393600618088102e-05,
+      "loss": 0.9728,
+      "step": 7508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391433130967667e-05,
+      "loss": 0.9125,
+      "step": 7509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0389265625429328e-05,
+      "loss": 0.9551,
+      "step": 7510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387098101575067e-05,
+      "loss": 0.9206,
+      "step": 7511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384930559506871e-05,
+      "loss": 0.9527,
+      "step": 7512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382762999326735e-05,
+      "loss": 1.0264,
+      "step": 7513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0380595421136641e-05,
+      "loss": 0.9593,
+      "step": 7514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0378427825038582e-05,
+      "loss": 1.001,
+      "step": 7515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376260211134548e-05,
+      "loss": 1.0313,
+      "step": 7516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374092579526533e-05,
+      "loss": 1.0999,
+      "step": 7517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371924930316525e-05,
+      "loss": 0.9386,
+      "step": 7518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0369757263606525e-05,
+      "loss": 0.9062,
+      "step": 7519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0367589579498518e-05,
+      "loss": 0.9459,
+      "step": 7520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365421878094506e-05,
+      "loss": 0.9039,
+      "step": 7521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363254159496479e-05,
+      "loss": 1.0241,
+      "step": 7522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361086423806438e-05,
+      "loss": 0.9423,
+      "step": 7523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358918671126379e-05,
+      "loss": 0.9836,
+      "step": 7524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0356750901558302e-05,
+      "loss": 0.9232,
+      "step": 7525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354583115204202e-05,
+      "loss": 1.0225,
+      "step": 7526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352415312166077e-05,
+      "loss": 1.0237,
+      "step": 7527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0350247492545937e-05,
+      "loss": 0.99,
+      "step": 7528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348079656445775e-05,
+      "loss": 1.015,
+      "step": 7529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0345911803967599e-05,
+      "loss": 0.9262,
+      "step": 7530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343743935213404e-05,
+      "loss": 0.9065,
+      "step": 7531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03415760502852e-05,
+      "loss": 0.9963,
+      "step": 7532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339408149284991e-05,
+      "loss": 0.9813,
+      "step": 7533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033724023231478e-05,
+      "loss": 0.9341,
+      "step": 7534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335072299476571e-05,
+      "loss": 0.9571,
+      "step": 7535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332904350872377e-05,
+      "loss": 1.0096,
+      "step": 7536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0330736386604198e-05,
+      "loss": 1.0309,
+      "step": 7537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328568406774053e-05,
+      "loss": 0.9199,
+      "step": 7538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326400411483939e-05,
+      "loss": 0.9669,
+      "step": 7539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032423240083587e-05,
+      "loss": 0.996,
+      "step": 7540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0322064374931857e-05,
+      "loss": 1.0033,
+      "step": 7541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0319896333873911e-05,
+      "loss": 1.0516,
+      "step": 7542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317728277764046e-05,
+      "loss": 0.9285,
+      "step": 7543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031556020670427e-05,
+      "loss": 0.9785,
+      "step": 7544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313392120796595e-05,
+      "loss": 1.0186,
+      "step": 7545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0311224020143041e-05,
+      "loss": 1.0024,
+      "step": 7546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0309055904845617e-05,
+      "loss": 0.9541,
+      "step": 7547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306887775006347e-05,
+      "loss": 0.9569,
+      "step": 7548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304719630727233e-05,
+      "loss": 1.0008,
+      "step": 7549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0302551472110304e-05,
+      "loss": 1.0153,
+      "step": 7550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0300383299257573e-05,
+      "loss": 1.1132,
+      "step": 7551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0298215112271058e-05,
+      "loss": 0.945,
+      "step": 7552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296046911252771e-05,
+      "loss": 0.9827,
+      "step": 7553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293878696304742e-05,
+      "loss": 0.9574,
+      "step": 7554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291710467528987e-05,
+      "loss": 0.9827,
+      "step": 7555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0289542225027523e-05,
+      "loss": 0.9795,
+      "step": 7556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0287373968902376e-05,
+      "loss": 1.0387,
+      "step": 7557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0285205699255567e-05,
+      "loss": 0.9423,
+      "step": 7558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283037416189116e-05,
+      "loss": 0.9974,
+      "step": 7559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280869119805045e-05,
+      "loss": 0.9734,
+      "step": 7560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278700810205385e-05,
+      "loss": 0.9323,
+      "step": 7561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0276532487492155e-05,
+      "loss": 0.9371,
+      "step": 7562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027436415176738e-05,
+      "loss": 0.9187,
+      "step": 7563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272195803133086e-05,
+      "loss": 0.9543,
+      "step": 7564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0270027441691299e-05,
+      "loss": 0.91,
+      "step": 7565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267859067544045e-05,
+      "loss": 0.9388,
+      "step": 7566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265690680793358e-05,
+      "loss": 1.0158,
+      "step": 7567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0263522281541256e-05,
+      "loss": 1.014,
+      "step": 7568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0261353869889777e-05,
+      "loss": 0.9644,
+      "step": 7569
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025918544594094e-05,
+      "loss": 0.9546,
+      "step": 7570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257017009796787e-05,
+      "loss": 0.8872,
+      "step": 7571
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254848561559334e-05,
+      "loss": 0.9489,
+      "step": 7572
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0252680101330626e-05,
+      "loss": 0.9125,
+      "step": 7573
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0250511629212683e-05,
+      "loss": 0.9766,
+      "step": 7574
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248343145307547e-05,
+      "loss": 0.9609,
+      "step": 7575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0246174649717241e-05,
+      "loss": 0.931,
+      "step": 7576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0244006142543807e-05,
+      "loss": 0.9583,
+      "step": 7577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241837623889272e-05,
+      "loss": 1.0015,
+      "step": 7578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239669093855675e-05,
+      "loss": 0.9405,
+      "step": 7579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237500552545048e-05,
+      "loss": 0.9556,
+      "step": 7580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0235332000059425e-05,
+      "loss": 1.0104,
+      "step": 7581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0233163436500844e-05,
+      "loss": 0.9973,
+      "step": 7582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230994861971342e-05,
+      "loss": 0.8737,
+      "step": 7583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0228826276572956e-05,
+      "loss": 1.0029,
+      "step": 7584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0226657680407718e-05,
+      "loss": 1.0392,
+      "step": 7585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224489073577673e-05,
+      "loss": 0.9197,
+      "step": 7586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0222320456184857e-05,
+      "loss": 1.0503,
+      "step": 7587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220151828331309e-05,
+      "loss": 0.9804,
+      "step": 7588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0217983190119064e-05,
+      "loss": 1.024,
+      "step": 7589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215814541650166e-05,
+      "loss": 0.966,
+      "step": 7590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0213645883026657e-05,
+      "loss": 0.9742,
+      "step": 7591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211477214350575e-05,
+      "loss": 0.9985,
+      "step": 7592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020930853572396e-05,
+      "loss": 1.0531,
+      "step": 7593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0207139847248857e-05,
+      "loss": 1.0039,
+      "step": 7594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0204971149027307e-05,
+      "loss": 0.9703,
+      "step": 7595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202802441161355e-05,
+      "loss": 1.0727,
+      "step": 7596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200633723753037e-05,
+      "loss": 0.9737,
+      "step": 7597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198464996904403e-05,
+      "loss": 0.8977,
+      "step": 7598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196296260717493e-05,
+      "loss": 1.013,
+      "step": 7599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0194127515294359e-05,
+      "loss": 0.9926,
+      "step": 7600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0191958760737036e-05,
+      "loss": 0.9732,
+      "step": 7601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0189789997147577e-05,
+      "loss": 0.9759,
+      "step": 7602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018762122462802e-05,
+      "loss": 0.9311,
+      "step": 7603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185452443280418e-05,
+      "loss": 0.9963,
+      "step": 7604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0183283653206816e-05,
+      "loss": 0.9477,
+      "step": 7605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181114854509261e-05,
+      "loss": 1.0459,
+      "step": 7606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01789460472898e-05,
+      "loss": 1.0139,
+      "step": 7607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176777231650478e-05,
+      "loss": 1.0236,
+      "step": 7608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174608407693349e-05,
+      "loss": 1.0271,
+      "step": 7609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0172439575520457e-05,
+      "loss": 1.0156,
+      "step": 7610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0170270735233849e-05,
+      "loss": 0.9665,
+      "step": 7611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0168101886935581e-05,
+      "loss": 0.9329,
+      "step": 7612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165933030727701e-05,
+      "loss": 1.0018,
+      "step": 7613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163764166712256e-05,
+      "loss": 1.0049,
+      "step": 7614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161595294991295e-05,
+      "loss": 0.9513,
+      "step": 7615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0159426415666876e-05,
+      "loss": 0.9859,
+      "step": 7616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0157257528841042e-05,
+      "loss": 0.961,
+      "step": 7617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0155088634615848e-05,
+      "loss": 0.9614,
+      "step": 7618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015291973309335e-05,
+      "loss": 0.9684,
+      "step": 7619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150750824375596e-05,
+      "loss": 0.8976,
+      "step": 7620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148581908564639e-05,
+      "loss": 1.0059,
+      "step": 7621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014641298576253e-05,
+      "loss": 1.0388,
+      "step": 7622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144244056071327e-05,
+      "loss": 0.9377,
+      "step": 7623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014207511959308e-05,
+      "loss": 1.0196,
+      "step": 7624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139906176429843e-05,
+      "loss": 0.96,
+      "step": 7625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013773722668367e-05,
+      "loss": 0.9067,
+      "step": 7626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013556827045662e-05,
+      "loss": 1.0089,
+      "step": 7627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133399307850742e-05,
+      "loss": 0.9245,
+      "step": 7628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131230338968094e-05,
+      "loss": 1.0737,
+      "step": 7629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0129061363910729e-05,
+      "loss": 1.0121,
+      "step": 7630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0126892382780707e-05,
+      "loss": 1.0487,
+      "step": 7631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0124723395680079e-05,
+      "loss": 0.9188,
+      "step": 7632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122554402710906e-05,
+      "loss": 0.9496,
+      "step": 7633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120385403975242e-05,
+      "loss": 0.9675,
+      "step": 7634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0118216399575143e-05,
+      "loss": 1.0403,
+      "step": 7635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0116047389612665e-05,
+      "loss": 0.9966,
+      "step": 7636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0113878374189872e-05,
+      "loss": 1.0085,
+      "step": 7637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0111709353408813e-05,
+      "loss": 1.0019,
+      "step": 7638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109540327371554e-05,
+      "loss": 0.9957,
+      "step": 7639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107371296180144e-05,
+      "loss": 0.8782,
+      "step": 7640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010520225993665e-05,
+      "loss": 0.9194,
+      "step": 7641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0103033218743128e-05,
+      "loss": 0.933,
+      "step": 7642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0100864172701632e-05,
+      "loss": 0.9422,
+      "step": 7643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0098695121914226e-05,
+      "loss": 0.918,
+      "step": 7644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009652606648297e-05,
+      "loss": 0.9493,
+      "step": 7645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094357006509921e-05,
+      "loss": 0.9684,
+      "step": 7646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092187942097138e-05,
+      "loss": 0.983,
+      "step": 7647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0090018873346682e-05,
+      "loss": 1.038,
+      "step": 7648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0087849800360611e-05,
+      "loss": 0.9778,
+      "step": 7649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0085680723240992e-05,
+      "loss": 1.0155,
+      "step": 7650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083511642089876e-05,
+      "loss": 0.9387,
+      "step": 7651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081342557009335e-05,
+      "loss": 1.0652,
+      "step": 7652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079173468101418e-05,
+      "loss": 1.0147,
+      "step": 7653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0077004375468194e-05,
+      "loss": 1.0107,
+      "step": 7654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007483527921172e-05,
+      "loss": 1.0469,
+      "step": 7655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0072666179434062e-05,
+      "loss": 1.0208,
+      "step": 7656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070497076237278e-05,
+      "loss": 1.0104,
+      "step": 7657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006832796972343e-05,
+      "loss": 0.9345,
+      "step": 7658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066158859994576e-05,
+      "loss": 1.0278,
+      "step": 7659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006398974715279e-05,
+      "loss": 1.0362,
+      "step": 7660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006182063130012e-05,
+      "loss": 1.0405,
+      "step": 7661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059651512538643e-05,
+      "loss": 1.0392,
+      "step": 7662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0057482390970406e-05,
+      "loss": 0.9893,
+      "step": 7663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0055313266697482e-05,
+      "loss": 1.0524,
+      "step": 7664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053144139821927e-05,
+      "loss": 1.0026,
+      "step": 7665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050975010445814e-05,
+      "loss": 0.9832,
+      "step": 7666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0048805878671193e-05,
+      "loss": 1.0106,
+      "step": 7667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046636744600138e-05,
+      "loss": 0.9437,
+      "step": 7668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044467608334702e-05,
+      "loss": 1.0185,
+      "step": 7669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004229846997696e-05,
+      "loss": 0.9476,
+      "step": 7670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0040129329628967e-05,
+      "loss": 0.9485,
+      "step": 7671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003796018739279e-05,
+      "loss": 0.9631,
+      "step": 7672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003579104337049e-05,
+      "loss": 0.9791,
+      "step": 7673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033621897664132e-05,
+      "loss": 1.0154,
+      "step": 7674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003145275037578e-05,
+      "loss": 0.9472,
+      "step": 7675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00292836016075e-05,
+      "loss": 0.9191,
+      "step": 7676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002711445146135e-05,
+      "loss": 0.9948,
+      "step": 7677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00249453000394e-05,
+      "loss": 0.9481,
+      "step": 7678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022776147443709e-05,
+      "loss": 0.9444,
+      "step": 7679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0020606993776343e-05,
+      "loss": 0.9837,
+      "step": 7680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001843783913937e-05,
+      "loss": 0.9502,
+      "step": 7681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016268683634847e-05,
+      "loss": 0.9792,
+      "step": 7682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014099527364843e-05,
+      "loss": 0.9597,
+      "step": 7683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001193037043142e-05,
+      "loss": 0.9484,
+      "step": 7684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009761212936646e-05,
+      "loss": 0.9471,
+      "step": 7685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000759205498258e-05,
+      "loss": 0.9922,
+      "step": 7686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005422896671291e-05,
+      "loss": 0.978,
+      "step": 7687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003253738104838e-05,
+      "loss": 0.9985,
+      "step": 7688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0001084579385292e-05,
+      "loss": 0.9661,
+      "step": 7689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998915420614713e-06,
+      "loss": 0.9844,
+      "step": 7690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.996746261895163e-06,
+      "loss": 1.0089,
+      "step": 7691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.994577103328712e-06,
+      "loss": 1.0077,
+      "step": 7692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992407945017426e-06,
+      "loss": 0.9963,
+      "step": 7693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990238787063355e-06,
+      "loss": 0.9105,
+      "step": 7694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988069629568584e-06,
+      "loss": 0.9866,
+      "step": 7695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985900472635159e-06,
+      "loss": 1.0679,
+      "step": 7696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.983731316365156e-06,
+      "loss": 1.0457,
+      "step": 7697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981562160860636e-06,
+      "loss": 0.992,
+      "step": 7698
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.979393006223658e-06,
+      "loss": 0.9881,
+      "step": 7699
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977223852556294e-06,
+      "loss": 0.9373,
+      "step": 7700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975054699960602e-06,
+      "loss": 0.8822,
+      "step": 7701
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972885548538651e-06,
+      "loss": 0.9462,
+      "step": 7702
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970716398392505e-06,
+      "loss": 0.9713,
+      "step": 7703
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.968547249624221e-06,
+      "loss": 0.93,
+      "step": 7704
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96637810233587e-06,
+      "loss": 0.9968,
+      "step": 7705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96420895662951e-06,
+      "loss": 1.038,
+      "step": 7706
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.962039812607213e-06,
+      "loss": 0.9576,
+      "step": 7707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959870670371035e-06,
+      "loss": 1.0156,
+      "step": 7708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95770153002304e-06,
+      "loss": 0.9381,
+      "step": 7709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9555323916653e-06,
+      "loss": 1.0001,
+      "step": 7710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953363255399869e-06,
+      "loss": 0.9261,
+      "step": 7711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95119412132881e-06,
+      "loss": 0.9197,
+      "step": 7712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949024989554192e-06,
+      "loss": 0.9529,
+      "step": 7713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946855860178073e-06,
+      "loss": 0.8537,
+      "step": 7714
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944686733302523e-06,
+      "loss": 0.9931,
+      "step": 7715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942517609029598e-06,
+      "loss": 0.9449,
+      "step": 7716
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940348487461362e-06,
+      "loss": 0.9477,
+      "step": 7717
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938179368699883e-06,
+      "loss": 0.9756,
+      "step": 7718
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93601025284721e-06,
+      "loss": 0.9459,
+      "step": 7719
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.933841140005425e-06,
+      "loss": 1.0251,
+      "step": 7720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931672030276576e-06,
+      "loss": 0.916,
+      "step": 7721
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929502923762725e-06,
+      "loss": 0.9651,
+      "step": 7722
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927333820565942e-06,
+      "loss": 0.9881,
+      "step": 7723
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.925164720788281e-06,
+      "loss": 0.9373,
+      "step": 7724
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.922995624531808e-06,
+      "loss": 0.9195,
+      "step": 7725
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920826531898587e-06,
+      "loss": 0.9514,
+      "step": 7726
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918657442990669e-06,
+      "loss": 0.9006,
+      "step": 7727
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916488357910127e-06,
+      "loss": 0.9227,
+      "step": 7728
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.914319276759011e-06,
+      "loss": 0.9441,
+      "step": 7729
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91215019963939e-06,
+      "loss": 0.9214,
+      "step": 7730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.909981126653323e-06,
+      "loss": 0.9012,
+      "step": 7731
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907812057902865e-06,
+      "loss": 1.0096,
+      "step": 7732
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905642993490084e-06,
+      "loss": 0.973,
+      "step": 7733
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903473933517031e-06,
+      "loss": 0.9909,
+      "step": 7734
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901304878085777e-06,
+      "loss": 0.986,
+      "step": 7735
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.899135827298371e-06,
+      "loss": 0.9369,
+      "step": 7736
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896966781256875e-06,
+      "loss": 1.0046,
+      "step": 7737
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894797740063353e-06,
+      "loss": 0.947,
+      "step": 7738
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892628703819856e-06,
+      "loss": 1.0116,
+      "step": 7739
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890459672628449e-06,
+      "loss": 1.0084,
+      "step": 7740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.888290646591188e-06,
+      "loss": 0.8871,
+      "step": 7741
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88612162581013e-06,
+      "loss": 0.9996,
+      "step": 7742
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883952610387337e-06,
+      "loss": 0.9417,
+      "step": 7743
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.881783600424857e-06,
+      "loss": 0.9996,
+      "step": 7744
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879614596024763e-06,
+      "loss": 0.9426,
+      "step": 7745
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877445597289099e-06,
+      "loss": 0.939,
+      "step": 7746
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.875276604319921e-06,
+      "loss": 1.0285,
+      "step": 7747
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873107617219296e-06,
+      "loss": 0.9196,
+      "step": 7748
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870938636089276e-06,
+      "loss": 1.0172,
+      "step": 7749
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868769661031908e-06,
+      "loss": 0.9332,
+      "step": 7750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866600692149263e-06,
+      "loss": 1.0571,
+      "step": 7751
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864431729543382e-06,
+      "loss": 0.9388,
+      "step": 7752
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862262773316332e-06,
+      "loss": 1.0139,
+      "step": 7753
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86009382357016e-06,
+      "loss": 0.9758,
+      "step": 7754
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857924880406923e-06,
+      "loss": 0.9359,
+      "step": 7755
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855755943928676e-06,
+      "loss": 1.078,
+      "step": 7756
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85358701423747e-06,
+      "loss": 0.9316,
+      "step": 7757
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851418091435363e-06,
+      "loss": 1.0111,
+      "step": 7758
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849249175624409e-06,
+      "loss": 0.9381,
+      "step": 7759
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.847080266906653e-06,
+      "loss": 0.9724,
+      "step": 7760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.844911365384153e-06,
+      "loss": 0.9521,
+      "step": 7761
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84274247115896e-06,
+      "loss": 1.0364,
+      "step": 7762
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.840573584333129e-06,
+      "loss": 0.9269,
+      "step": 7763
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838404705008708e-06,
+      "loss": 1.0212,
+      "step": 7764
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.836235833287747e-06,
+      "loss": 1.0014,
+      "step": 7765
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.834066969272302e-06,
+      "loss": 0.8843,
+      "step": 7766
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831898113064419e-06,
+      "loss": 0.9565,
+      "step": 7767
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.829729264766153e-06,
+      "loss": 0.9314,
+      "step": 7768
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827560424479548e-06,
+      "loss": 0.9833,
+      "step": 7769
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.825391592306653e-06,
+      "loss": 1.0473,
+      "step": 7770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823222768349524e-06,
+      "loss": 0.9936,
+      "step": 7771
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8210539527102e-06,
+      "loss": 0.8753,
+      "step": 7772
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.81888514549074e-06,
+      "loss": 0.9919,
+      "step": 7773
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.816716346793187e-06,
+      "loss": 0.9236,
+      "step": 7774
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814547556719582e-06,
+      "loss": 0.9929,
+      "step": 7775
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812378775371982e-06,
+      "loss": 1.0084,
+      "step": 7776
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810210002852424e-06,
+      "loss": 1.0438,
+      "step": 7777
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.808041239262966e-06,
+      "loss": 1.0741,
+      "step": 7778
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.805872484705646e-06,
+      "loss": 1.0255,
+      "step": 7779
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.803703739282507e-06,
+      "loss": 0.9967,
+      "step": 7780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8015350030956e-06,
+      "loss": 0.9254,
+      "step": 7781
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.799366276246965e-06,
+      "loss": 0.9177,
+      "step": 7782
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.797197558838649e-06,
+      "loss": 0.9236,
+      "step": 7783
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.795028850972696e-06,
+      "loss": 0.9731,
+      "step": 7784
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792860152751143e-06,
+      "loss": 1.0298,
+      "step": 7785
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.790691464276043e-06,
+      "loss": 0.9638,
+      "step": 7786
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78852278564943e-06,
+      "loss": 1.0779,
+      "step": 7787
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786354116973345e-06,
+      "loss": 0.9957,
+      "step": 7788
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784185458349837e-06,
+      "loss": 0.9441,
+      "step": 7789
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782016809880938e-06,
+      "loss": 1.0366,
+      "step": 7790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.779848171668695e-06,
+      "loss": 0.9266,
+      "step": 7791
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777679543815148e-06,
+      "loss": 1.0065,
+      "step": 7792
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775510926422329e-06,
+      "loss": 0.9666,
+      "step": 7793
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.773342319592285e-06,
+      "loss": 1.0224,
+      "step": 7794
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771173723427047e-06,
+      "loss": 1.0398,
+      "step": 7795
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76900513802866e-06,
+      "loss": 0.9005,
+      "step": 7796
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.766836563499157e-06,
+      "loss": 0.9401,
+      "step": 7797
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764667999940576e-06,
+      "loss": 0.9189,
+      "step": 7798
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762499447454955e-06,
+      "loss": 1.0258,
+      "step": 7799
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760330906144326e-06,
+      "loss": 0.9592,
+      "step": 7800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75816237611073e-06,
+      "loss": 0.9273,
+      "step": 7801
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.755993857456198e-06,
+      "loss": 0.8649,
+      "step": 7802
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75382535028276e-06,
+      "loss": 0.9926,
+      "step": 7803
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751656854692458e-06,
+      "loss": 0.8482,
+      "step": 7804
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.749488370787318e-06,
+      "loss": 1.0723,
+      "step": 7805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747319898669378e-06,
+      "loss": 1.0018,
+      "step": 7806
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74515143844067e-06,
+      "loss": 0.9958,
+      "step": 7807
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742982990203217e-06,
+      "loss": 0.9317,
+      "step": 7808
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740814554059064e-06,
+      "loss": 0.907,
+      "step": 7809
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.738646130110225e-06,
+      "loss": 0.9957,
+      "step": 7810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.736477718458746e-06,
+      "loss": 0.9785,
+      "step": 7811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734309319206647e-06,
+      "loss": 1.0292,
+      "step": 7812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732140932455956e-06,
+      "loss": 0.9893,
+      "step": 7813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729972558308703e-06,
+      "loss": 1.0463,
+      "step": 7814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.727804196866916e-06,
+      "loss": 1.0762,
+      "step": 7815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.725635848232623e-06,
+      "loss": 0.9044,
+      "step": 7816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72346751250785e-06,
+      "loss": 0.9188,
+      "step": 7817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721299189794617e-06,
+      "loss": 0.9713,
+      "step": 7818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719130880194957e-06,
+      "loss": 0.963,
+      "step": 7819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716962583810886e-06,
+      "loss": 0.9695,
+      "step": 7820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714794300744438e-06,
+      "loss": 1.0425,
+      "step": 7821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712626031097626e-06,
+      "loss": 0.983,
+      "step": 7822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.710457774972478e-06,
+      "loss": 1.0368,
+      "step": 7823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.708289532471016e-06,
+      "loss": 0.9591,
+      "step": 7824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706121303695263e-06,
+      "loss": 0.981,
+      "step": 7825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70395308874723e-06,
+      "loss": 0.9892,
+      "step": 7826
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.701784887728949e-06,
+      "loss": 1.0147,
+      "step": 7827
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69961670074243e-06,
+      "loss": 0.9529,
+      "step": 7828
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697448527889697e-06,
+      "loss": 0.983,
+      "step": 7829
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695280369272769e-06,
+      "loss": 1.0631,
+      "step": 7830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693112224993657e-06,
+      "loss": 1.0354,
+      "step": 7831
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690944095154385e-06,
+      "loss": 0.9888,
+      "step": 7832
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688775979856959e-06,
+      "loss": 0.96,
+      "step": 7833
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.686607879203408e-06,
+      "loss": 0.9384,
+      "step": 7834
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684439793295737e-06,
+      "loss": 0.9843,
+      "step": 7835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.682271722235959e-06,
+      "loss": 0.8957,
+      "step": 7836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680103666126092e-06,
+      "loss": 0.994,
+      "step": 7837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.677935625068145e-06,
+      "loss": 0.9368,
+      "step": 7838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.675767599164133e-06,
+      "loss": 0.9711,
+      "step": 7839
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673599588516066e-06,
+      "loss": 0.9519,
+      "step": 7840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67143159322595e-06,
+      "loss": 0.8776,
+      "step": 7841
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669263613395803e-06,
+      "loss": 0.9398,
+      "step": 7842
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667095649127623e-06,
+      "loss": 0.9577,
+      "step": 7843
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66492770052343e-06,
+      "loss": 0.9849,
+      "step": 7844
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662759767685225e-06,
+      "loss": 0.9253,
+      "step": 7845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660591850715012e-06,
+      "loss": 0.9349,
+      "step": 7846
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658423949714803e-06,
+      "loss": 0.9086,
+      "step": 7847
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656256064786598e-06,
+      "loss": 1.0426,
+      "step": 7848
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654088196032405e-06,
+      "loss": 1.0443,
+      "step": 7849
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.651920343554229e-06,
+      "loss": 0.9864,
+      "step": 7850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649752507454067e-06,
+      "loss": 0.9711,
+      "step": 7851
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647584687833925e-06,
+      "loss": 0.96,
+      "step": 7852
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645416884795802e-06,
+      "loss": 0.9877,
+      "step": 7853
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643249098441703e-06,
+      "loss": 0.9712,
+      "step": 7854
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641081328873625e-06,
+      "loss": 0.9846,
+      "step": 7855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.638913576193563e-06,
+      "loss": 1.0514,
+      "step": 7856
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636745840503525e-06,
+      "loss": 0.9674,
+      "step": 7857
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634578121905494e-06,
+      "loss": 0.9795,
+      "step": 7858
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.632410420501484e-06,
+      "loss": 0.9475,
+      "step": 7859
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63024273639348e-06,
+      "loss": 0.9866,
+      "step": 7860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628075069683475e-06,
+      "loss": 1.003,
+      "step": 7861
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62590742047347e-06,
+      "loss": 0.9193,
+      "step": 7862
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623739788865456e-06,
+      "loss": 0.8929,
+      "step": 7863
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62157217496142e-06,
+      "loss": 0.9971,
+      "step": 7864
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.619404578863364e-06,
+      "loss": 1.0675,
+      "step": 7865
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617237000673267e-06,
+      "loss": 0.9889,
+      "step": 7866
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615069440493132e-06,
+      "loss": 0.9428,
+      "step": 7867
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612901898424938e-06,
+      "loss": 0.94,
+      "step": 7868
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610734374570675e-06,
+      "loss": 0.9827,
+      "step": 7869
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608566869032334e-06,
+      "loss": 0.9641,
+      "step": 7870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606399381911898e-06,
+      "loss": 0.9617,
+      "step": 7871
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604231913311356e-06,
+      "loss": 1.0062,
+      "step": 7872
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602064463332693e-06,
+      "loss": 1.1016,
+      "step": 7873
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.599897032077884e-06,
+      "loss": 1.0146,
+      "step": 7874
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597729619648927e-06,
+      "loss": 0.9351,
+      "step": 7875
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59556222614779e-06,
+      "loss": 0.9269,
+      "step": 7876
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593394851676464e-06,
+      "loss": 1.0068,
+      "step": 7877
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591227496336924e-06,
+      "loss": 1.0001,
+      "step": 7878
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58906016023115e-06,
+      "loss": 0.9237,
+      "step": 7879
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586892843461126e-06,
+      "loss": 0.9699,
+      "step": 7880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584725546128821e-06,
+      "loss": 0.9442,
+      "step": 7881
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.582558268336218e-06,
+      "loss": 0.9294,
+      "step": 7882
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580391010185294e-06,
+      "loss": 0.9231,
+      "step": 7883
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578223771778018e-06,
+      "loss": 0.9729,
+      "step": 7884
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576056553216369e-06,
+      "loss": 0.9431,
+      "step": 7885
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573889354602315e-06,
+      "loss": 0.9792,
+      "step": 7886
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.571722176037833e-06,
+      "loss": 1.0256,
+      "step": 7887
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.569555017624893e-06,
+      "loss": 0.9507,
+      "step": 7888
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567387879465463e-06,
+      "loss": 1.0081,
+      "step": 7889
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.565220761661517e-06,
+      "loss": 0.9465,
+      "step": 7890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563053664315015e-06,
+      "loss": 0.9217,
+      "step": 7891
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560886587527934e-06,
+      "loss": 1.0341,
+      "step": 7892
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558719531402236e-06,
+      "loss": 0.9469,
+      "step": 7893
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556552496039881e-06,
+      "loss": 0.9637,
+      "step": 7894
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554385481542842e-06,
+      "loss": 0.9349,
+      "step": 7895
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552218488013082e-06,
+      "loss": 0.9603,
+      "step": 7896
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550051515552556e-06,
+      "loss": 0.9959,
+      "step": 7897
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547884564263234e-06,
+      "loss": 0.9405,
+      "step": 7898
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.545717634247065e-06,
+      "loss": 0.9676,
+      "step": 7899
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543550725606025e-06,
+      "loss": 0.9693,
+      "step": 7900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54138383844206e-06,
+      "loss": 0.925,
+      "step": 7901
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.539216972857129e-06,
+      "loss": 1.0225,
+      "step": 7902
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537050128953192e-06,
+      "loss": 0.9718,
+      "step": 7903
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.534883306832202e-06,
+      "loss": 0.9112,
+      "step": 7904
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532716506596116e-06,
+      "loss": 1.0437,
+      "step": 7905
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530549728346886e-06,
+      "loss": 0.9397,
+      "step": 7906
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528382972186459e-06,
+      "loss": 0.9608,
+      "step": 7907
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.526216238216797e-06,
+      "loss": 0.9543,
+      "step": 7908
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52404952653984e-06,
+      "loss": 0.9944,
+      "step": 7909
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.521882837257543e-06,
+      "loss": 1.0413,
+      "step": 7910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519716170471854e-06,
+      "loss": 1.0117,
+      "step": 7911
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517549526284717e-06,
+      "loss": 0.9729,
+      "step": 7912
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51538290479808e-06,
+      "loss": 0.9713,
+      "step": 7913
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.513216306113886e-06,
+      "loss": 0.995,
+      "step": 7914
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.511049730334086e-06,
+      "loss": 0.913,
+      "step": 7915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.508883177560616e-06,
+      "loss": 0.9567,
+      "step": 7916
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506716647895415e-06,
+      "loss": 0.9041,
+      "step": 7917
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.504550141440432e-06,
+      "loss": 1.0018,
+      "step": 7918
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502383658297599e-06,
+      "loss": 1.03,
+      "step": 7919
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500217198568859e-06,
+      "loss": 0.9941,
+      "step": 7920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49805076235615e-06,
+      "loss": 0.9537,
+      "step": 7921
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495884349761405e-06,
+      "loss": 1.0757,
+      "step": 7922
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49371796088656e-06,
+      "loss": 0.9557,
+      "step": 7923
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491551595833548e-06,
+      "loss": 1.0256,
+      "step": 7924
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489385254704308e-06,
+      "loss": 0.9938,
+      "step": 7925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487218937600765e-06,
+      "loss": 1.0057,
+      "step": 7926
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485052644624848e-06,
+      "loss": 0.9521,
+      "step": 7927
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482886375878496e-06,
+      "loss": 0.9888,
+      "step": 7928
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480720131463627e-06,
+      "loss": 0.9998,
+      "step": 7929
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478553911482176e-06,
+      "loss": 0.9986,
+      "step": 7930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476387716036067e-06,
+      "loss": 0.9692,
+      "step": 7931
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474221545227218e-06,
+      "loss": 1.0356,
+      "step": 7932
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472055399157565e-06,
+      "loss": 1.0188,
+      "step": 7933
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.469889277929022e-06,
+      "loss": 0.9485,
+      "step": 7934
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46772318164351e-06,
+      "loss": 1.0434,
+      "step": 7935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465557110402954e-06,
+      "loss": 0.91,
+      "step": 7936
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46339106430927e-06,
+      "loss": 1.0131,
+      "step": 7937
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461225043464378e-06,
+      "loss": 0.979,
+      "step": 7938
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.459059047970195e-06,
+      "loss": 0.9439,
+      "step": 7939
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45689307792863e-06,
+      "loss": 0.9169,
+      "step": 7940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454727133441606e-06,
+      "loss": 0.9108,
+      "step": 7941
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452561214611028e-06,
+      "loss": 1.0858,
+      "step": 7942
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450395321538816e-06,
+      "loss": 0.9932,
+      "step": 7943
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448229454326878e-06,
+      "loss": 0.9693,
+      "step": 7944
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446063613077119e-06,
+      "loss": 1.0669,
+      "step": 7945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443897797891453e-06,
+      "loss": 0.9797,
+      "step": 7946
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441732008871781e-06,
+      "loss": 0.9725,
+      "step": 7947
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439566246120018e-06,
+      "loss": 0.8919,
+      "step": 7948
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43740050973806e-06,
+      "loss": 0.9762,
+      "step": 7949
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435234799827813e-06,
+      "loss": 0.9817,
+      "step": 7950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.433069116491179e-06,
+      "loss": 0.9186,
+      "step": 7951
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430903459830058e-06,
+      "loss": 0.9762,
+      "step": 7952
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428737829946353e-06,
+      "loss": 1.0062,
+      "step": 7953
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426572226941962e-06,
+      "loss": 0.9571,
+      "step": 7954
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.424406650918774e-06,
+      "loss": 0.9274,
+      "step": 7955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.422241101978696e-06,
+      "loss": 1.0145,
+      "step": 7956
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420075580223611e-06,
+      "loss": 0.8985,
+      "step": 7957
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.417910085755426e-06,
+      "loss": 0.9986,
+      "step": 7958
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415744618676023e-06,
+      "loss": 1.0134,
+      "step": 7959
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413579179087292e-06,
+      "loss": 0.9903,
+      "step": 7960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.411413767091128e-06,
+      "loss": 0.9873,
+      "step": 7961
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409248382789413e-06,
+      "loss": 1.0244,
+      "step": 7962
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40708302628404e-06,
+      "loss": 0.8685,
+      "step": 7963
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.404917697676895e-06,
+      "loss": 0.9341,
+      "step": 7964
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402752397069851e-06,
+      "loss": 0.9843,
+      "step": 7965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400587124564806e-06,
+      "loss": 0.96,
+      "step": 7966
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.398421880263629e-06,
+      "loss": 0.8806,
+      "step": 7967
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.396256664268206e-06,
+      "loss": 0.9168,
+      "step": 7968
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394091476680418e-06,
+      "loss": 0.9971,
+      "step": 7969
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.391926317602134e-06,
+      "loss": 0.9019,
+      "step": 7970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38976118713524e-06,
+      "loss": 0.9052,
+      "step": 7971
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.387596085381606e-06,
+      "loss": 0.9201,
+      "step": 7972
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.385431012443103e-06,
+      "loss": 0.9934,
+      "step": 7973
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383265968421608e-06,
+      "loss": 0.9629,
+      "step": 7974
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.381100953418988e-06,
+      "loss": 0.958,
+      "step": 7975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.378935967537116e-06,
+      "loss": 0.9082,
+      "step": 7976
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376771010877858e-06,
+      "loss": 0.9705,
+      "step": 7977
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374606083543078e-06,
+      "loss": 0.9742,
+      "step": 7978
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.372441185634645e-06,
+      "loss": 0.9328,
+      "step": 7979
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37027631725442e-06,
+      "loss": 0.9751,
+      "step": 7980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368111478504273e-06,
+      "loss": 0.9657,
+      "step": 7981
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365946669486056e-06,
+      "loss": 0.9387,
+      "step": 7982
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36378189030163e-06,
+      "loss": 1.0228,
+      "step": 7983
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.361617141052856e-06,
+      "loss": 1.062,
+      "step": 7984
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.35945242184159e-06,
+      "loss": 1.0331,
+      "step": 7985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357287732769689e-06,
+      "loss": 1.0213,
+      "step": 7986
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.355123073939008e-06,
+      "loss": 0.9699,
+      "step": 7987
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.352958445451388e-06,
+      "loss": 1.0321,
+      "step": 7988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.350793847408699e-06,
+      "loss": 1.0405,
+      "step": 7989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348629279912773e-06,
+      "loss": 0.9557,
+      "step": 7990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346464743065473e-06,
+      "loss": 0.9739,
+      "step": 7991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344300236968636e-06,
+      "loss": 0.9642,
+      "step": 7992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.342135761724111e-06,
+      "loss": 0.9875,
+      "step": 7993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339971317433741e-06,
+      "loss": 1.0007,
+      "step": 7994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337806904199367e-06,
+      "loss": 0.9178,
+      "step": 7995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335642522122836e-06,
+      "loss": 0.9687,
+      "step": 7996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.333478171305986e-06,
+      "loss": 0.9878,
+      "step": 7997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331313851850648e-06,
+      "loss": 0.9941,
+      "step": 7998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.329149563858665e-06,
+      "loss": 0.9399,
+      "step": 7999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32698530743187e-06,
+      "loss": 1.0209,
+      "step": 8000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.324821082672098e-06,
+      "loss": 0.9805,
+      "step": 8001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.322656889681179e-06,
+      "loss": 1.0323,
+      "step": 8002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320492728560946e-06,
+      "loss": 0.8955,
+      "step": 8003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.318328599413228e-06,
+      "loss": 0.9762,
+      "step": 8004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31616450233985e-06,
+      "loss": 0.9954,
+      "step": 8005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314000437442645e-06,
+      "loss": 0.9232,
+      "step": 8006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31183640482343e-06,
+      "loss": 1.0301,
+      "step": 8007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309672404584029e-06,
+      "loss": 0.9412,
+      "step": 8008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307508436826269e-06,
+      "loss": 1.0348,
+      "step": 8009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305344501651966e-06,
+      "loss": 0.8992,
+      "step": 8010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303180599162936e-06,
+      "loss": 0.9162,
+      "step": 8011
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301016729461003e-06,
+      "loss": 0.9901,
+      "step": 8012
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.298852892647974e-06,
+      "loss": 0.9638,
+      "step": 8013
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296689088825674e-06,
+      "loss": 0.9854,
+      "step": 8014
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294525318095906e-06,
+      "loss": 0.9539,
+      "step": 8015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.292361580560483e-06,
+      "loss": 0.9318,
+      "step": 8016
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.290197876321217e-06,
+      "loss": 1.0049,
+      "step": 8017
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288034205479911e-06,
+      "loss": 1.0367,
+      "step": 8018
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285870568138376e-06,
+      "loss": 0.9745,
+      "step": 8019
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283706964398417e-06,
+      "loss": 0.968,
+      "step": 8020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.281543394361827e-06,
+      "loss": 0.944,
+      "step": 8021
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.279379858130423e-06,
+      "loss": 1.0064,
+      "step": 8022
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277216355805988e-06,
+      "loss": 0.9332,
+      "step": 8023
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.275052887490338e-06,
+      "loss": 0.9896,
+      "step": 8024
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272889453285256e-06,
+      "loss": 0.9361,
+      "step": 8025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27072605329254e-06,
+      "loss": 0.9484,
+      "step": 8026
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268562687613987e-06,
+      "loss": 0.9625,
+      "step": 8027
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.266399356351385e-06,
+      "loss": 0.8829,
+      "step": 8028
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264236059606525e-06,
+      "loss": 0.9579,
+      "step": 8029
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2620727974812e-06,
+      "loss": 0.9891,
+      "step": 8030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25990957007719e-06,
+      "loss": 1.008,
+      "step": 8031
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257746377496285e-06,
+      "loss": 0.9933,
+      "step": 8032
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255583219840263e-06,
+      "loss": 0.9543,
+      "step": 8033
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253420097210913e-06,
+      "loss": 0.9222,
+      "step": 8034
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.251257009710011e-06,
+      "loss": 0.9493,
+      "step": 8035
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249093957439337e-06,
+      "loss": 1.0144,
+      "step": 8036
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.246930940500671e-06,
+      "loss": 0.9187,
+      "step": 8037
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244767958995778e-06,
+      "loss": 0.9423,
+      "step": 8038
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.242605013026447e-06,
+      "loss": 1.0214,
+      "step": 8039
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24044210269444e-06,
+      "loss": 0.9482,
+      "step": 8040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.238279228101524e-06,
+      "loss": 0.9466,
+      "step": 8041
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236116389349477e-06,
+      "loss": 0.9188,
+      "step": 8042
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23395358654006e-06,
+      "loss": 1.0189,
+      "step": 8043
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231790819775042e-06,
+      "loss": 0.9492,
+      "step": 8044
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229628089156188e-06,
+      "loss": 1.0179,
+      "step": 8045
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227465394785248e-06,
+      "loss": 0.9856,
+      "step": 8046
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225302736763998e-06,
+      "loss": 0.9568,
+      "step": 8047
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.223140115194186e-06,
+      "loss": 0.9908,
+      "step": 8048
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.220977530177572e-06,
+      "loss": 0.9572,
+      "step": 8049
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218814981815912e-06,
+      "loss": 1.0468,
+      "step": 8050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216652470210957e-06,
+      "loss": 0.9361,
+      "step": 8051
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214489995464461e-06,
+      "loss": 0.9391,
+      "step": 8052
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.212327557678175e-06,
+      "loss": 1.0126,
+      "step": 8053
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21016515695384e-06,
+      "loss": 0.9248,
+      "step": 8054
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20800279339321e-06,
+      "loss": 0.9934,
+      "step": 8055
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205840467098025e-06,
+      "loss": 0.956,
+      "step": 8056
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20367817817003e-06,
+      "loss": 1.0048,
+      "step": 8057
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.201515926710968e-06,
+      "loss": 0.9545,
+      "step": 8058
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.199353712822573e-06,
+      "loss": 0.9967,
+      "step": 8059
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197191536606588e-06,
+      "loss": 0.9439,
+      "step": 8060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195029398164744e-06,
+      "loss": 0.9968,
+      "step": 8061
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19286729759878e-06,
+      "loss": 0.9069,
+      "step": 8062
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190705235010427e-06,
+      "loss": 0.9494,
+      "step": 8063
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.188543210501409e-06,
+      "loss": 0.9204,
+      "step": 8064
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186381224173463e-06,
+      "loss": 0.9476,
+      "step": 8065
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18421927612831e-06,
+      "loss": 0.9979,
+      "step": 8066
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18205736646768e-06,
+      "loss": 0.9446,
+      "step": 8067
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179895495293295e-06,
+      "loss": 0.9126,
+      "step": 8068
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17773366270687e-06,
+      "loss": 0.9312,
+      "step": 8069
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.175571868810139e-06,
+      "loss": 0.9442,
+      "step": 8070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1734101137048e-06,
+      "loss": 1.006,
+      "step": 8071
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171248397492588e-06,
+      "loss": 0.9675,
+      "step": 8072
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.169086720275207e-06,
+      "loss": 0.9973,
+      "step": 8073
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166925082154368e-06,
+      "loss": 0.9372,
+      "step": 8074
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16476348323179e-06,
+      "loss": 0.9989,
+      "step": 8075
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16260192360917e-06,
+      "loss": 1.0229,
+      "step": 8076
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.160440403388225e-06,
+      "loss": 0.9751,
+      "step": 8077
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158278922670658e-06,
+      "loss": 1.0292,
+      "step": 8078
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.156117481558165e-06,
+      "loss": 0.9328,
+      "step": 8079
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153956080152456e-06,
+      "loss": 1.003,
+      "step": 8080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151794718555225e-06,
+      "loss": 1.0231,
+      "step": 8081
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149633396868172e-06,
+      "loss": 0.9594,
+      "step": 8082
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147472115192991e-06,
+      "loss": 0.8959,
+      "step": 8083
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145310873631375e-06,
+      "loss": 1.02,
+      "step": 8084
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.143149672285018e-06,
+      "loss": 1.0512,
+      "step": 8085
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140988511255612e-06,
+      "loss": 1.0169,
+      "step": 8086
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138827390644838e-06,
+      "loss": 1.0283,
+      "step": 8087
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.13666631055439e-06,
+      "loss": 1.0125,
+      "step": 8088
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134505271085945e-06,
+      "loss": 0.9699,
+      "step": 8089
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132344272341192e-06,
+      "loss": 0.9933,
+      "step": 8090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130183314421806e-06,
+      "loss": 1.0033,
+      "step": 8091
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.128022397429466e-06,
+      "loss": 0.9846,
+      "step": 8092
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125861521465853e-06,
+      "loss": 0.9274,
+      "step": 8093
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123700686632636e-06,
+      "loss": 0.9468,
+      "step": 8094
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121539893031496e-06,
+      "loss": 0.9672,
+      "step": 8095
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.119379140764094e-06,
+      "loss": 1.0366,
+      "step": 8096
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117218429932103e-06,
+      "loss": 1.0526,
+      "step": 8097
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11505776063719e-06,
+      "loss": 0.9442,
+      "step": 8098
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112897132981019e-06,
+      "loss": 0.9197,
+      "step": 8099
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110736547065256e-06,
+      "loss": 0.9734,
+      "step": 8100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.108576002991558e-06,
+      "loss": 0.9666,
+      "step": 8101
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106415500861585e-06,
+      "loss": 0.9524,
+      "step": 8102
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104255040776998e-06,
+      "loss": 1.0252,
+      "step": 8103
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.102094622839443e-06,
+      "loss": 0.9812,
+      "step": 8104
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.099934247150584e-06,
+      "loss": 1.0349,
+      "step": 8105
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097773913812065e-06,
+      "loss": 0.9782,
+      "step": 8106
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095613622925535e-06,
+      "loss": 1.0235,
+      "step": 8107
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093453374592645e-06,
+      "loss": 1.0048,
+      "step": 8108
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091293168915036e-06,
+      "loss": 0.9609,
+      "step": 8109
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089133005994355e-06,
+      "loss": 1.0131,
+      "step": 8110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086972885932244e-06,
+      "loss": 0.991,
+      "step": 8111
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084812808830331e-06,
+      "loss": 0.9541,
+      "step": 8112
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082652774790268e-06,
+      "loss": 0.9207,
+      "step": 8113
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.080492783913682e-06,
+      "loss": 0.9351,
+      "step": 8114
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078332836302205e-06,
+      "loss": 1.0054,
+      "step": 8115
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076172932057474e-06,
+      "loss": 1.016,
+      "step": 8116
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07401307128111e-06,
+      "loss": 0.9453,
+      "step": 8117
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.071853254074744e-06,
+      "loss": 0.9871,
+      "step": 8118
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069693480540002e-06,
+      "loss": 0.9525,
+      "step": 8119
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067533750778506e-06,
+      "loss": 0.964,
+      "step": 8120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065374064891879e-06,
+      "loss": 0.9466,
+      "step": 8121
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063214422981731e-06,
+      "loss": 0.9985,
+      "step": 8122
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.061054825149688e-06,
+      "loss": 0.9141,
+      "step": 8123
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058895271497358e-06,
+      "loss": 0.9407,
+      "step": 8124
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056735762126356e-06,
+      "loss": 1.0526,
+      "step": 8125
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054576297138296e-06,
+      "loss": 0.9847,
+      "step": 8126
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.052416876634777e-06,
+      "loss": 1.0163,
+      "step": 8127
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050257500717417e-06,
+      "loss": 0.9897,
+      "step": 8128
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048098169487812e-06,
+      "loss": 0.9955,
+      "step": 8129
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.045938883047563e-06,
+      "loss": 0.9517,
+      "step": 8130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.043779641498275e-06,
+      "loss": 1.0426,
+      "step": 8131
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04162044494154e-06,
+      "loss": 0.968,
+      "step": 8132
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039461293478961e-06,
+      "loss": 0.8847,
+      "step": 8133
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037302187212128e-06,
+      "loss": 0.9706,
+      "step": 8134
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035143126242624e-06,
+      "loss": 0.9633,
+      "step": 8135
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.032984110672055e-06,
+      "loss": 1.012,
+      "step": 8136
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.03082514060199e-06,
+      "loss": 1.0324,
+      "step": 8137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028666216134032e-06,
+      "loss": 0.9179,
+      "step": 8138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02650733736975e-06,
+      "loss": 0.9914,
+      "step": 8139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024348504410728e-06,
+      "loss": 1.0045,
+      "step": 8140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.022189717358549e-06,
+      "loss": 0.9573,
+      "step": 8141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020030976314782e-06,
+      "loss": 0.9436,
+      "step": 8142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017872281381008e-06,
+      "loss": 0.9278,
+      "step": 8143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.0157136326588e-06,
+      "loss": 0.9645,
+      "step": 8144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.013555030249715e-06,
+      "loss": 1.0052,
+      "step": 8145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.011396474255338e-06,
+      "loss": 1.0046,
+      "step": 8146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009237964777221e-06,
+      "loss": 0.9028,
+      "step": 8147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007079501916935e-06,
+      "loss": 0.9613,
+      "step": 8148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00492108577604e-06,
+      "loss": 0.8805,
+      "step": 8149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.002762716456089e-06,
+      "loss": 0.9008,
+      "step": 8150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000604394058646e-06,
+      "loss": 0.9972,
+      "step": 8151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99844611868526e-06,
+      "loss": 0.9448,
+      "step": 8152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996287890437489e-06,
+      "loss": 0.9827,
+      "step": 8153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994129709416878e-06,
+      "loss": 0.9116,
+      "step": 8154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991971575724976e-06,
+      "loss": 0.9754,
+      "step": 8155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98981348946333e-06,
+      "loss": 0.9938,
+      "step": 8156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.987655450733481e-06,
+      "loss": 1.0525,
+      "step": 8157
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.985497459636973e-06,
+      "loss": 0.964,
+      "step": 8158
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983339516275344e-06,
+      "loss": 0.973,
+      "step": 8159
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98118162075013e-06,
+      "loss": 0.9518,
+      "step": 8160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.979023773162866e-06,
+      "loss": 1.0541,
+      "step": 8161
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.976865973615084e-06,
+      "loss": 0.957,
+      "step": 8162
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974708222208308e-06,
+      "loss": 0.9127,
+      "step": 8163
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972550519044078e-06,
+      "loss": 1.0063,
+      "step": 8164
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970392864223907e-06,
+      "loss": 0.9465,
+      "step": 8165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968235257849325e-06,
+      "loss": 0.978,
+      "step": 8166
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966077700021856e-06,
+      "loss": 0.904,
+      "step": 8167
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963920190843007e-06,
+      "loss": 1.0652,
+      "step": 8168
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961762730414305e-06,
+      "loss": 0.9798,
+      "step": 8169
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959605318837254e-06,
+      "loss": 0.9447,
+      "step": 8170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957447956213378e-06,
+      "loss": 0.9983,
+      "step": 8171
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955290642644178e-06,
+      "loss": 0.9203,
+      "step": 8172
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95313337823116e-06,
+      "loss": 0.9855,
+      "step": 8173
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950976163075835e-06,
+      "loss": 1.008,
+      "step": 8174
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.948818997279698e-06,
+      "loss": 0.9261,
+      "step": 8175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946661880944255e-06,
+      "loss": 0.9946,
+      "step": 8176
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.944504814171004e-06,
+      "loss": 0.963,
+      "step": 8177
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.942347797061433e-06,
+      "loss": 0.9788,
+      "step": 8178
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940190829717043e-06,
+      "loss": 0.9105,
+      "step": 8179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938033912239319e-06,
+      "loss": 1.0212,
+      "step": 8180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935877044729752e-06,
+      "loss": 1.0598,
+      "step": 8181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933720227289831e-06,
+      "loss": 0.984,
+      "step": 8182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931563460021034e-06,
+      "loss": 0.9149,
+      "step": 8183
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929406743024846e-06,
+      "loss": 0.9266,
+      "step": 8184
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927250076402743e-06,
+      "loss": 0.9728,
+      "step": 8185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925093460256207e-06,
+      "loss": 0.8975,
+      "step": 8186
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922936894686706e-06,
+      "loss": 1.0153,
+      "step": 8187
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.920780379795714e-06,
+      "loss": 0.9897,
+      "step": 8188
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918623915684701e-06,
+      "loss": 1.0122,
+      "step": 8189
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.916467502455133e-06,
+      "loss": 0.9673,
+      "step": 8190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914311140208477e-06,
+      "loss": 0.9147,
+      "step": 8191
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912154829046196e-06,
+      "loss": 0.9961,
+      "step": 8192
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90999856906974e-06,
+      "loss": 0.9913,
+      "step": 8193
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90784236038058e-06,
+      "loss": 0.9696,
+      "step": 8194
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90568620308016e-06,
+      "loss": 0.9674,
+      "step": 8195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903530097269944e-06,
+      "loss": 0.8825,
+      "step": 8196
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901374043051373e-06,
+      "loss": 0.9748,
+      "step": 8197
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.899218040525894e-06,
+      "loss": 0.9328,
+      "step": 8198
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897062089794958e-06,
+      "loss": 0.9231,
+      "step": 8199
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89490619096001e-06,
+      "loss": 0.9105,
+      "step": 8200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.892750344122477e-06,
+      "loss": 1.0833,
+      "step": 8201
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890594549383812e-06,
+      "loss": 1.0181,
+      "step": 8202
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88843880684544e-06,
+      "loss": 0.9645,
+      "step": 8203
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886283116608805e-06,
+      "loss": 0.9689,
+      "step": 8204
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884127478775327e-06,
+      "loss": 0.9445,
+      "step": 8205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881971893446437e-06,
+      "loss": 0.9616,
+      "step": 8206
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879816360723566e-06,
+      "loss": 0.9782,
+      "step": 8207
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877660880708128e-06,
+      "loss": 0.9838,
+      "step": 8208
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875505453501552e-06,
+      "loss": 0.9131,
+      "step": 8209
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.873350079205257e-06,
+      "loss": 0.9721,
+      "step": 8210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.871194757920649e-06,
+      "loss": 0.9131,
+      "step": 8211
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869039489749149e-06,
+      "loss": 0.9622,
+      "step": 8212
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866884274792163e-06,
+      "loss": 0.9987,
+      "step": 8213
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864729113151106e-06,
+      "loss": 0.9496,
+      "step": 8214
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86257400492738e-06,
+      "loss": 0.9561,
+      "step": 8215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860418950222384e-06,
+      "loss": 1.0404,
+      "step": 8216
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858263949137528e-06,
+      "loss": 0.9664,
+      "step": 8217
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.856109001774198e-06,
+      "loss": 0.983,
+      "step": 8218
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853954108233803e-06,
+      "loss": 0.9988,
+      "step": 8219
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851799268617728e-06,
+      "loss": 1.0462,
+      "step": 8220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.849644483027363e-06,
+      "loss": 0.9453,
+      "step": 8221
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8474897515641e-06,
+      "loss": 0.9257,
+      "step": 8222
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845335074329322e-06,
+      "loss": 1.0047,
+      "step": 8223
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.843180451424414e-06,
+      "loss": 0.9434,
+      "step": 8224
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.841025882950758e-06,
+      "loss": 0.9935,
+      "step": 8225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838871369009722e-06,
+      "loss": 0.9441,
+      "step": 8226
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.836716909702696e-06,
+      "loss": 0.8944,
+      "step": 8227
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.834562505131035e-06,
+      "loss": 1.0196,
+      "step": 8228
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83240815539613e-06,
+      "loss": 0.9635,
+      "step": 8229
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830253860599332e-06,
+      "loss": 1.0639,
+      "step": 8230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.828099620842011e-06,
+      "loss": 0.9604,
+      "step": 8231
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825945436225532e-06,
+      "loss": 0.9592,
+      "step": 8232
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82379130685125e-06,
+      "loss": 1.0179,
+      "step": 8233
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821637232820528e-06,
+      "loss": 0.9248,
+      "step": 8234
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819483214234718e-06,
+      "loss": 0.9969,
+      "step": 8235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.817329251195168e-06,
+      "loss": 0.9068,
+      "step": 8236
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.815175343803233e-06,
+      "loss": 0.9068,
+      "step": 8237
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813021492160257e-06,
+      "loss": 1.0315,
+      "step": 8238
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810867696367584e-06,
+      "loss": 0.946,
+      "step": 8239
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808713956526555e-06,
+      "loss": 0.9753,
+      "step": 8240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80656027273851e-06,
+      "loss": 1.0201,
+      "step": 8241
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804406645104787e-06,
+      "loss": 0.9781,
+      "step": 8242
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80225307372672e-06,
+      "loss": 0.9956,
+      "step": 8243
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800099558705632e-06,
+      "loss": 0.8937,
+      "step": 8244
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79794610014286e-06,
+      "loss": 0.986,
+      "step": 8245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795792698139723e-06,
+      "loss": 0.9872,
+      "step": 8246
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79363935279755e-06,
+      "loss": 1.0129,
+      "step": 8247
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791486064217661e-06,
+      "loss": 0.9217,
+      "step": 8248
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.789332832501367e-06,
+      "loss": 0.9647,
+      "step": 8249
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78717965774999e-06,
+      "loss": 0.9859,
+      "step": 8250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785026540064833e-06,
+      "loss": 0.9604,
+      "step": 8251
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782873479547222e-06,
+      "loss": 0.9526,
+      "step": 8252
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78072047629845e-06,
+      "loss": 1.0042,
+      "step": 8253
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77856753041982e-06,
+      "loss": 0.9802,
+      "step": 8254
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.776414642012644e-06,
+      "loss": 1.0205,
+      "step": 8255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774261811178212e-06,
+      "loss": 0.9767,
+      "step": 8256
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.772109038017825e-06,
+      "loss": 1.0278,
+      "step": 8257
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.769956322632778e-06,
+      "loss": 1.0014,
+      "step": 8258
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76780366512435e-06,
+      "loss": 0.9923,
+      "step": 8259
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.765651065593846e-06,
+      "loss": 0.9846,
+      "step": 8260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763498524142536e-06,
+      "loss": 0.9547,
+      "step": 8261
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.761346040871713e-06,
+      "loss": 0.9734,
+      "step": 8262
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759193615882652e-06,
+      "loss": 0.9356,
+      "step": 8263
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757041249276628e-06,
+      "loss": 1.0497,
+      "step": 8264
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75488894115492e-06,
+      "loss": 1.0073,
+      "step": 8265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.752736691618795e-06,
+      "loss": 0.8892,
+      "step": 8266
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.750584500769526e-06,
+      "loss": 0.9289,
+      "step": 8267
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74843236870838e-06,
+      "loss": 0.9122,
+      "step": 8268
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74628029553661e-06,
+      "loss": 0.9254,
+      "step": 8269
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744128281355488e-06,
+      "loss": 0.9922,
+      "step": 8270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741976326266266e-06,
+      "loss": 0.9173,
+      "step": 8271
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739824430370201e-06,
+      "loss": 1.0468,
+      "step": 8272
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737672593768544e-06,
+      "loss": 0.9715,
+      "step": 8273
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735520816562542e-06,
+      "loss": 0.8881,
+      "step": 8274
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73336909885345e-06,
+      "loss": 0.9903,
+      "step": 8275
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731217440742502e-06,
+      "loss": 0.9737,
+      "step": 8276
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729065842330939e-06,
+      "loss": 0.9724,
+      "step": 8277
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726914303720007e-06,
+      "loss": 0.9095,
+      "step": 8278
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724762825010934e-06,
+      "loss": 0.9106,
+      "step": 8279
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722611406304956e-06,
+      "loss": 0.9448,
+      "step": 8280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720460047703303e-06,
+      "loss": 0.9731,
+      "step": 8281
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718308749307198e-06,
+      "loss": 0.98,
+      "step": 8282
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716157511217872e-06,
+      "loss": 1.0175,
+      "step": 8283
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714006333536536e-06,
+      "loss": 0.9429,
+      "step": 8284
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711855216364421e-06,
+      "loss": 0.9117,
+      "step": 8285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709704159802733e-06,
+      "loss": 0.9622,
+      "step": 8286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707553163952684e-06,
+      "loss": 0.9906,
+      "step": 8287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70540222891549e-06,
+      "loss": 1.0059,
+      "step": 8288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703251354792353e-06,
+      "loss": 0.9886,
+      "step": 8289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70110054168448e-06,
+      "loss": 0.9791,
+      "step": 8290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698949789693072e-06,
+      "loss": 0.9292,
+      "step": 8291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696799098919322e-06,
+      "loss": 0.9876,
+      "step": 8292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.694648469464433e-06,
+      "loss": 0.9395,
+      "step": 8293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692497901429592e-06,
+      "loss": 1.0017,
+      "step": 8294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69034739491599e-06,
+      "loss": 0.9313,
+      "step": 8295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688196950024817e-06,
+      "loss": 0.9866,
+      "step": 8296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68604656685725e-06,
+      "loss": 0.9806,
+      "step": 8297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683896245514477e-06,
+      "loss": 1.0293,
+      "step": 8298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68174598609767e-06,
+      "loss": 0.9255,
+      "step": 8299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67959578870801e-06,
+      "loss": 0.9388,
+      "step": 8300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677445653446668e-06,
+      "loss": 0.9674,
+      "step": 8301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675295580414807e-06,
+      "loss": 0.9545,
+      "step": 8302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6731455697136e-06,
+      "loss": 1.0192,
+      "step": 8303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670995621444206e-06,
+      "loss": 1.0182,
+      "step": 8304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668845735707789e-06,
+      "loss": 0.8741,
+      "step": 8305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666695912605506e-06,
+      "loss": 0.9046,
+      "step": 8306
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.664546152238508e-06,
+      "loss": 0.9263,
+      "step": 8307
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.662396454707953e-06,
+      "loss": 0.9256,
+      "step": 8308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660246820114984e-06,
+      "loss": 1.0167,
+      "step": 8309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658097248560745e-06,
+      "loss": 0.9593,
+      "step": 8310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655947740146386e-06,
+      "loss": 1.0226,
+      "step": 8311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65379829497304e-06,
+      "loss": 0.9268,
+      "step": 8312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65164891314185e-06,
+      "loss": 0.9807,
+      "step": 8313
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649499594753947e-06,
+      "loss": 0.9459,
+      "step": 8314
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647350339910455e-06,
+      "loss": 1.041,
+      "step": 8315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645201148712515e-06,
+      "loss": 0.9744,
+      "step": 8316
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643052021261237e-06,
+      "loss": 0.8899,
+      "step": 8317
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64090295765776e-06,
+      "loss": 0.9692,
+      "step": 8318
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63875395800319e-06,
+      "loss": 0.9021,
+      "step": 8319
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636605022398643e-06,
+      "loss": 0.939,
+      "step": 8320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634456150945238e-06,
+      "loss": 0.896,
+      "step": 8321
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632307343744079e-06,
+      "loss": 0.9587,
+      "step": 8322
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.630158600896278e-06,
+      "loss": 0.951,
+      "step": 8323
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62800992250294e-06,
+      "loss": 0.9594,
+      "step": 8324
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625861308665155e-06,
+      "loss": 1.0696,
+      "step": 8325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.623712759484033e-06,
+      "loss": 0.9364,
+      "step": 8326
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62156427506066e-06,
+      "loss": 0.9927,
+      "step": 8327
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.619415855496133e-06,
+      "loss": 1.0024,
+      "step": 8328
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.617267500891537e-06,
+      "loss": 1.0121,
+      "step": 8329
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615119211347958e-06,
+      "loss": 1.0283,
+      "step": 8330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61297098696648e-06,
+      "loss": 1.0056,
+      "step": 8331
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61082282784818e-06,
+      "loss": 1.0062,
+      "step": 8332
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60867473409414e-06,
+      "loss": 0.9951,
+      "step": 8333
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606526705805427e-06,
+      "loss": 0.9449,
+      "step": 8334
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604378743083112e-06,
+      "loss": 0.891,
+      "step": 8335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602230846028264e-06,
+      "loss": 0.9559,
+      "step": 8336
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600083014741942e-06,
+      "loss": 0.9318,
+      "step": 8337
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597935249325215e-06,
+      "loss": 0.912,
+      "step": 8338
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595787549879137e-06,
+      "loss": 1.0013,
+      "step": 8339
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.593639916504759e-06,
+      "loss": 0.9366,
+      "step": 8340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591492349303139e-06,
+      "loss": 1.0204,
+      "step": 8341
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589344848375316e-06,
+      "loss": 0.9642,
+      "step": 8342
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.587197413822349e-06,
+      "loss": 0.9418,
+      "step": 8343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58505004574527e-06,
+      "loss": 0.9823,
+      "step": 8344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.582902744245117e-06,
+      "loss": 0.9912,
+      "step": 8345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580755509422932e-06,
+      "loss": 0.9479,
+      "step": 8346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578608341379748e-06,
+      "loss": 0.9676,
+      "step": 8347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.576461240216585e-06,
+      "loss": 0.9207,
+      "step": 8348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574314206034483e-06,
+      "loss": 0.91,
+      "step": 8349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.572167238934452e-06,
+      "loss": 0.9272,
+      "step": 8350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570020339017525e-06,
+      "loss": 0.9353,
+      "step": 8351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567873506384711e-06,
+      "loss": 0.9328,
+      "step": 8352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565726741137023e-06,
+      "loss": 0.972,
+      "step": 8353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563580043375477e-06,
+      "loss": 0.9004,
+      "step": 8354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561433413201075e-06,
+      "loss": 0.8933,
+      "step": 8355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.559286850714825e-06,
+      "loss": 1.0762,
+      "step": 8356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557140356017732e-06,
+      "loss": 1.0286,
+      "step": 8357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554993929210782e-06,
+      "loss": 0.9659,
+      "step": 8358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55284757039498e-06,
+      "loss": 0.9667,
+      "step": 8359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550701279671314e-06,
+      "loss": 0.944,
+      "step": 8360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.548555057140773e-06,
+      "loss": 0.9104,
+      "step": 8361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546408902904342e-06,
+      "loss": 0.9568,
+      "step": 8362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544262817063002e-06,
+      "loss": 0.9127,
+      "step": 8363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542116799717732e-06,
+      "loss": 1.0506,
+      "step": 8364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539970850969508e-06,
+      "loss": 0.9798,
+      "step": 8365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537824970919306e-06,
+      "loss": 0.9606,
+      "step": 8366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535679159668089e-06,
+      "loss": 0.9875,
+      "step": 8367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.533533417316823e-06,
+      "loss": 0.9601,
+      "step": 8368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531387743966475e-06,
+      "loss": 1.0529,
+      "step": 8369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529242139718e-06,
+      "loss": 0.9494,
+      "step": 8370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.527096604672356e-06,
+      "loss": 0.9945,
+      "step": 8371
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5249511389305e-06,
+      "loss": 0.9664,
+      "step": 8372
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52280574259337e-06,
+      "loss": 0.9331,
+      "step": 8373
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520660415761924e-06,
+      "loss": 0.9038,
+      "step": 8374
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518515158537096e-06,
+      "loss": 0.9944,
+      "step": 8375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516369971019837e-06,
+      "loss": 0.915,
+      "step": 8376
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514224853311072e-06,
+      "loss": 0.8533,
+      "step": 8377
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51207980551174e-06,
+      "loss": 0.9091,
+      "step": 8378
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.509934827722769e-06,
+      "loss": 1.0104,
+      "step": 8379
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507789920045085e-06,
+      "loss": 1.0364,
+      "step": 8380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505645082579614e-06,
+      "loss": 0.9891,
+      "step": 8381
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503500315427276e-06,
+      "loss": 0.9598,
+      "step": 8382
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.501355618688982e-06,
+      "loss": 0.9814,
+      "step": 8383
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.499210992465654e-06,
+      "loss": 0.9145,
+      "step": 8384
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497066436858195e-06,
+      "loss": 0.9689,
+      "step": 8385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494921951967512e-06,
+      "loss": 0.9391,
+      "step": 8386
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492777537894514e-06,
+      "loss": 1.0199,
+      "step": 8387
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.490633194740093e-06,
+      "loss": 1.041,
+      "step": 8388
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488488922605153e-06,
+      "loss": 0.9334,
+      "step": 8389
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486344721590587e-06,
+      "loss": 0.9817,
+      "step": 8390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48420059179728e-06,
+      "loss": 0.936,
+      "step": 8391
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48205653332612e-06,
+      "loss": 0.9547,
+      "step": 8392
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479912546277991e-06,
+      "loss": 1.0167,
+      "step": 8393
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.477768630753777e-06,
+      "loss": 0.979,
+      "step": 8394
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47562478685435e-06,
+      "loss": 0.9916,
+      "step": 8395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473481014680581e-06,
+      "loss": 0.9512,
+      "step": 8396
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471337314333347e-06,
+      "loss": 0.9568,
+      "step": 8397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469193685913507e-06,
+      "loss": 0.9797,
+      "step": 8398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.467050129521933e-06,
+      "loss": 0.9642,
+      "step": 8399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464906645259476e-06,
+      "loss": 1.0022,
+      "step": 8400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.462763233226993e-06,
+      "loss": 1.0189,
+      "step": 8401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460619893525344e-06,
+      "loss": 0.9065,
+      "step": 8402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45847662625537e-06,
+      "loss": 0.9434,
+      "step": 8403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.456333431517922e-06,
+      "loss": 0.9577,
+      "step": 8404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.454190309413845e-06,
+      "loss": 0.9691,
+      "step": 8405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45204726004397e-06,
+      "loss": 0.9399,
+      "step": 8406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449904283509142e-06,
+      "loss": 0.9637,
+      "step": 8407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447761379910185e-06,
+      "loss": 1.0505,
+      "step": 8408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445618549347938e-06,
+      "loss": 0.9341,
+      "step": 8409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443475791923216e-06,
+      "loss": 1.0373,
+      "step": 8410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441333107736847e-06,
+      "loss": 0.9708,
+      "step": 8411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439190496889649e-06,
+      "loss": 0.9309,
+      "step": 8412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437047959482433e-06,
+      "loss": 0.9425,
+      "step": 8413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43490549561602e-06,
+      "loss": 1.0216,
+      "step": 8414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.432763105391211e-06,
+      "loss": 0.9728,
+      "step": 8415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43062078890881e-06,
+      "loss": 0.992,
+      "step": 8416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.428478546269622e-06,
+      "loss": 0.9481,
+      "step": 8417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426336377574443e-06,
+      "loss": 0.9751,
+      "step": 8418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42419428292407e-06,
+      "loss": 0.9866,
+      "step": 8419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42205226241929e-06,
+      "loss": 0.921,
+      "step": 8420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.419910316160892e-06,
+      "loss": 0.9462,
+      "step": 8421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417768444249662e-06,
+      "loss": 1.0145,
+      "step": 8422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41562664678638e-06,
+      "loss": 0.968,
+      "step": 8423
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413484923871817e-06,
+      "loss": 0.9735,
+      "step": 8424
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.411343275606755e-06,
+      "loss": 1.0075,
+      "step": 8425
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409201702091958e-06,
+      "loss": 0.9776,
+      "step": 8426
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407060203428195e-06,
+      "loss": 0.9094,
+      "step": 8427
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.404918779716228e-06,
+      "loss": 0.9593,
+      "step": 8428
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402777431056814e-06,
+      "loss": 0.8667,
+      "step": 8429
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400636157550717e-06,
+      "loss": 0.9888,
+      "step": 8430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398494959298675e-06,
+      "loss": 1.0285,
+      "step": 8431
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396353836401453e-06,
+      "loss": 0.9241,
+      "step": 8432
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394212788959787e-06,
+      "loss": 0.9914,
+      "step": 8433
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392071817074418e-06,
+      "loss": 1.0065,
+      "step": 8434
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389930920846088e-06,
+      "loss": 1.0061,
+      "step": 8435
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.387790100375526e-06,
+      "loss": 0.8836,
+      "step": 8436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385649355763469e-06,
+      "loss": 0.9642,
+      "step": 8437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383508687110646e-06,
+      "loss": 0.9686,
+      "step": 8438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.38136809451777e-06,
+      "loss": 0.9638,
+      "step": 8439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379227578085573e-06,
+      "loss": 0.884,
+      "step": 8440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.377087137914764e-06,
+      "loss": 1.0141,
+      "step": 8441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374946774106061e-06,
+      "loss": 1.0473,
+      "step": 8442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37280648676017e-06,
+      "loss": 1.0205,
+      "step": 8443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370666275977796e-06,
+      "loss": 1.0059,
+      "step": 8444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368526141859647e-06,
+      "loss": 0.9998,
+      "step": 8445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366386084506414e-06,
+      "loss": 0.9621,
+      "step": 8446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364246104018798e-06,
+      "loss": 1.0633,
+      "step": 8447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.362106200497492e-06,
+      "loss": 0.9828,
+      "step": 8448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359966374043176e-06,
+      "loss": 0.9066,
+      "step": 8449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35782662475654e-06,
+      "loss": 0.991,
+      "step": 8450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35568695273826e-06,
+      "loss": 0.9625,
+      "step": 8451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353547358089019e-06,
+      "loss": 0.9652,
+      "step": 8452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.351407840909487e-06,
+      "loss": 0.9782,
+      "step": 8453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349268401300332e-06,
+      "loss": 0.9102,
+      "step": 8454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347129039362227e-06,
+      "loss": 0.9977,
+      "step": 8455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34498975519582e-06,
+      "loss": 1.0171,
+      "step": 8456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342850548901788e-06,
+      "loss": 1.0019,
+      "step": 8457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340711420580772e-06,
+      "loss": 0.9193,
+      "step": 8458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338572370333428e-06,
+      "loss": 1.0362,
+      "step": 8459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336433398260405e-06,
+      "loss": 1.0353,
+      "step": 8460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.334294504462347e-06,
+      "loss": 0.9911,
+      "step": 8461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33215568903989e-06,
+      "loss": 0.9979,
+      "step": 8462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330016952093676e-06,
+      "loss": 0.9419,
+      "step": 8463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.327878293724331e-06,
+      "loss": 0.9646,
+      "step": 8464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.325739714032495e-06,
+      "loss": 0.9351,
+      "step": 8465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.323601213118786e-06,
+      "loss": 0.9689,
+      "step": 8466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321462791083823e-06,
+      "loss": 0.9219,
+      "step": 8467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319324448028232e-06,
+      "loss": 0.9534,
+      "step": 8468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317186184052622e-06,
+      "loss": 0.9836,
+      "step": 8469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315047999257607e-06,
+      "loss": 0.9526,
+      "step": 8470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312909893743793e-06,
+      "loss": 0.9484,
+      "step": 8471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310771867611776e-06,
+      "loss": 0.984,
+      "step": 8472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30863392096217e-06,
+      "loss": 1.0019,
+      "step": 8473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306496053895556e-06,
+      "loss": 1.0446,
+      "step": 8474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304358266512537e-06,
+      "loss": 0.9598,
+      "step": 8475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302220558913695e-06,
+      "loss": 0.8744,
+      "step": 8476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300082931199616e-06,
+      "loss": 0.9498,
+      "step": 8477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29794538347088e-06,
+      "loss": 0.9549,
+      "step": 8478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295807915828066e-06,
+      "loss": 0.9553,
+      "step": 8479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293670528371747e-06,
+      "loss": 0.9877,
+      "step": 8480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291533221202494e-06,
+      "loss": 0.95,
+      "step": 8481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289395994420866e-06,
+      "loss": 0.8771,
+      "step": 8482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28725884812743e-06,
+      "loss": 0.967,
+      "step": 8483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285121782422743e-06,
+      "loss": 0.9391,
+      "step": 8484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28298479740736e-06,
+      "loss": 0.93,
+      "step": 8485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280847893181832e-06,
+      "loss": 0.9941,
+      "step": 8486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2787110698467e-06,
+      "loss": 0.9415,
+      "step": 8487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27657432750252e-06,
+      "loss": 0.9662,
+      "step": 8488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274437666249812e-06,
+      "loss": 0.95,
+      "step": 8489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272301086189131e-06,
+      "loss": 1.0259,
+      "step": 8490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.270164587420995e-06,
+      "loss": 0.9259,
+      "step": 8491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268028170045934e-06,
+      "loss": 0.9148,
+      "step": 8492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.265891834164477e-06,
+      "loss": 0.8923,
+      "step": 8493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.263755579877137e-06,
+      "loss": 0.9885,
+      "step": 8494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261619407284438e-06,
+      "loss": 1.0345,
+      "step": 8495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.259483316486888e-06,
+      "loss": 1.0259,
+      "step": 8496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257347307584988e-06,
+      "loss": 1.015,
+      "step": 8497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255211380679258e-06,
+      "loss": 0.9399,
+      "step": 8498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253075535870189e-06,
+      "loss": 0.9854,
+      "step": 8499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250939773258276e-06,
+      "loss": 1.0123,
+      "step": 8500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248804092944018e-06,
+      "loss": 0.9692,
+      "step": 8501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246668495027899e-06,
+      "loss": 0.9259,
+      "step": 8502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244532979610408e-06,
+      "loss": 1.0013,
+      "step": 8503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.242397546792029e-06,
+      "loss": 0.9692,
+      "step": 8504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240262196673228e-06,
+      "loss": 0.9326,
+      "step": 8505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238126929354491e-06,
+      "loss": 0.9574,
+      "step": 8506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23599174493628e-06,
+      "loss": 1.0228,
+      "step": 8507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233856643519066e-06,
+      "loss": 0.9108,
+      "step": 8508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.231721625203308e-06,
+      "loss": 0.9852,
+      "step": 8509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22958669008946e-06,
+      "loss": 0.9517,
+      "step": 8510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227451838277984e-06,
+      "loss": 0.9376,
+      "step": 8511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225317069869324e-06,
+      "loss": 0.8982,
+      "step": 8512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223182384963934e-06,
+      "loss": 0.983,
+      "step": 8513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221047783662247e-06,
+      "loss": 0.9829,
+      "step": 8514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218913266064703e-06,
+      "loss": 0.9581,
+      "step": 8515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216778832271742e-06,
+      "loss": 0.9853,
+      "step": 8516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214644482383788e-06,
+      "loss": 0.9189,
+      "step": 8517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.212510216501271e-06,
+      "loss": 0.9584,
+      "step": 8518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210376034724615e-06,
+      "loss": 0.9478,
+      "step": 8519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208241937154233e-06,
+      "loss": 1.0467,
+      "step": 8520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20610792389055e-06,
+      "loss": 0.9401,
+      "step": 8521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20397399503396e-06,
+      "loss": 0.9989,
+      "step": 8522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.201840150684887e-06,
+      "loss": 0.9984,
+      "step": 8523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199706390943725e-06,
+      "loss": 1.021,
+      "step": 8524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197572715910872e-06,
+      "loss": 0.9027,
+      "step": 8525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195439125686726e-06,
+      "loss": 0.9199,
+      "step": 8526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193305620371676e-06,
+      "loss": 0.9006,
+      "step": 8527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.191172200066108e-06,
+      "loss": 1.0532,
+      "step": 8528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18903886487041e-06,
+      "loss": 0.9658,
+      "step": 8529
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186905614884952e-06,
+      "loss": 0.9485,
+      "step": 8530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184772450210117e-06,
+      "loss": 0.9371,
+      "step": 8531
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182639370946268e-06,
+      "loss": 0.9497,
+      "step": 8532
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18050637719378e-06,
+      "loss": 0.9226,
+      "step": 8533
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17837346905301e-06,
+      "loss": 1.0807,
+      "step": 8534
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.176240646624318e-06,
+      "loss": 1.0002,
+      "step": 8535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174107910008059e-06,
+      "loss": 0.9414,
+      "step": 8536
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171975259304585e-06,
+      "loss": 0.9858,
+      "step": 8537
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.169842694614238e-06,
+      "loss": 0.9491,
+      "step": 8538
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.167710216037366e-06,
+      "loss": 0.9237,
+      "step": 8539
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165577823674302e-06,
+      "loss": 0.993,
+      "step": 8540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163445517625385e-06,
+      "loss": 0.9462,
+      "step": 8541
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.161313297990944e-06,
+      "loss": 1.049,
+      "step": 8542
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159181164871303e-06,
+      "loss": 0.9009,
+      "step": 8543
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157049118366787e-06,
+      "loss": 0.9742,
+      "step": 8544
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154917158577712e-06,
+      "loss": 0.9299,
+      "step": 8545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.152785285604397e-06,
+      "loss": 0.9577,
+      "step": 8546
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.150653499547148e-06,
+      "loss": 0.9555,
+      "step": 8547
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148521800506267e-06,
+      "loss": 0.9477,
+      "step": 8548
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14639018858206e-06,
+      "loss": 1.0018,
+      "step": 8549
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144258663874826e-06,
+      "loss": 0.9585,
+      "step": 8550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142127226484857e-06,
+      "loss": 0.9115,
+      "step": 8551
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139995876512445e-06,
+      "loss": 1.0336,
+      "step": 8552
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137864614057867e-06,
+      "loss": 0.9456,
+      "step": 8553
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135733439221416e-06,
+      "loss": 0.9997,
+      "step": 8554
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133602352103356e-06,
+      "loss": 0.9762,
+      "step": 8555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131471352803975e-06,
+      "loss": 0.9252,
+      "step": 8556
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12934044142353e-06,
+      "loss": 1.009,
+      "step": 8557
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127209618062288e-06,
+      "loss": 0.9749,
+      "step": 8558
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125078882820514e-06,
+      "loss": 0.9898,
+      "step": 8559
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12294823579846e-06,
+      "loss": 0.9525,
+      "step": 8560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12081767709638e-06,
+      "loss": 0.9405,
+      "step": 8561
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.118687206814526e-06,
+      "loss": 0.9464,
+      "step": 8562
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116556825053131e-06,
+      "loss": 0.893,
+      "step": 8563
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.114426531912446e-06,
+      "loss": 0.9975,
+      "step": 8564
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1122963274927e-06,
+      "loss": 0.9325,
+      "step": 8565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.110166211894129e-06,
+      "loss": 0.9944,
+      "step": 8566
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108036185216957e-06,
+      "loss": 0.9462,
+      "step": 8567
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105906247561408e-06,
+      "loss": 0.9762,
+      "step": 8568
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1037763990277e-06,
+      "loss": 0.928,
+      "step": 8569
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10164663971605e-06,
+      "loss": 1.0106,
+      "step": 8570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.099516969726669e-06,
+      "loss": 0.9343,
+      "step": 8571
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097387389159761e-06,
+      "loss": 1.0177,
+      "step": 8572
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095257898115525e-06,
+      "loss": 0.9666,
+      "step": 8573
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093128496694164e-06,
+      "loss": 0.9943,
+      "step": 8574
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.090999184995872e-06,
+      "loss": 0.9776,
+      "step": 8575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088869963120834e-06,
+      "loss": 0.8831,
+      "step": 8576
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.08674083116924e-06,
+      "loss": 0.9921,
+      "step": 8577
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084611789241264e-06,
+      "loss": 0.9834,
+      "step": 8578
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082482837437094e-06,
+      "loss": 0.9049,
+      "step": 8579
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080353975856893e-06,
+      "loss": 0.8776,
+      "step": 8580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07822520460083e-06,
+      "loss": 0.9185,
+      "step": 8581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076096523769074e-06,
+      "loss": 0.9724,
+      "step": 8582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07396793346178e-06,
+      "loss": 1.0024,
+      "step": 8583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071839433779106e-06,
+      "loss": 0.8879,
+      "step": 8584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.069711024821205e-06,
+      "loss": 0.9711,
+      "step": 8585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067582706688217e-06,
+      "loss": 0.9797,
+      "step": 8586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065454479480295e-06,
+      "loss": 1.0074,
+      "step": 8587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.063326343297563e-06,
+      "loss": 0.9499,
+      "step": 8588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.061198298240173e-06,
+      "loss": 1.0067,
+      "step": 8589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059070344408243e-06,
+      "loss": 0.9515,
+      "step": 8590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.056942481901898e-06,
+      "loss": 0.9494,
+      "step": 8591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054814710821267e-06,
+      "loss": 0.9801,
+      "step": 8592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052687031266457e-06,
+      "loss": 0.8616,
+      "step": 8593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05055944333759e-06,
+      "loss": 0.9073,
+      "step": 8594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048431947134771e-06,
+      "loss": 0.9747,
+      "step": 8595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046304542758101e-06,
+      "loss": 0.9924,
+      "step": 8596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044177230307684e-06,
+      "loss": 0.9713,
+      "step": 8597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042050009883611e-06,
+      "loss": 0.9454,
+      "step": 8598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039922881585977e-06,
+      "loss": 0.9222,
+      "step": 8599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.037795845514867e-06,
+      "loss": 0.9056,
+      "step": 8600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035668901770362e-06,
+      "loss": 0.9715,
+      "step": 8601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033542050452544e-06,
+      "loss": 0.9142,
+      "step": 8602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03141529166148e-06,
+      "loss": 0.9317,
+      "step": 8603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02928862549725e-06,
+      "loss": 0.9628,
+      "step": 8604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02716205205991e-06,
+      "loss": 1.0278,
+      "step": 8605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02503557144952e-06,
+      "loss": 0.9976,
+      "step": 8606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022909183766143e-06,
+      "loss": 0.9791,
+      "step": 8607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020782889109822e-06,
+      "loss": 1.0111,
+      "step": 8608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.018656687580614e-06,
+      "loss": 0.9636,
+      "step": 8609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016530579278559e-06,
+      "loss": 0.974,
+      "step": 8610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01440456430369e-06,
+      "loss": 1.0002,
+      "step": 8611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01227864275605e-06,
+      "loss": 0.9351,
+      "step": 8612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010152814735663e-06,
+      "loss": 0.9105,
+      "step": 8613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008027080342555e-06,
+      "loss": 1.0165,
+      "step": 8614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00590143967675e-06,
+      "loss": 0.9318,
+      "step": 8615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003775892838259e-06,
+      "loss": 0.9808,
+      "step": 8616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.001650439927102e-06,
+      "loss": 1.0423,
+      "step": 8617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.999525081043284e-06,
+      "loss": 0.9656,
+      "step": 8618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997399816286801e-06,
+      "loss": 0.9244,
+      "step": 8619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995274645757666e-06,
+      "loss": 1.0715,
+      "step": 8620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.993149569555864e-06,
+      "loss": 1.0014,
+      "step": 8621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991024587781386e-06,
+      "loss": 0.9191,
+      "step": 8622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988899700534221e-06,
+      "loss": 0.9527,
+      "step": 8623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986774907914345e-06,
+      "loss": 1.0074,
+      "step": 8624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984650210021741e-06,
+      "loss": 0.944,
+      "step": 8625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.982525606956377e-06,
+      "loss": 0.9137,
+      "step": 8626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980401098818223e-06,
+      "loss": 0.9247,
+      "step": 8627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978276685707244e-06,
+      "loss": 0.9926,
+      "step": 8628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976152367723393e-06,
+      "loss": 0.9907,
+      "step": 8629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97402814496663e-06,
+      "loss": 0.9701,
+      "step": 8630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9719040175369e-06,
+      "loss": 1.03,
+      "step": 8631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969779985534154e-06,
+      "loss": 0.9177,
+      "step": 8632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.967656049058332e-06,
+      "loss": 0.9463,
+      "step": 8633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965532208209366e-06,
+      "loss": 0.9137,
+      "step": 8634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963408463087194e-06,
+      "loss": 1.0166,
+      "step": 8635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.961284813791736e-06,
+      "loss": 0.9506,
+      "step": 8636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.959161260422926e-06,
+      "loss": 0.9724,
+      "step": 8637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957037803080673e-06,
+      "loss": 0.9444,
+      "step": 8638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954914441864894e-06,
+      "loss": 0.9761,
+      "step": 8639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952791176875499e-06,
+      "loss": 0.9513,
+      "step": 8640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950668008212392e-06,
+      "loss": 0.9281,
+      "step": 8641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948544935975475e-06,
+      "loss": 1.0027,
+      "step": 8642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.946421960264643e-06,
+      "loss": 1.0346,
+      "step": 8643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944299081179783e-06,
+      "loss": 1.0057,
+      "step": 8644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942176298820792e-06,
+      "loss": 0.8593,
+      "step": 8645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940053613287538e-06,
+      "loss": 0.9061,
+      "step": 8646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937931024679915e-06,
+      "loss": 0.9805,
+      "step": 8647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.935808533097785e-06,
+      "loss": 0.9475,
+      "step": 8648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933686138641016e-06,
+      "loss": 0.8635,
+      "step": 8649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93156384140948e-06,
+      "loss": 1.0226,
+      "step": 8650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929441641503031e-06,
+      "loss": 1.0155,
+      "step": 8651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.927319539021518e-06,
+      "loss": 0.9288,
+      "step": 8652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925197534064805e-06,
+      "loss": 1.018,
+      "step": 8653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.923075626732727e-06,
+      "loss": 0.9254,
+      "step": 8654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92095381712513e-06,
+      "loss": 0.9836,
+      "step": 8655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918832105341849e-06,
+      "loss": 0.915,
+      "step": 8656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916710491482713e-06,
+      "loss": 0.9395,
+      "step": 8657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.914588975647554e-06,
+      "loss": 0.968,
+      "step": 8658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91246755793619e-06,
+      "loss": 0.9585,
+      "step": 8659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910346238448443e-06,
+      "loss": 0.9896,
+      "step": 8660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908225017284129e-06,
+      "loss": 0.9988,
+      "step": 8661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.906103894543048e-06,
+      "loss": 0.9624,
+      "step": 8662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.903982870325008e-06,
+      "loss": 0.9966,
+      "step": 8663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90186194472981e-06,
+      "loss": 0.9762,
+      "step": 8664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899741117857249e-06,
+      "loss": 0.9714,
+      "step": 8665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.897620389807115e-06,
+      "loss": 0.9729,
+      "step": 8666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89549976067919e-06,
+      "loss": 1.0181,
+      "step": 8667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893379230573263e-06,
+      "loss": 0.9673,
+      "step": 8668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891258799589097e-06,
+      "loss": 0.9796,
+      "step": 8669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.889138467826479e-06,
+      "loss": 0.9787,
+      "step": 8670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887018235385167e-06,
+      "loss": 0.9102,
+      "step": 8671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88489810236492e-06,
+      "loss": 0.9967,
+      "step": 8672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882778068865505e-06,
+      "loss": 0.9492,
+      "step": 8673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880658134986666e-06,
+      "loss": 1.0649,
+      "step": 8674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878538300828159e-06,
+      "loss": 0.9381,
+      "step": 8675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.876418566489725e-06,
+      "loss": 0.9921,
+      "step": 8676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.874298932071096e-06,
+      "loss": 0.9502,
+      "step": 8677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872179397672018e-06,
+      "loss": 0.9121,
+      "step": 8678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870059963392208e-06,
+      "loss": 0.9435,
+      "step": 8679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8679406293314e-06,
+      "loss": 0.9824,
+      "step": 8680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86582139558931e-06,
+      "loss": 0.9186,
+      "step": 8681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.863702262265654e-06,
+      "loss": 0.9923,
+      "step": 8682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86158322946014e-06,
+      "loss": 0.9717,
+      "step": 8683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859464297272484e-06,
+      "loss": 0.9298,
+      "step": 8684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.857345465802369e-06,
+      "loss": 0.9308,
+      "step": 8685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85522673514951e-06,
+      "loss": 0.9468,
+      "step": 8686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853108105413585e-06,
+      "loss": 0.906,
+      "step": 8687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850989576694288e-06,
+      "loss": 1.021,
+      "step": 8688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8488711490913e-06,
+      "loss": 0.9424,
+      "step": 8689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846752822704294e-06,
+      "loss": 1.0586,
+      "step": 8690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844634597632948e-06,
+      "loss": 0.9849,
+      "step": 8691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.842516473976927e-06,
+      "loss": 0.9445,
+      "step": 8692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840398451835897e-06,
+      "loss": 0.9022,
+      "step": 8693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.838280531309513e-06,
+      "loss": 0.886,
+      "step": 8694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.836162712497428e-06,
+      "loss": 0.9825,
+      "step": 8695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.834044995499292e-06,
+      "loss": 1.0096,
+      "step": 8696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831927380414749e-06,
+      "loss": 0.9912,
+      "step": 8697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829809867343439e-06,
+      "loss": 0.9123,
+      "step": 8698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827692456384995e-06,
+      "loss": 1.0256,
+      "step": 8699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825575147639046e-06,
+      "loss": 1.07,
+      "step": 8700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82345794120522e-06,
+      "loss": 0.9567,
+      "step": 8701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821340837183128e-06,
+      "loss": 1.0373,
+      "step": 8702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.819223835672398e-06,
+      "loss": 0.9747,
+      "step": 8703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817106936772632e-06,
+      "loss": 0.9072,
+      "step": 8704
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814990140583435e-06,
+      "loss": 0.9687,
+      "step": 8705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.812873447204411e-06,
+      "loss": 1.0295,
+      "step": 8706
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810756856735152e-06,
+      "loss": 0.9385,
+      "step": 8707
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808640369275253e-06,
+      "loss": 0.9553,
+      "step": 8708
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806523984924299e-06,
+      "loss": 0.8301,
+      "step": 8709
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804407703781866e-06,
+      "loss": 0.9421,
+      "step": 8710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80229152594754e-06,
+      "loss": 0.9621,
+      "step": 8711
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800175451520882e-06,
+      "loss": 0.95,
+      "step": 8712
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.798059480601467e-06,
+      "loss": 0.9657,
+      "step": 8713
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.795943613288853e-06,
+      "loss": 0.9308,
+      "step": 8714
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793827849682594e-06,
+      "loss": 0.9242,
+      "step": 8715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791712189882249e-06,
+      "loss": 0.9686,
+      "step": 8716
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789596633987356e-06,
+      "loss": 1.012,
+      "step": 8717
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.787481182097468e-06,
+      "loss": 0.9425,
+      "step": 8718
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785365834312116e-06,
+      "loss": 0.8837,
+      "step": 8719
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783250590730829e-06,
+      "loss": 0.9734,
+      "step": 8720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781135451453141e-06,
+      "loss": 1.0057,
+      "step": 8721
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779020416578573e-06,
+      "loss": 0.9998,
+      "step": 8722
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77690548620664e-06,
+      "loss": 0.917,
+      "step": 8723
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774790660436857e-06,
+      "loss": 0.9405,
+      "step": 8724
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77267593936873e-06,
+      "loss": 0.967,
+      "step": 8725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770561323101768e-06,
+      "loss": 1.0334,
+      "step": 8726
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768446811735462e-06,
+      "loss": 0.9261,
+      "step": 8727
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.766332405369304e-06,
+      "loss": 0.9903,
+      "step": 8728
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76421810410279e-06,
+      "loss": 0.9084,
+      "step": 8729
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762103908035394e-06,
+      "loss": 0.9989,
+      "step": 8730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.759989817266602e-06,
+      "loss": 0.9925,
+      "step": 8731
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.757875831895886e-06,
+      "loss": 1.0017,
+      "step": 8732
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755761952022706e-06,
+      "loss": 0.9198,
+      "step": 8733
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.753648177746538e-06,
+      "loss": 1.02,
+      "step": 8734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751534509166827e-06,
+      "loss": 0.9734,
+      "step": 8735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.749420946383041e-06,
+      "loss": 0.8596,
+      "step": 8736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747307489494618e-06,
+      "loss": 0.9949,
+      "step": 8737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745194138601003e-06,
+      "loss": 0.9651,
+      "step": 8738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743080893801635e-06,
+      "loss": 0.9419,
+      "step": 8739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740967755195948e-06,
+      "loss": 0.9828,
+      "step": 8740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738854722883373e-06,
+      "loss": 0.9187,
+      "step": 8741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736741796963333e-06,
+      "loss": 0.9573,
+      "step": 8742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734628977535238e-06,
+      "loss": 1.078,
+      "step": 8743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732516264698512e-06,
+      "loss": 0.9464,
+      "step": 8744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.730403658552558e-06,
+      "loss": 0.948,
+      "step": 8745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.728291159196782e-06,
+      "loss": 1.0183,
+      "step": 8746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72617876673058e-06,
+      "loss": 0.8457,
+      "step": 8747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724066481253344e-06,
+      "loss": 0.9853,
+      "step": 8748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72195430286447e-06,
+      "loss": 0.9844,
+      "step": 8749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71984223166333e-06,
+      "loss": 0.9027,
+      "step": 8750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717730267749315e-06,
+      "loss": 0.9651,
+      "step": 8751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715618411221787e-06,
+      "loss": 0.8867,
+      "step": 8752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713506662180119e-06,
+      "loss": 0.9284,
+      "step": 8753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711395020723674e-06,
+      "loss": 0.9869,
+      "step": 8754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709283486951809e-06,
+      "loss": 1.005,
+      "step": 8755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707172060963878e-06,
+      "loss": 1.0113,
+      "step": 8756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705060742859229e-06,
+      "loss": 0.9647,
+      "step": 8757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702949532737203e-06,
+      "loss": 0.8885,
+      "step": 8758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700838430697141e-06,
+      "loss": 0.9028,
+      "step": 8759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698727436838373e-06,
+      "loss": 0.8949,
+      "step": 8760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.696616551260223e-06,
+      "loss": 0.9501,
+      "step": 8761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694505774062022e-06,
+      "loss": 0.9953,
+      "step": 8762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69239510534308e-06,
+      "loss": 0.9485,
+      "step": 8763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690284545202714e-06,
+      "loss": 0.9718,
+      "step": 8764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688174093740231e-06,
+      "loss": 0.9232,
+      "step": 8765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686063751054924e-06,
+      "loss": 0.9689,
+      "step": 8766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683953517246105e-06,
+      "loss": 1.0275,
+      "step": 8767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68184339241305e-06,
+      "loss": 0.978,
+      "step": 8768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679733376655061e-06,
+      "loss": 0.9725,
+      "step": 8769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677623470071409e-06,
+      "loss": 0.9749,
+      "step": 8770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675513672761372e-06,
+      "loss": 0.9568,
+      "step": 8771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673403984824225e-06,
+      "loss": 1.0119,
+      "step": 8772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67129440635923e-06,
+      "loss": 1.0217,
+      "step": 8773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66918493746565e-06,
+      "loss": 0.8975,
+      "step": 8774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667075578242744e-06,
+      "loss": 0.9668,
+      "step": 8775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664966328789754e-06,
+      "loss": 0.9982,
+      "step": 8776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662857189205932e-06,
+      "loss": 0.9345,
+      "step": 8777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.660748159590517e-06,
+      "loss": 0.9294,
+      "step": 8778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658639240042744e-06,
+      "loss": 1.0088,
+      "step": 8779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656530430661843e-06,
+      "loss": 0.9696,
+      "step": 8780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654421731547037e-06,
+      "loss": 0.9332,
+      "step": 8781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.652313142797548e-06,
+      "loss": 0.9568,
+      "step": 8782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65020466451259e-06,
+      "loss": 0.8632,
+      "step": 8783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648096296791373e-06,
+      "loss": 1.0072,
+      "step": 8784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645988039733099e-06,
+      "loss": 0.9933,
+      "step": 8785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643879893436965e-06,
+      "loss": 0.9397,
+      "step": 8786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641771858002169e-06,
+      "loss": 0.8945,
+      "step": 8787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639663933527896e-06,
+      "loss": 0.9967,
+      "step": 8788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.637556120113332e-06,
+      "loss": 0.9267,
+      "step": 8789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635448417857654e-06,
+      "loss": 0.9858,
+      "step": 8790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633340826860029e-06,
+      "loss": 0.8798,
+      "step": 8791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.631233347219635e-06,
+      "loss": 1.0224,
+      "step": 8792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6291259790356235e-06,
+      "loss": 0.9814,
+      "step": 8793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.627018722407163e-06,
+      "loss": 0.9827,
+      "step": 8794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624911577433395e-06,
+      "loss": 1.0141,
+      "step": 8795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.622804544213468e-06,
+      "loss": 0.9877,
+      "step": 8796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620697622846527e-06,
+      "loss": 1.0076,
+      "step": 8797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618590813431708e-06,
+      "loss": 1.0187,
+      "step": 8798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616484116068131e-06,
+      "loss": 0.922,
+      "step": 8799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614377530854938e-06,
+      "loss": 1.0019,
+      "step": 8800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6122710578912365e-06,
+      "loss": 0.9053,
+      "step": 8801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610164697276148e-06,
+      "loss": 0.9238,
+      "step": 8802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608058449108778e-06,
+      "loss": 0.9469,
+      "step": 8803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605952313488232e-06,
+      "loss": 1.0138,
+      "step": 8804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.60384629051361e-06,
+      "loss": 0.9792,
+      "step": 8805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601740380284004e-06,
+      "loss": 0.9288,
+      "step": 8806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599634582898505e-06,
+      "loss": 0.9816,
+      "step": 8807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597528898456197e-06,
+      "loss": 0.9563,
+      "step": 8808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5954233270561496e-06,
+      "loss": 1.0363,
+      "step": 8809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593317868797443e-06,
+      "loss": 0.9599,
+      "step": 8810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591212523779141e-06,
+      "loss": 0.9593,
+      "step": 8811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589107292100307e-06,
+      "loss": 0.9496,
+      "step": 8812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.587002173859997e-06,
+      "loss": 0.9562,
+      "step": 8813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.58489716915726e-06,
+      "loss": 1.0287,
+      "step": 8814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582792278091148e-06,
+      "loss": 1.0279,
+      "step": 8815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.58068750076069e-06,
+      "loss": 0.8766,
+      "step": 8816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.578582837264934e-06,
+      "loss": 0.9173,
+      "step": 8817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576478287702901e-06,
+      "loss": 0.8551,
+      "step": 8818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574373852173616e-06,
+      "loss": 0.9797,
+      "step": 8819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572269530776102e-06,
+      "loss": 0.9649,
+      "step": 8820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570165323609369e-06,
+      "loss": 1.0004,
+      "step": 8821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568061230772428e-06,
+      "loss": 0.9819,
+      "step": 8822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565957252364282e-06,
+      "loss": 0.9678,
+      "step": 8823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563853388483921e-06,
+      "loss": 0.99,
+      "step": 8824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561749639230349e-06,
+      "loss": 0.9575,
+      "step": 8825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559646004702541e-06,
+      "loss": 0.8777,
+      "step": 8826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.557542484999489e-06,
+      "loss": 1.0293,
+      "step": 8827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5554390802201625e-06,
+      "loss": 0.8839,
+      "step": 8828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553335790463531e-06,
+      "loss": 1.0468,
+      "step": 8829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5512326158285635e-06,
+      "loss": 0.9666,
+      "step": 8830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5491295564142155e-06,
+      "loss": 0.9992,
+      "step": 8831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547026612319446e-06,
+      "loss": 0.9083,
+      "step": 8832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544923783643202e-06,
+      "loss": 0.9293,
+      "step": 8833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542821070484423e-06,
+      "loss": 0.9394,
+      "step": 8834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540718472942052e-06,
+      "loss": 0.935,
+      "step": 8835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538615991115019e-06,
+      "loss": 0.9684,
+      "step": 8836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53651362510225e-06,
+      "loss": 0.9006,
+      "step": 8837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53441137500267e-06,
+      "loss": 1.0046,
+      "step": 8838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532309240915191e-06,
+      "loss": 0.9835,
+      "step": 8839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530207222938728e-06,
+      "loss": 1.009,
+      "step": 8840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528105321172186e-06,
+      "loss": 0.9863,
+      "step": 8841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526003535714458e-06,
+      "loss": 0.9972,
+      "step": 8842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5239018666644455e-06,
+      "loss": 0.9892,
+      "step": 8843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521800314121032e-06,
+      "loss": 0.9445,
+      "step": 8844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519698878183107e-06,
+      "loss": 0.9555,
+      "step": 8845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.517597558949544e-06,
+      "loss": 0.9349,
+      "step": 8846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5154963565192145e-06,
+      "loss": 0.9351,
+      "step": 8847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51339527099099e-06,
+      "loss": 0.9459,
+      "step": 8848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511294302463725e-06,
+      "loss": 1.0016,
+      "step": 8849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5091934510362845e-06,
+      "loss": 1.0114,
+      "step": 8850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50709271680751e-06,
+      "loss": 0.9655,
+      "step": 8851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504992099876249e-06,
+      "loss": 0.9838,
+      "step": 8852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502891600341343e-06,
+      "loss": 0.9297,
+      "step": 8853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500791218301623e-06,
+      "loss": 0.972,
+      "step": 8854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4986909538559204e-06,
+      "loss": 0.9745,
+      "step": 8855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4965908071030565e-06,
+      "loss": 1.0096,
+      "step": 8856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.494490778141843e-06,
+      "loss": 0.9275,
+      "step": 8857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.492390867071103e-06,
+      "loss": 1.0142,
+      "step": 8858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490291073989632e-06,
+      "loss": 0.9325,
+      "step": 8859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488191398996236e-06,
+      "loss": 0.8935,
+      "step": 8860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486091842189709e-06,
+      "loss": 0.9094,
+      "step": 8861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483992403668836e-06,
+      "loss": 1.0046,
+      "step": 8862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4818930835324075e-06,
+      "loss": 0.9408,
+      "step": 8863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479793881879198e-06,
+      "loss": 0.9523,
+      "step": 8864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477694798807983e-06,
+      "loss": 1.0556,
+      "step": 8865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.47559583441753e-06,
+      "loss": 0.9459,
+      "step": 8866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473496988806592e-06,
+      "loss": 0.8885,
+      "step": 8867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471398262073936e-06,
+      "loss": 0.9987,
+      "step": 8868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4692996543183036e-06,
+      "loss": 0.9294,
+      "step": 8869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.467201165638447e-06,
+      "loss": 0.9915,
+      "step": 8870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4651027961331005e-06,
+      "loss": 0.9384,
+      "step": 8871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463004545900997e-06,
+      "loss": 0.9806,
+      "step": 8872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460906415040872e-06,
+      "loss": 1.0025,
+      "step": 8873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4588084036514375e-06,
+      "loss": 0.9555,
+      "step": 8874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.456710511831414e-06,
+      "loss": 0.9721,
+      "step": 8875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4546127396795156e-06,
+      "loss": 0.8647,
+      "step": 8876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452515087294443e-06,
+      "loss": 0.9792,
+      "step": 8877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4504175547749e-06,
+      "loss": 0.9913,
+      "step": 8878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4483201422195805e-06,
+      "loss": 0.9578,
+      "step": 8879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4462228497271695e-06,
+      "loss": 0.9912,
+      "step": 8880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.444125677396356e-06,
+      "loss": 0.9732,
+      "step": 8881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442028625325805e-06,
+      "loss": 0.9735,
+      "step": 8882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439931693614205e-06,
+      "loss": 0.9418,
+      "step": 8883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.437834882360211e-06,
+      "loss": 0.903,
+      "step": 8884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435738191662483e-06,
+      "loss": 0.9887,
+      "step": 8885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433641621619679e-06,
+      "loss": 0.9983,
+      "step": 8886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431545172330445e-06,
+      "loss": 0.8711,
+      "step": 8887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429448843893427e-06,
+      "loss": 0.8824,
+      "step": 8888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427352636407265e-06,
+      "loss": 0.9383,
+      "step": 8889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.425256549970581e-06,
+      "loss": 1.0037,
+      "step": 8890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423160584682013e-06,
+      "loss": 1.004,
+      "step": 8891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421064740640172e-06,
+      "loss": 0.9423,
+      "step": 8892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.418969017943678e-06,
+      "loss": 1.0343,
+      "step": 8893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416873416691139e-06,
+      "loss": 0.9979,
+      "step": 8894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414777936981155e-06,
+      "loss": 0.9855,
+      "step": 8895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.412682578912329e-06,
+      "loss": 0.9339,
+      "step": 8896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410587342583247e-06,
+      "loss": 0.9192,
+      "step": 8897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.408492228092503e-06,
+      "loss": 1.01,
+      "step": 8898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406397235538671e-06,
+      "loss": 0.9617,
+      "step": 8899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404302365020324e-06,
+      "loss": 1.0033,
+      "step": 8900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.402207616636037e-06,
+      "loss": 1.0424,
+      "step": 8901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.400112990484367e-06,
+      "loss": 0.8891,
+      "step": 8902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398018486663877e-06,
+      "loss": 0.9942,
+      "step": 8903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.395924105273117e-06,
+      "loss": 0.9589,
+      "step": 8904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393829846410629e-06,
+      "loss": 1.0357,
+      "step": 8905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.391735710174961e-06,
+      "loss": 0.995,
+      "step": 8906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389641696664635e-06,
+      "loss": 0.9781,
+      "step": 8907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387547805978196e-06,
+      "loss": 1.0324,
+      "step": 8908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.385454038214155e-06,
+      "loss": 0.9516,
+      "step": 8909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383360393471028e-06,
+      "loss": 0.9733,
+      "step": 8910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381266871847334e-06,
+      "loss": 1.0243,
+      "step": 8911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.379173473441576e-06,
+      "loss": 0.9774,
+      "step": 8912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.377080198352246e-06,
+      "loss": 0.9122,
+      "step": 8913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.37498704667785e-06,
+      "loss": 1.0345,
+      "step": 8914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372894018516864e-06,
+      "loss": 0.9087,
+      "step": 8915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3708011139677825e-06,
+      "loss": 0.949,
+      "step": 8916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368708333129073e-06,
+      "loss": 0.9441,
+      "step": 8917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366615676099207e-06,
+      "loss": 0.948,
+      "step": 8918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3645231429766525e-06,
+      "loss": 0.9729,
+      "step": 8919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.362430733859866e-06,
+      "loss": 0.9294,
+      "step": 8920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.360338448847302e-06,
+      "loss": 0.8715,
+      "step": 8921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3582462880374095e-06,
+      "loss": 0.9114,
+      "step": 8922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.356154251528622e-06,
+      "loss": 0.9918,
+      "step": 8923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354062339419387e-06,
+      "loss": 0.8983,
+      "step": 8924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351970551808123e-06,
+      "loss": 0.9202,
+      "step": 8925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34987888879326e-06,
+      "loss": 0.9632,
+      "step": 8926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3477873504732155e-06,
+      "loss": 0.9767,
+      "step": 8927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345695936946398e-06,
+      "loss": 1.0038,
+      "step": 8928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.343604648311219e-06,
+      "loss": 0.8885,
+      "step": 8929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.341513484666074e-06,
+      "loss": 0.9371,
+      "step": 8930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.339422446109364e-06,
+      "loss": 0.9419,
+      "step": 8931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33733153273947e-06,
+      "loss": 0.9429,
+      "step": 8932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335240744654777e-06,
+      "loss": 0.915,
+      "step": 8933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333150081953665e-06,
+      "loss": 0.9262,
+      "step": 8934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3310595447345e-06,
+      "loss": 0.9917,
+      "step": 8935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.328969133095651e-06,
+      "loss": 0.9575,
+      "step": 8936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326878847135478e-06,
+      "loss": 0.9524,
+      "step": 8937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.324788686952329e-06,
+      "loss": 0.8553,
+      "step": 8938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.322698652644557e-06,
+      "loss": 0.96,
+      "step": 8939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320608744310495e-06,
+      "loss": 1.0131,
+      "step": 8940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318518962048491e-06,
+      "loss": 0.9239,
+      "step": 8941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.316429305956865e-06,
+      "loss": 1.0061,
+      "step": 8942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31433977613394e-06,
+      "loss": 0.9533,
+      "step": 8943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31225037267804e-06,
+      "loss": 0.9108,
+      "step": 8944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.310161095687471e-06,
+      "loss": 0.8996,
+      "step": 8945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308071945260542e-06,
+      "loss": 0.9322,
+      "step": 8946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.305982921495554e-06,
+      "loss": 0.9666,
+      "step": 8947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303894024490793e-06,
+      "loss": 0.8791,
+      "step": 8948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3018052543445585e-06,
+      "loss": 0.8919,
+      "step": 8949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.299716611155123e-06,
+      "loss": 1.0142,
+      "step": 8950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2976280950207635e-06,
+      "loss": 0.9693,
+      "step": 8951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295539706039756e-06,
+      "loss": 0.8985,
+      "step": 8952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293451444310356e-06,
+      "loss": 0.9748,
+      "step": 8953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291363309930828e-06,
+      "loss": 1.0107,
+      "step": 8954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2892753029994246e-06,
+      "loss": 0.9705,
+      "step": 8955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.287187423614384e-06,
+      "loss": 0.9429,
+      "step": 8956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285099671873954e-06,
+      "loss": 0.9167,
+      "step": 8957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283012047876362e-06,
+      "loss": 0.9838,
+      "step": 8958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280924551719842e-06,
+      "loss": 0.9824,
+      "step": 8959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278837183502614e-06,
+      "loss": 1.0149,
+      "step": 8960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276749943322891e-06,
+      "loss": 0.875,
+      "step": 8961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.274662831278887e-06,
+      "loss": 0.9154,
+      "step": 8962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272575847468801e-06,
+      "loss": 0.9108,
+      "step": 8963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2704889919908385e-06,
+      "loss": 0.9875,
+      "step": 8964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.268402264943185e-06,
+      "loss": 0.9437,
+      "step": 8965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266315666424026e-06,
+      "loss": 1.0334,
+      "step": 8966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264229196531542e-06,
+      "loss": 0.9524,
+      "step": 8967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2621428553639095e-06,
+      "loss": 1.016,
+      "step": 8968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260056643019293e-06,
+      "loss": 0.9917,
+      "step": 8969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257970559595859e-06,
+      "loss": 0.9237,
+      "step": 8970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255884605191752e-06,
+      "loss": 0.9812,
+      "step": 8971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.253798779905135e-06,
+      "loss": 0.989,
+      "step": 8972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251713083834137e-06,
+      "loss": 0.9757,
+      "step": 8973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24962751707691e-06,
+      "loss": 0.9688,
+      "step": 8974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.247542079731577e-06,
+      "loss": 0.9745,
+      "step": 8975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2454567718962594e-06,
+      "loss": 0.9602,
+      "step": 8976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243371593669086e-06,
+      "loss": 0.9306,
+      "step": 8977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241286545148161e-06,
+      "loss": 0.8751,
+      "step": 8978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.239201626431595e-06,
+      "loss": 0.9447,
+      "step": 8979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237116837617493e-06,
+      "loss": 0.9218,
+      "step": 8980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.235032178803936e-06,
+      "loss": 0.9759,
+      "step": 8981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.232947650089028e-06,
+      "loss": 1.0314,
+      "step": 8982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230863251570841e-06,
+      "loss": 0.9724,
+      "step": 8983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.228778983347456e-06,
+      "loss": 0.9255,
+      "step": 8984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226694845516941e-06,
+      "loss": 0.9026,
+      "step": 8985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22461083817736e-06,
+      "loss": 0.9613,
+      "step": 8986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222526961426772e-06,
+      "loss": 0.9646,
+      "step": 8987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.220443215363231e-06,
+      "loss": 0.9385,
+      "step": 8988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218359600084775e-06,
+      "loss": 0.9158,
+      "step": 8989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216276115689449e-06,
+      "loss": 0.9243,
+      "step": 8990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2141927622752824e-06,
+      "loss": 0.9147,
+      "step": 8991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212109539940307e-06,
+      "loss": 0.9109,
+      "step": 8992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210026448782541e-06,
+      "loss": 0.9965,
+      "step": 8993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207943488899997e-06,
+      "loss": 1.0087,
+      "step": 8994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.205860660390691e-06,
+      "loss": 0.9357,
+      "step": 8995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2037779633526115e-06,
+      "loss": 0.9172,
+      "step": 8996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2016953978837715e-06,
+      "loss": 0.9653,
+      "step": 8997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199612964082149e-06,
+      "loss": 1.0034,
+      "step": 8998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1975306620457285e-06,
+      "loss": 0.9366,
+      "step": 8999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1954484918724935e-06,
+      "loss": 1.0095,
+      "step": 9000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1933664536604085e-06,
+      "loss": 0.9391,
+      "step": 9001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191284547507446e-06,
+      "loss": 0.8545,
+      "step": 9002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.189202773511561e-06,
+      "loss": 0.9812,
+      "step": 9003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.187121131770702e-06,
+      "loss": 0.9821,
+      "step": 9004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185039622382823e-06,
+      "loss": 0.9575,
+      "step": 9005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.182958245445857e-06,
+      "loss": 1.0276,
+      "step": 9006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1808770010577465e-06,
+      "loss": 0.905,
+      "step": 9007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.178795889316413e-06,
+      "loss": 0.9386,
+      "step": 9008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.176714910319777e-06,
+      "loss": 0.9213,
+      "step": 9009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17463406416576e-06,
+      "loss": 0.9693,
+      "step": 9010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1725533509522625e-06,
+      "loss": 0.9281,
+      "step": 9011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170472770777196e-06,
+      "loss": 0.9288,
+      "step": 9012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168392323738454e-06,
+      "loss": 0.9526,
+      "step": 9013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.166312009933923e-06,
+      "loss": 0.9919,
+      "step": 9014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.16423182946149e-06,
+      "loss": 0.9282,
+      "step": 9015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.162151782419031e-06,
+      "loss": 0.9536,
+      "step": 9016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160071868904421e-06,
+      "loss": 0.9388,
+      "step": 9017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157992089015524e-06,
+      "loss": 0.888,
+      "step": 9018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155912442850194e-06,
+      "loss": 0.9222,
+      "step": 9019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.15383293050629e-06,
+      "loss": 0.9878,
+      "step": 9020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151753552081653e-06,
+      "loss": 0.9888,
+      "step": 9021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14967430767413e-06,
+      "loss": 0.8909,
+      "step": 9022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147595197381548e-06,
+      "loss": 1.0064,
+      "step": 9023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145516221301735e-06,
+      "loss": 0.9699,
+      "step": 9024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1434373795325155e-06,
+      "loss": 0.8661,
+      "step": 9025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.141358672171701e-06,
+      "loss": 0.9897,
+      "step": 9026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.139280099317101e-06,
+      "loss": 0.9223,
+      "step": 9027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137201661066519e-06,
+      "loss": 0.8887,
+      "step": 9028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135123357517746e-06,
+      "loss": 0.9288,
+      "step": 9029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133045188768579e-06,
+      "loss": 0.8664,
+      "step": 9030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130967154916795e-06,
+      "loss": 0.9725,
+      "step": 9031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128889256060171e-06,
+      "loss": 0.9466,
+      "step": 9032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.126811492296479e-06,
+      "loss": 0.9033,
+      "step": 9033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124733863723483e-06,
+      "loss": 0.9123,
+      "step": 9034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1226563704389404e-06,
+      "loss": 0.9808,
+      "step": 9035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.120579012540606e-06,
+      "loss": 0.9788,
+      "step": 9036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118501790126214e-06,
+      "loss": 0.9795,
+      "step": 9037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116424703293516e-06,
+      "loss": 0.9399,
+      "step": 9038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114347752140235e-06,
+      "loss": 1.0483,
+      "step": 9039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112270936764101e-06,
+      "loss": 0.9453,
+      "step": 9040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1101942572628326e-06,
+      "loss": 0.8875,
+      "step": 9041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1081177137341395e-06,
+      "loss": 0.9514,
+      "step": 9042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106041306275734e-06,
+      "loss": 0.931,
+      "step": 9043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.103965034985312e-06,
+      "loss": 1.0207,
+      "step": 9044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101888899960569e-06,
+      "loss": 0.9324,
+      "step": 9045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.099812901299194e-06,
+      "loss": 0.9903,
+      "step": 9046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097737039098862e-06,
+      "loss": 0.9396,
+      "step": 9047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.095661313457254e-06,
+      "loss": 1.0015,
+      "step": 9048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093585724472033e-06,
+      "loss": 0.911,
+      "step": 9049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091510272240865e-06,
+      "loss": 0.8946,
+      "step": 9050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089434956861404e-06,
+      "loss": 0.9155,
+      "step": 9051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087359778431295e-06,
+      "loss": 1.0089,
+      "step": 9052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085284737048188e-06,
+      "loss": 0.9194,
+      "step": 9053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083209832809708e-06,
+      "loss": 1.0073,
+      "step": 9054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081135065813498e-06,
+      "loss": 0.9909,
+      "step": 9055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07906043615717e-06,
+      "loss": 1.06,
+      "step": 9056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076985943938343e-06,
+      "loss": 0.9008,
+      "step": 9057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074911589254629e-06,
+      "loss": 0.9003,
+      "step": 9058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.072837372203629e-06,
+      "loss": 1.0407,
+      "step": 9059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070763292882945e-06,
+      "loss": 0.9423,
+      "step": 9060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068689351390167e-06,
+      "loss": 1.0307,
+      "step": 9061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06661554782287e-06,
+      "loss": 1.0243,
+      "step": 9062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.064541882278644e-06,
+      "loss": 0.9621,
+      "step": 9063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.062468354855052e-06,
+      "loss": 0.9826,
+      "step": 9064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0603949656496595e-06,
+      "loss": 1.0,
+      "step": 9065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058321714760028e-06,
+      "loss": 1.0204,
+      "step": 9066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056248602283704e-06,
+      "loss": 0.9801,
+      "step": 9067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.054175628318241e-06,
+      "loss": 1.0098,
+      "step": 9068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052102792961172e-06,
+      "loss": 0.9396,
+      "step": 9069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0500300963100235e-06,
+      "loss": 1.0471,
+      "step": 9070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.047957538462334e-06,
+      "loss": 0.9308,
+      "step": 9071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045885119515612e-06,
+      "loss": 0.897,
+      "step": 9072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043812839567376e-06,
+      "loss": 0.9655,
+      "step": 9073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0417406987151295e-06,
+      "loss": 0.9745,
+      "step": 9074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039668697056369e-06,
+      "loss": 1.0394,
+      "step": 9075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037596834688592e-06,
+      "loss": 0.8996,
+      "step": 9076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0355251117092825e-06,
+      "loss": 1.0054,
+      "step": 9077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033453528215924e-06,
+      "loss": 1.0073,
+      "step": 9078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.031382084305986e-06,
+      "loss": 0.9272,
+      "step": 9079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.029310780076932e-06,
+      "loss": 0.9704,
+      "step": 9080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.027239615626229e-06,
+      "loss": 1.0013,
+      "step": 9081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025168591051325e-06,
+      "loss": 0.9683,
+      "step": 9082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023097706449671e-06,
+      "loss": 0.8719,
+      "step": 9083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0210269619187056e-06,
+      "loss": 0.9096,
+      "step": 9084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0189563575558595e-06,
+      "loss": 0.9847,
+      "step": 9085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.016885893458567e-06,
+      "loss": 0.9785,
+      "step": 9086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014815569724238e-06,
+      "loss": 0.9765,
+      "step": 9087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0127453864502995e-06,
+      "loss": 0.9496,
+      "step": 9088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.010675343734149e-06,
+      "loss": 0.8638,
+      "step": 9089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008605441673189e-06,
+      "loss": 0.9102,
+      "step": 9090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006535680364816e-06,
+      "loss": 0.8966,
+      "step": 9091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004466059906413e-06,
+      "loss": 0.9145,
+      "step": 9092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.002396580395368e-06,
+      "loss": 0.9171,
+      "step": 9093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000327241929052e-06,
+      "loss": 0.9486,
+      "step": 9094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.998258044604824e-06,
+      "loss": 0.9724,
+      "step": 9095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996188988520061e-06,
+      "loss": 0.9514,
+      "step": 9096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.994120073772106e-06,
+      "loss": 1.0292,
+      "step": 9097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992051300458307e-06,
+      "loss": 0.9325,
+      "step": 9098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.98998266867601e-06,
+      "loss": 0.9578,
+      "step": 9099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987914178522543e-06,
+      "loss": 1.0585,
+      "step": 9100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985845830095242e-06,
+      "loss": 0.9796,
+      "step": 9101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983777623491424e-06,
+      "loss": 0.977,
+      "step": 9102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981709558808394e-06,
+      "loss": 0.9856,
+      "step": 9103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979641636143477e-06,
+      "loss": 1.0275,
+      "step": 9104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9775738555939594e-06,
+      "loss": 0.913,
+      "step": 9105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975506217257144e-06,
+      "loss": 0.9594,
+      "step": 9106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973438721230317e-06,
+      "loss": 0.9917,
+      "step": 9107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.971371367610753e-06,
+      "loss": 0.985,
+      "step": 9108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969304156495733e-06,
+      "loss": 0.9306,
+      "step": 9109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967237087982521e-06,
+      "loss": 0.9888,
+      "step": 9110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965170162168384e-06,
+      "loss": 0.9627,
+      "step": 9111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9631033791505685e-06,
+      "loss": 0.992,
+      "step": 9112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961036739026321e-06,
+      "loss": 0.9155,
+      "step": 9113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9589702418928885e-06,
+      "loss": 0.9843,
+      "step": 9114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956903887847499e-06,
+      "loss": 0.9714,
+      "step": 9115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954837676987386e-06,
+      "loss": 1.0198,
+      "step": 9116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952771609409765e-06,
+      "loss": 0.9279,
+      "step": 9117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.950705685211849e-06,
+      "loss": 0.9814,
+      "step": 9118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948639904490851e-06,
+      "loss": 0.9797,
+      "step": 9119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.946574267343962e-06,
+      "loss": 0.9457,
+      "step": 9120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.944508773868384e-06,
+      "loss": 0.969,
+      "step": 9121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9424434241613e-06,
+      "loss": 0.9212,
+      "step": 9122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940378218319886e-06,
+      "loss": 0.8886,
+      "step": 9123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938313156441323e-06,
+      "loss": 0.9772,
+      "step": 9124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93624823862277e-06,
+      "loss": 1.0513,
+      "step": 9125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9341834649613915e-06,
+      "loss": 1.0038,
+      "step": 9126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93211883555434e-06,
+      "loss": 0.9555,
+      "step": 9127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9300543504987564e-06,
+      "loss": 0.9239,
+      "step": 9128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927990009891788e-06,
+      "loss": 0.9195,
+      "step": 9129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.925925813830558e-06,
+      "loss": 0.8422,
+      "step": 9130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.923861762412198e-06,
+      "loss": 1.0168,
+      "step": 9131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921797855733827e-06,
+      "loss": 0.9159,
+      "step": 9132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919734093892554e-06,
+      "loss": 0.9287,
+      "step": 9133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917670476985486e-06,
+      "loss": 0.9377,
+      "step": 9134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.915607005109723e-06,
+      "loss": 0.9641,
+      "step": 9135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913543678362351e-06,
+      "loss": 0.9097,
+      "step": 9136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.91148049684046e-06,
+      "loss": 0.9191,
+      "step": 9137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909417460641123e-06,
+      "loss": 0.9689,
+      "step": 9138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.907354569861417e-06,
+      "loss": 1.0106,
+      "step": 9139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.905291824598403e-06,
+      "loss": 0.9058,
+      "step": 9140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903229224949136e-06,
+      "loss": 0.9128,
+      "step": 9141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90116677101067e-06,
+      "loss": 0.8665,
+      "step": 9142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.899104462880046e-06,
+      "loss": 0.9466,
+      "step": 9143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8970423006543075e-06,
+      "loss": 0.9292,
+      "step": 9144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.894980284430476e-06,
+      "loss": 0.9414,
+      "step": 9145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.892918414305576e-06,
+      "loss": 1.0321,
+      "step": 9146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.890856690376626e-06,
+      "loss": 1.0096,
+      "step": 9147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888795112740634e-06,
+      "loss": 0.9375,
+      "step": 9148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.886733681494604e-06,
+      "loss": 1.0104,
+      "step": 9149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884672396735533e-06,
+      "loss": 0.9129,
+      "step": 9150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8826112585604e-06,
+      "loss": 0.9634,
+      "step": 9151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8805502670662e-06,
+      "loss": 0.9974,
+      "step": 9152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878489422349895e-06,
+      "loss": 0.9959,
+      "step": 9153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.876428724508466e-06,
+      "loss": 0.9114,
+      "step": 9154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8743681736388655e-06,
+      "loss": 0.9418,
+      "step": 9155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.872307769838047e-06,
+      "loss": 1.0297,
+      "step": 9156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.870247513202963e-06,
+      "loss": 0.9512,
+      "step": 9157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868187403830547e-06,
+      "loss": 1.001,
+      "step": 9158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.866127441817739e-06,
+      "loss": 1.0143,
+      "step": 9159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864067627261465e-06,
+      "loss": 0.9438,
+      "step": 9160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8620079602586345e-06,
+      "loss": 1.0168,
+      "step": 9161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.859948440906174e-06,
+      "loss": 1.0154,
+      "step": 9162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857889069300978e-06,
+      "loss": 0.9107,
+      "step": 9163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8558298455399515e-06,
+      "loss": 0.8846,
+      "step": 9164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853770769719985e-06,
+      "loss": 0.9709,
+      "step": 9165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8517118419379594e-06,
+      "loss": 0.9579,
+      "step": 9166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849653062290758e-06,
+      "loss": 1.0348,
+      "step": 9167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847594430875244e-06,
+      "loss": 0.9391,
+      "step": 9168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845535947788293e-06,
+      "loss": 0.8962,
+      "step": 9169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8434776131267515e-06,
+      "loss": 0.8958,
+      "step": 9170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.84141942698747e-06,
+      "loss": 0.8805,
+      "step": 9171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.839361389467296e-06,
+      "loss": 0.8827,
+      "step": 9172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837303500663063e-06,
+      "loss": 0.9454,
+      "step": 9173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835245760671597e-06,
+      "loss": 0.9677,
+      "step": 9174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833188169589729e-06,
+      "loss": 0.9325,
+      "step": 9175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.831130727514259e-06,
+      "loss": 0.9213,
+      "step": 9176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829073434542011e-06,
+      "loss": 0.9094,
+      "step": 9177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.827016290769775e-06,
+      "loss": 0.9038,
+      "step": 9178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824959296294344e-06,
+      "loss": 0.9964,
+      "step": 9179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.822902451212513e-06,
+      "loss": 0.8983,
+      "step": 9180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820845755621055e-06,
+      "loss": 0.9226,
+      "step": 9181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818789209616745e-06,
+      "loss": 1.0351,
+      "step": 9182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816732813296352e-06,
+      "loss": 0.9255,
+      "step": 9183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.814676566756625e-06,
+      "loss": 0.9631,
+      "step": 9184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812620470094328e-06,
+      "loss": 0.997,
+      "step": 9185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.810564523406191e-06,
+      "loss": 0.9171,
+      "step": 9186
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808508726788969e-06,
+      "loss": 1.005,
+      "step": 9187
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806453080339378e-06,
+      "loss": 0.9831,
+      "step": 9188
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804397584154146e-06,
+      "loss": 0.9612,
+      "step": 9189
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802342238329989e-06,
+      "loss": 0.8711,
+      "step": 9190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800287042963616e-06,
+      "loss": 0.9725,
+      "step": 9191
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798231998151732e-06,
+      "loss": 0.9092,
+      "step": 9192
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796177103991031e-06,
+      "loss": 1.0273,
+      "step": 9193
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7941223605781945e-06,
+      "loss": 1.035,
+      "step": 9194
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792067768009911e-06,
+      "loss": 1.0204,
+      "step": 9195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790013326382851e-06,
+      "loss": 0.9788,
+      "step": 9196
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.787959035793683e-06,
+      "loss": 1.0309,
+      "step": 9197
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7859048963390636e-06,
+      "loss": 0.8895,
+      "step": 9198
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783850908115646e-06,
+      "loss": 0.9928,
+      "step": 9199
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781797071220078e-06,
+      "loss": 0.9962,
+      "step": 9200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.779743385748994e-06,
+      "loss": 0.9702,
+      "step": 9201
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777689851799032e-06,
+      "loss": 0.9652,
+      "step": 9202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7756364694668075e-06,
+      "loss": 0.9782,
+      "step": 9203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77358323884894e-06,
+      "loss": 0.9071,
+      "step": 9204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77153016004204e-06,
+      "loss": 0.9817,
+      "step": 9205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769477233142708e-06,
+      "loss": 0.9847,
+      "step": 9206
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.767424458247545e-06,
+      "loss": 1.0013,
+      "step": 9207
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7653718354531364e-06,
+      "loss": 0.9269,
+      "step": 9208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.763319364856056e-06,
+      "loss": 0.9799,
+      "step": 9209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.761267046552888e-06,
+      "loss": 0.9653,
+      "step": 9210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7592148806401945e-06,
+      "loss": 1.0238,
+      "step": 9211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.757162867214535e-06,
+      "loss": 1.0187,
+      "step": 9212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.755111006372463e-06,
+      "loss": 0.9393,
+      "step": 9213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75305929821052e-06,
+      "loss": 0.9934,
+      "step": 9214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7510077428252495e-06,
+      "loss": 0.9414,
+      "step": 9215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748956340313182e-06,
+      "loss": 0.9603,
+      "step": 9216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746905090770832e-06,
+      "loss": 0.9473,
+      "step": 9217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744853994294731e-06,
+      "loss": 0.8873,
+      "step": 9218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742803050981375e-06,
+      "loss": 0.9158,
+      "step": 9219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740752260927272e-06,
+      "loss": 1.0144,
+      "step": 9220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738701624228916e-06,
+      "loss": 0.8542,
+      "step": 9221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7366511409827924e-06,
+      "loss": 0.9544,
+      "step": 9222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734600811285388e-06,
+      "loss": 1.0747,
+      "step": 9223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732550635233166e-06,
+      "loss": 1.0305,
+      "step": 9224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730500612922601e-06,
+      "loss": 0.8511,
+      "step": 9225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728450744450152e-06,
+      "loss": 0.8931,
+      "step": 9226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.726401029912262e-06,
+      "loss": 0.9501,
+      "step": 9227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724351469405381e-06,
+      "loss": 0.9642,
+      "step": 9228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722302063025944e-06,
+      "loss": 0.8676,
+      "step": 9229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7202528108703835e-06,
+      "loss": 0.8669,
+      "step": 9230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718203713035121e-06,
+      "loss": 0.8542,
+      "step": 9231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716154769616567e-06,
+      "loss": 0.9938,
+      "step": 9232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.714105980711138e-06,
+      "loss": 0.9605,
+      "step": 9233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712057346415224e-06,
+      "loss": 0.9574,
+      "step": 9234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71000886682523e-06,
+      "loss": 0.9512,
+      "step": 9235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707960542037535e-06,
+      "loss": 0.9423,
+      "step": 9236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705912372148516e-06,
+      "loss": 0.94,
+      "step": 9237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70386435725455e-06,
+      "loss": 1.022,
+      "step": 9238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701816497451999e-06,
+      "loss": 0.9848,
+      "step": 9239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699768792837218e-06,
+      "loss": 0.9452,
+      "step": 9240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697721243506563e-06,
+      "loss": 0.8887,
+      "step": 9241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695673849556365e-06,
+      "loss": 0.9255,
+      "step": 9242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693626611082971e-06,
+      "loss": 0.8893,
+      "step": 9243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691579528182696e-06,
+      "loss": 0.8611,
+      "step": 9244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689532600951877e-06,
+      "loss": 0.9841,
+      "step": 9245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.687485829486815e-06,
+      "loss": 1.0077,
+      "step": 9246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.685439213883816e-06,
+      "loss": 0.9081,
+      "step": 9247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.683392754239182e-06,
+      "loss": 0.8946,
+      "step": 9248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681346450649205e-06,
+      "loss": 0.977,
+      "step": 9249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679300303210161e-06,
+      "loss": 0.9671,
+      "step": 9250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.677254312018337e-06,
+      "loss": 0.9711,
+      "step": 9251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.675208477169995e-06,
+      "loss": 0.9703,
+      "step": 9252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673162798761399e-06,
+      "loss": 0.9875,
+      "step": 9253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6711172768888035e-06,
+      "loss": 1.0316,
+      "step": 9254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669071911648452e-06,
+      "loss": 0.9338,
+      "step": 9255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.667026703136589e-06,
+      "loss": 0.9543,
+      "step": 9256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664981651449443e-06,
+      "loss": 0.9523,
+      "step": 9257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662936756683241e-06,
+      "loss": 0.8758,
+      "step": 9258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.660892018934202e-06,
+      "loss": 0.9747,
+      "step": 9259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.65884743829853e-06,
+      "loss": 0.9357,
+      "step": 9260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656803014872434e-06,
+      "loss": 1.0291,
+      "step": 9261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6547587487521035e-06,
+      "loss": 0.9523,
+      "step": 9262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652714640033733e-06,
+      "loss": 0.9287,
+      "step": 9263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650670688813497e-06,
+      "loss": 1.0096,
+      "step": 9264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6486268951875695e-06,
+      "loss": 0.9962,
+      "step": 9265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6465832592521214e-06,
+      "loss": 0.9156,
+      "step": 9266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644539781103302e-06,
+      "loss": 0.9606,
+      "step": 9267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.642496460837272e-06,
+      "loss": 0.9266,
+      "step": 9268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640453298550168e-06,
+      "loss": 0.9975,
+      "step": 9269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638410294338125e-06,
+      "loss": 1.0101,
+      "step": 9270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6363674482972766e-06,
+      "loss": 1.0106,
+      "step": 9271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6343247605237385e-06,
+      "loss": 0.9244,
+      "step": 9272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632282231113629e-06,
+      "loss": 0.9827,
+      "step": 9273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6302398601630545e-06,
+      "loss": 1.0526,
+      "step": 9274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.628197647768105e-06,
+      "loss": 0.9126,
+      "step": 9275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.626155594024883e-06,
+      "loss": 1.0137,
+      "step": 9276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624113699029465e-06,
+      "loss": 0.9407,
+      "step": 9277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62207196287793e-06,
+      "loss": 0.8873,
+      "step": 9278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620030385666347e-06,
+      "loss": 0.9962,
+      "step": 9279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.617988967490776e-06,
+      "loss": 0.8986,
+      "step": 9280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.615947708447273e-06,
+      "loss": 0.9267,
+      "step": 9281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613906608631878e-06,
+      "loss": 0.9856,
+      "step": 9282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.611865668140638e-06,
+      "loss": 0.9015,
+      "step": 9283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.609824887069584e-06,
+      "loss": 0.8981,
+      "step": 9284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607784265514731e-06,
+      "loss": 1.0157,
+      "step": 9285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6057438035721046e-06,
+      "loss": 1.0179,
+      "step": 9286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60370350133771e-06,
+      "loss": 0.9476,
+      "step": 9287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.601663358907546e-06,
+      "loss": 0.9776,
+      "step": 9288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599623376377612e-06,
+      "loss": 0.951,
+      "step": 9289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597583553843888e-06,
+      "loss": 0.8803,
+      "step": 9290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.595543891402361e-06,
+      "loss": 0.9792,
+      "step": 9291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.593504389148995e-06,
+      "loss": 0.9692,
+      "step": 9292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591465047179755e-06,
+      "loss": 0.9651,
+      "step": 9293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589425865590599e-06,
+      "loss": 0.9256,
+      "step": 9294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587386844477472e-06,
+      "loss": 0.9528,
+      "step": 9295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.585347983936321e-06,
+      "loss": 0.9727,
+      "step": 9296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5833092840630755e-06,
+      "loss": 0.8985,
+      "step": 9297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581270744953661e-06,
+      "loss": 0.9316,
+      "step": 9298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579232366704e-06,
+      "loss": 0.9442,
+      "step": 9299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577194149409994e-06,
+      "loss": 0.9895,
+      "step": 9300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575156093167559e-06,
+      "loss": 1.0084,
+      "step": 9301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573118198072583e-06,
+      "loss": 0.9575,
+      "step": 9302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571080464220951e-06,
+      "loss": 0.9638,
+      "step": 9303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569042891708552e-06,
+      "loss": 0.9334,
+      "step": 9304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567005480631251e-06,
+      "loss": 1.0663,
+      "step": 9305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56496823108492e-06,
+      "loss": 0.8828,
+      "step": 9306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562931143165414e-06,
+      "loss": 0.8933,
+      "step": 9307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.560894216968578e-06,
+      "loss": 0.957,
+      "step": 9308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558857452590264e-06,
+      "loss": 0.9928,
+      "step": 9309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556820850126297e-06,
+      "loss": 0.9441,
+      "step": 9310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5547844096725145e-06,
+      "loss": 0.9531,
+      "step": 9311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.552748131324728e-06,
+      "loss": 0.9785,
+      "step": 9312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550712015178751e-06,
+      "loss": 0.8639,
+      "step": 9313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.54867606133039e-06,
+      "loss": 0.9416,
+      "step": 9314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5466402698754395e-06,
+      "loss": 0.9051,
+      "step": 9315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544604640909694e-06,
+      "loss": 0.9438,
+      "step": 9316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542569174528929e-06,
+      "loss": 0.8568,
+      "step": 9317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.540533870828918e-06,
+      "loss": 0.9253,
+      "step": 9318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538498729905429e-06,
+      "loss": 0.9383,
+      "step": 9319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53646375185422e-06,
+      "loss": 0.9211,
+      "step": 9320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534428936771047e-06,
+      "loss": 0.9427,
+      "step": 9321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5323942847516455e-06,
+      "loss": 0.9975,
+      "step": 9322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530359795891753e-06,
+      "loss": 1.0007,
+      "step": 9323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528325470287103e-06,
+      "loss": 0.9378,
+      "step": 9324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526291308033408e-06,
+      "loss": 0.9477,
+      "step": 9325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524257309226381e-06,
+      "loss": 0.9975,
+      "step": 9326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5222234739617304e-06,
+      "loss": 0.9128,
+      "step": 9327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5201898023351505e-06,
+      "loss": 0.9461,
+      "step": 9328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.518156294442335e-06,
+      "loss": 0.9538,
+      "step": 9329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.516122950378965e-06,
+      "loss": 0.9263,
+      "step": 9330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514089770240704e-06,
+      "loss": 0.9897,
+      "step": 9331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512056754123233e-06,
+      "loss": 0.9476,
+      "step": 9332
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510023902122198e-06,
+      "loss": 1.0032,
+      "step": 9333
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50799121433326e-06,
+      "loss": 0.9447,
+      "step": 9334
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505958690852057e-06,
+      "loss": 0.9752,
+      "step": 9335
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.503926331774222e-06,
+      "loss": 0.9571,
+      "step": 9336
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501894137195387e-06,
+      "loss": 0.9811,
+      "step": 9337
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499862107211168e-06,
+      "loss": 0.9198,
+      "step": 9338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.497830241917182e-06,
+      "loss": 0.996,
+      "step": 9339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.495798541409033e-06,
+      "loss": 0.8746,
+      "step": 9340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.493767005782308e-06,
+      "loss": 0.9566,
+      "step": 9341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49173563513261e-06,
+      "loss": 0.8299,
+      "step": 9342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489704429555508e-06,
+      "loss": 0.9078,
+      "step": 9343
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487673389146583e-06,
+      "loss": 0.9152,
+      "step": 9344
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.485642514001398e-06,
+      "loss": 0.9414,
+      "step": 9345
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483611804215508e-06,
+      "loss": 0.9147,
+      "step": 9346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.481581259884467e-06,
+      "loss": 0.9603,
+      "step": 9347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4795508811038155e-06,
+      "loss": 1.0274,
+      "step": 9348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.47752066796909e-06,
+      "loss": 0.8997,
+      "step": 9349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.475490620575815e-06,
+      "loss": 0.8383,
+      "step": 9350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.473460739019507e-06,
+      "loss": 0.8843,
+      "step": 9351
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471431023395682e-06,
+      "loss": 0.9371,
+      "step": 9352
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469401473799838e-06,
+      "loss": 0.987,
+      "step": 9353
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467372090327475e-06,
+      "loss": 0.9391,
+      "step": 9354
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465342873074079e-06,
+      "loss": 0.937,
+      "step": 9355
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.463313822135126e-06,
+      "loss": 0.927,
+      "step": 9356
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461284937606098e-06,
+      "loss": 0.9708,
+      "step": 9357
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.459256219582446e-06,
+      "loss": 0.9019,
+      "step": 9358
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457227668159637e-06,
+      "loss": 0.9765,
+      "step": 9359
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.455199283433113e-06,
+      "loss": 0.9213,
+      "step": 9360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453171065498316e-06,
+      "loss": 0.9635,
+      "step": 9361
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45114301445068e-06,
+      "loss": 0.9612,
+      "step": 9362
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449115130385632e-06,
+      "loss": 0.9062,
+      "step": 9363
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44708741339858e-06,
+      "loss": 0.9781,
+      "step": 9364
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445059863584947e-06,
+      "loss": 1.0099,
+      "step": 9365
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.443032481040119e-06,
+      "loss": 0.9545,
+      "step": 9366
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4410052658595036e-06,
+      "loss": 0.9094,
+      "step": 9367
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438978218138479e-06,
+      "loss": 0.9596,
+      "step": 9368
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436951337972419e-06,
+      "loss": 0.9861,
+      "step": 9369
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434924625456702e-06,
+      "loss": 1.0332,
+      "step": 9370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4328980806866825e-06,
+      "loss": 0.943,
+      "step": 9371
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.430871703757721e-06,
+      "loss": 0.9571,
+      "step": 9372
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4288454947651615e-06,
+      "loss": 1.004,
+      "step": 9373
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4268194538043384e-06,
+      "loss": 0.9198,
+      "step": 9374
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424793580970586e-06,
+      "loss": 0.9634,
+      "step": 9375
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422767876359224e-06,
+      "loss": 0.9837,
+      "step": 9376
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42074234006557e-06,
+      "loss": 0.896,
+      "step": 9377
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418716972184929e-06,
+      "loss": 1.0669,
+      "step": 9378
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416691772812598e-06,
+      "loss": 0.9708,
+      "step": 9379
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41466674204387e-06,
+      "loss": 1.0386,
+      "step": 9380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412641879974026e-06,
+      "loss": 0.9716,
+      "step": 9381
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410617186698344e-06,
+      "loss": 0.9504,
+      "step": 9382
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40859266231209e-06,
+      "loss": 1.0151,
+      "step": 9383
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406568306910517e-06,
+      "loss": 0.9454,
+      "step": 9384
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404544120588883e-06,
+      "loss": 0.9874,
+      "step": 9385
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402520103442426e-06,
+      "loss": 0.9554,
+      "step": 9386
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.400496255566388e-06,
+      "loss": 0.889,
+      "step": 9387
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398472577055992e-06,
+      "loss": 0.9511,
+      "step": 9388
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396449068006451e-06,
+      "loss": 0.9409,
+      "step": 9389
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394425728512987e-06,
+      "loss": 0.9264,
+      "step": 9390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3924025586707935e-06,
+      "loss": 0.9622,
+      "step": 9391
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390379558575076e-06,
+      "loss": 1.0071,
+      "step": 9392
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.388356728321015e-06,
+      "loss": 0.9023,
+      "step": 9393
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.386334068003788e-06,
+      "loss": 0.9093,
+      "step": 9394
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384311577718569e-06,
+      "loss": 0.9401,
+      "step": 9395
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3822892575605225e-06,
+      "loss": 1.0282,
+      "step": 9396
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.380267107624803e-06,
+      "loss": 0.9589,
+      "step": 9397
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378245128006559e-06,
+      "loss": 1.03,
+      "step": 9398
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376223318800925e-06,
+      "loss": 0.9562,
+      "step": 9399
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374201680103036e-06,
+      "loss": 0.9571,
+      "step": 9400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372180212008014e-06,
+      "loss": 1.0071,
+      "step": 9401
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.370158914610974e-06,
+      "loss": 0.9672,
+      "step": 9402
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368137788007023e-06,
+      "loss": 0.9269,
+      "step": 9403
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.366116832291259e-06,
+      "loss": 0.9755,
+      "step": 9404
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364096047558776e-06,
+      "loss": 0.9624,
+      "step": 9405
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36207543390466e-06,
+      "loss": 0.9047,
+      "step": 9406
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360054991423975e-06,
+      "loss": 0.9555,
+      "step": 9407
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.358034720211798e-06,
+      "loss": 0.9023,
+      "step": 9408
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35601462036318e-06,
+      "loss": 0.9433,
+      "step": 9409
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35399469197318e-06,
+      "loss": 0.943,
+      "step": 9410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3519749351368345e-06,
+      "loss": 0.9387,
+      "step": 9411
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349955349949178e-06,
+      "loss": 0.9981,
+      "step": 9412
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347935936505245e-06,
+      "loss": 0.9087,
+      "step": 9413
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3459166949000406e-06,
+      "loss": 0.9313,
+      "step": 9414
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343897625228588e-06,
+      "loss": 0.921,
+      "step": 9415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341878727585884e-06,
+      "loss": 0.9449,
+      "step": 9416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.33986000206692e-06,
+      "loss": 0.9486,
+      "step": 9417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337841448766686e-06,
+      "loss": 0.9888,
+      "step": 9418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335823067780158e-06,
+      "loss": 0.9339,
+      "step": 9419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333804859202308e-06,
+      "loss": 0.9997,
+      "step": 9420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331786823128099e-06,
+      "loss": 0.9646,
+      "step": 9421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329768959652476e-06,
+      "loss": 1.0131,
+      "step": 9422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3277512688703954e-06,
+      "loss": 0.9954,
+      "step": 9423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325733750876786e-06,
+      "loss": 0.9637,
+      "step": 9424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3237164057665845e-06,
+      "loss": 0.9919,
+      "step": 9425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321699233634708e-06,
+      "loss": 0.9577,
+      "step": 9426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3196822345760675e-06,
+      "loss": 0.8613,
+      "step": 9427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317665408685574e-06,
+      "loss": 0.9092,
+      "step": 9428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315648756058117e-06,
+      "loss": 0.9731,
+      "step": 9429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313632276788591e-06,
+      "loss": 0.9771,
+      "step": 9430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.311615970971874e-06,
+      "loss": 0.9493,
+      "step": 9431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.309599838702836e-06,
+      "loss": 0.9406,
+      "step": 9432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307583880076344e-06,
+      "loss": 0.9922,
+      "step": 9433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305568095187252e-06,
+      "loss": 1.0378,
+      "step": 9434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.303552484130411e-06,
+      "loss": 0.9269,
+      "step": 9435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301537047000657e-06,
+      "loss": 0.9328,
+      "step": 9436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299521783892821e-06,
+      "loss": 0.9817,
+      "step": 9437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297506694901732e-06,
+      "loss": 0.9487,
+      "step": 9438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2954917801222e-06,
+      "loss": 0.9567,
+      "step": 9439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29347703964903e-06,
+      "loss": 0.8973,
+      "step": 9440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2914624735770255e-06,
+      "loss": 0.9478,
+      "step": 9441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.289448082000971e-06,
+      "loss": 0.9287,
+      "step": 9442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287433865015655e-06,
+      "loss": 0.8928,
+      "step": 9443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.285419822715849e-06,
+      "loss": 0.9593,
+      "step": 9444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2834059551963155e-06,
+      "loss": 0.9361,
+      "step": 9445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.28139226255182e-06,
+      "loss": 0.9746,
+      "step": 9446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2793787448771e-06,
+      "loss": 0.9666,
+      "step": 9447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2773654022669084e-06,
+      "loss": 0.9172,
+      "step": 9448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275352234815971e-06,
+      "loss": 0.924,
+      "step": 9449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273339242619012e-06,
+      "loss": 0.98,
+      "step": 9450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271326425770753e-06,
+      "loss": 0.9812,
+      "step": 9451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269313784365896e-06,
+      "loss": 1.0164,
+      "step": 9452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267301318499146e-06,
+      "loss": 0.9268,
+      "step": 9453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.265289028265194e-06,
+      "loss": 0.8819,
+      "step": 9454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.263276913758715e-06,
+      "loss": 0.9491,
+      "step": 9455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261264975074397e-06,
+      "loss": 1.0621,
+      "step": 9456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.259253212306895e-06,
+      "loss": 0.9363,
+      "step": 9457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.257241625550876e-06,
+      "loss": 1.0036,
+      "step": 9458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.255230214900987e-06,
+      "loss": 0.9396,
+      "step": 9459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253218980451867e-06,
+      "loss": 0.9636,
+      "step": 9460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251207922298155e-06,
+      "loss": 0.9613,
+      "step": 9461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2491970405344715e-06,
+      "loss": 0.8913,
+      "step": 9462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247186335255437e-06,
+      "loss": 0.9943,
+      "step": 9463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2451758065556626e-06,
+      "loss": 0.9417,
+      "step": 9464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24316545452974e-06,
+      "loss": 0.9621,
+      "step": 9465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241155279272268e-06,
+      "loss": 0.9641,
+      "step": 9466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239145280877828e-06,
+      "loss": 0.9859,
+      "step": 9467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.237135459440996e-06,
+      "loss": 0.9379,
+      "step": 9468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2351258150563395e-06,
+      "loss": 0.8704,
+      "step": 9469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2331163478184166e-06,
+      "loss": 0.8971,
+      "step": 9470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231107057821781e-06,
+      "loss": 0.9494,
+      "step": 9471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.22909794516097e-06,
+      "loss": 0.9708,
+      "step": 9472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2270890099305185e-06,
+      "loss": 0.9405,
+      "step": 9473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225080252224953e-06,
+      "loss": 0.9112,
+      "step": 9474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223071672138789e-06,
+      "loss": 0.9544,
+      "step": 9475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221063269766539e-06,
+      "loss": 0.9417,
+      "step": 9476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2190550452027e-06,
+      "loss": 0.965,
+      "step": 9477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.217046998541764e-06,
+      "loss": 0.9298,
+      "step": 9478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21503912987822e-06,
+      "loss": 0.9898,
+      "step": 9479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213031439306531e-06,
+      "loss": 0.9763,
+      "step": 9480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211023926921178e-06,
+      "loss": 0.8216,
+      "step": 9481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.209016592816611e-06,
+      "loss": 0.9334,
+      "step": 9482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20700943708728e-06,
+      "loss": 0.9325,
+      "step": 9483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.205002459827631e-06,
+      "loss": 0.8414,
+      "step": 9484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202995661132094e-06,
+      "loss": 1.0292,
+      "step": 9485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2009890410950955e-06,
+      "loss": 1.0156,
+      "step": 9486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198982599811054e-06,
+      "loss": 1.0053,
+      "step": 9487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196976337374369e-06,
+      "loss": 0.8828,
+      "step": 9488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1949702538794535e-06,
+      "loss": 0.9381,
+      "step": 9489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192964349420687e-06,
+      "loss": 0.9326,
+      "step": 9490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190958624092458e-06,
+      "loss": 0.9036,
+      "step": 9491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188953077989139e-06,
+      "loss": 0.983,
+      "step": 9492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1869477112050954e-06,
+      "loss": 0.9419,
+      "step": 9493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184942523834689e-06,
+      "loss": 0.9749,
+      "step": 9494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.182937515972263e-06,
+      "loss": 0.9249,
+      "step": 9495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180932687712164e-06,
+      "loss": 0.9563,
+      "step": 9496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17892803914872e-06,
+      "loss": 0.9161,
+      "step": 9497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.176923570376254e-06,
+      "loss": 0.9602,
+      "step": 9498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174919281489084e-06,
+      "loss": 0.95,
+      "step": 9499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172915172581514e-06,
+      "loss": 0.9797,
+      "step": 9500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170911243747846e-06,
+      "loss": 0.8746,
+      "step": 9501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.168907495082368e-06,
+      "loss": 0.9424,
+      "step": 9502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16690392667936e-06,
+      "loss": 0.9482,
+      "step": 9503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1649005386331005e-06,
+      "loss": 0.9218,
+      "step": 9504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1628973310378425e-06,
+      "loss": 0.954,
+      "step": 9505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160894303987856e-06,
+      "loss": 0.9562,
+      "step": 9506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.158891457577379e-06,
+      "loss": 0.9557,
+      "step": 9507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.15688879190065e-06,
+      "loss": 0.9381,
+      "step": 9508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154886307051906e-06,
+      "loss": 0.9492,
+      "step": 9509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152884003125368e-06,
+      "loss": 1.0322,
+      "step": 9510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150881880215238e-06,
+      "loss": 0.9398,
+      "step": 9511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148879938415737e-06,
+      "loss": 0.9124,
+      "step": 9512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.146878177821047e-06,
+      "loss": 0.9122,
+      "step": 9513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144876598525371e-06,
+      "loss": 0.9511,
+      "step": 9514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.142875200622877e-06,
+      "loss": 0.9046,
+      "step": 9515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140873984207736e-06,
+      "loss": 1.0168,
+      "step": 9516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138872949374115e-06,
+      "loss": 0.9531,
+      "step": 9517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1368720962161646e-06,
+      "loss": 0.9208,
+      "step": 9518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134871424828033e-06,
+      "loss": 1.0196,
+      "step": 9519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1328709353038564e-06,
+      "loss": 0.9881,
+      "step": 9520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130870627737755e-06,
+      "loss": 0.9029,
+      "step": 9521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.128870502223861e-06,
+      "loss": 0.9266,
+      "step": 9522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126870558856275e-06,
+      "loss": 0.959,
+      "step": 9523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124870797729106e-06,
+      "loss": 1.046,
+      "step": 9524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122871218936444e-06,
+      "loss": 1.0114,
+      "step": 9525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120871822572373e-06,
+      "loss": 0.9876,
+      "step": 9526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118872608730974e-06,
+      "loss": 1.058,
+      "step": 9527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1168735775063106e-06,
+      "loss": 0.9714,
+      "step": 9528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1148747289924485e-06,
+      "loss": 0.9316,
+      "step": 9529
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112876063283433e-06,
+      "loss": 0.9682,
+      "step": 9530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110877580473305e-06,
+      "loss": 1.0132,
+      "step": 9531
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.108879280656102e-06,
+      "loss": 0.9894,
+      "step": 9532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106881163925846e-06,
+      "loss": 1.0232,
+      "step": 9533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.104883230376557e-06,
+      "loss": 1.0008,
+      "step": 9534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102885480102243e-06,
+      "loss": 0.9162,
+      "step": 9535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.100887913196897e-06,
+      "loss": 0.9479,
+      "step": 9536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0988905297545175e-06,
+      "loss": 0.9474,
+      "step": 9537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096893329869078e-06,
+      "loss": 0.9646,
+      "step": 9538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.094896313634562e-06,
+      "loss": 1.0333,
+      "step": 9539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092899481144925e-06,
+      "loss": 1.0056,
+      "step": 9540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090902832494125e-06,
+      "loss": 0.9587,
+      "step": 9541
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088906367776113e-06,
+      "loss": 0.9226,
+      "step": 9542
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086910087084822e-06,
+      "loss": 0.8971,
+      "step": 9543
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.084913990514187e-06,
+      "loss": 0.9684,
+      "step": 9544
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082918078158131e-06,
+      "loss": 1.037,
+      "step": 9545
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080922350110556e-06,
+      "loss": 0.951,
+      "step": 9546
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078926806465381e-06,
+      "loss": 0.9636,
+      "step": 9547
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.07693144731649e-06,
+      "loss": 0.9294,
+      "step": 9548
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074936272757771e-06,
+      "loss": 0.9177,
+      "step": 9549
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072941282883107e-06,
+      "loss": 0.9103,
+      "step": 9550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070946477786361e-06,
+      "loss": 0.9686,
+      "step": 9551
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068951857561401e-06,
+      "loss": 0.9566,
+      "step": 9552
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.066957422302075e-06,
+      "loss": 0.9282,
+      "step": 9553
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064963172102223e-06,
+      "loss": 0.989,
+      "step": 9554
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.062969107055683e-06,
+      "loss": 0.9326,
+      "step": 9555
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060975227256278e-06,
+      "loss": 1.0096,
+      "step": 9556
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058981532797831e-06,
+      "loss": 0.9069,
+      "step": 9557
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056988023774145e-06,
+      "loss": 0.9575,
+      "step": 9558
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.05499470027902e-06,
+      "loss": 0.9512,
+      "step": 9559
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.053001562406249e-06,
+      "loss": 1.0096,
+      "step": 9560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051008610249611e-06,
+      "loss": 0.9186,
+      "step": 9561
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0490158439028855e-06,
+      "loss": 0.9437,
+      "step": 9562
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047023263459832e-06,
+      "loss": 0.9807,
+      "step": 9563
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0450308690142045e-06,
+      "loss": 0.9719,
+      "step": 9564
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0430386606597544e-06,
+      "loss": 0.9295,
+      "step": 9565
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.041046638490218e-06,
+      "loss": 1.002,
+      "step": 9566
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.039054802599327e-06,
+      "loss": 0.8921,
+      "step": 9567
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037063153080803e-06,
+      "loss": 0.8527,
+      "step": 9568
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.035071690028349e-06,
+      "loss": 0.9936,
+      "step": 9569
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033080413535684e-06,
+      "loss": 0.9916,
+      "step": 9570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031089323696484e-06,
+      "loss": 0.8161,
+      "step": 9571
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.029098420604453e-06,
+      "loss": 0.8638,
+      "step": 9572
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027107704353257e-06,
+      "loss": 0.968,
+      "step": 9573
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025117175036566e-06,
+      "loss": 0.9755,
+      "step": 9574
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.023126832748039e-06,
+      "loss": 1.0095,
+      "step": 9575
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021136677581327e-06,
+      "loss": 0.9746,
+      "step": 9576
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019146709630075e-06,
+      "loss": 0.9867,
+      "step": 9577
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017156928987915e-06,
+      "loss": 0.9151,
+      "step": 9578
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0151673357484665e-06,
+      "loss": 0.8718,
+      "step": 9579
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0131779300053495e-06,
+      "loss": 1.0086,
+      "step": 9580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011188711852165e-06,
+      "loss": 0.941,
+      "step": 9581
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009199681382518e-06,
+      "loss": 0.9922,
+      "step": 9582
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007210838689994e-06,
+      "loss": 0.9335,
+      "step": 9583
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005222183868171e-06,
+      "loss": 0.9398,
+      "step": 9584
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.003233717010623e-06,
+      "loss": 0.9964,
+      "step": 9585
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.001245438210914e-06,
+      "loss": 0.9613,
+      "step": 9586
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.999257347562592e-06,
+      "loss": 1.0036,
+      "step": 9587
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997269445159206e-06,
+      "loss": 0.9698,
+      "step": 9588
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9952817310942865e-06,
+      "loss": 0.9382,
+      "step": 9589
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.993294205461369e-06,
+      "loss": 1.0024,
+      "step": 9590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991306868353965e-06,
+      "loss": 0.9349,
+      "step": 9591
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989319719865584e-06,
+      "loss": 0.9861,
+      "step": 9592
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.987332760089731e-06,
+      "loss": 0.964,
+      "step": 9593
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985345989119888e-06,
+      "loss": 0.9337,
+      "step": 9594
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983359407049551e-06,
+      "loss": 0.9758,
+      "step": 9595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.981373013972185e-06,
+      "loss": 0.8684,
+      "step": 9596
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.979386809981253e-06,
+      "loss": 0.9157,
+      "step": 9597
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977400795170215e-06,
+      "loss": 0.9584,
+      "step": 9598
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975414969632516e-06,
+      "loss": 0.9499,
+      "step": 9599
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9734293334615955e-06,
+      "loss": 1.0259,
+      "step": 9600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9714438867508865e-06,
+      "loss": 0.9065,
+      "step": 9601
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969458629593797e-06,
+      "loss": 0.9635,
+      "step": 9602
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.967473562083753e-06,
+      "loss": 0.8648,
+      "step": 9603
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.965488684314143e-06,
+      "loss": 0.943,
+      "step": 9604
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.963503996378375e-06,
+      "loss": 0.984,
+      "step": 9605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961519498369823e-06,
+      "loss": 0.9532,
+      "step": 9606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959535190381865e-06,
+      "loss": 0.9801,
+      "step": 9607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.957551072507868e-06,
+      "loss": 0.8493,
+      "step": 9608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95556714484119e-06,
+      "loss": 0.9716,
+      "step": 9609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953583407475181e-06,
+      "loss": 0.9381,
+      "step": 9610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9515998605031814e-06,
+      "loss": 0.9286,
+      "step": 9611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949616504018515e-06,
+      "loss": 0.9988,
+      "step": 9612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947633338114511e-06,
+      "loss": 0.9386,
+      "step": 9613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94565036288448e-06,
+      "loss": 0.9476,
+      "step": 9614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.943667578421727e-06,
+      "loss": 0.9354,
+      "step": 9615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941684984819546e-06,
+      "loss": 0.9319,
+      "step": 9616
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.93970258217122e-06,
+      "loss": 0.9978,
+      "step": 9617
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937720370570033e-06,
+      "loss": 0.9421,
+      "step": 9618
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.935738350109245e-06,
+      "loss": 0.964,
+      "step": 9619
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933756520882124e-06,
+      "loss": 0.9429,
+      "step": 9620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9317748829819135e-06,
+      "loss": 0.9414,
+      "step": 9621
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929793436501854e-06,
+      "loss": 0.9487,
+      "step": 9622
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927812181535181e-06,
+      "loss": 0.8942,
+      "step": 9623
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925831118175116e-06,
+      "loss": 0.8921,
+      "step": 9624
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.923850246514871e-06,
+      "loss": 0.9719,
+      "step": 9625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.921869566647657e-06,
+      "loss": 0.9716,
+      "step": 9626
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919889078666661e-06,
+      "loss": 0.9682,
+      "step": 9627
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917908782665078e-06,
+      "loss": 0.9453,
+      "step": 9628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915928678736084e-06,
+      "loss": 1.006,
+      "step": 9629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913948766972843e-06,
+      "loss": 0.9242,
+      "step": 9630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.91196904746852e-06,
+      "loss": 0.9746,
+      "step": 9631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909989520316263e-06,
+      "loss": 0.9389,
+      "step": 9632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.908010185609217e-06,
+      "loss": 0.9772,
+      "step": 9633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906031043440512e-06,
+      "loss": 0.9543,
+      "step": 9634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904052093903269e-06,
+      "loss": 0.8903,
+      "step": 9635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9020733370906104e-06,
+      "loss": 1.011,
+      "step": 9636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900094773095632e-06,
+      "loss": 0.997,
+      "step": 9637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.898116402011438e-06,
+      "loss": 0.9892,
+      "step": 9638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896138223931112e-06,
+      "loss": 0.9891,
+      "step": 9639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.89416023894773e-06,
+      "loss": 0.9308,
+      "step": 9640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892182447154368e-06,
+      "loss": 0.8668,
+      "step": 9641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8902048486440785e-06,
+      "loss": 0.9448,
+      "step": 9642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.888227443509918e-06,
+      "loss": 0.8736,
+      "step": 9643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88625023184493e-06,
+      "loss": 0.9013,
+      "step": 9644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884273213742138e-06,
+      "loss": 0.9432,
+      "step": 9645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8822963892945726e-06,
+      "loss": 0.9031,
+      "step": 9646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8803197585952475e-06,
+      "loss": 0.8586,
+      "step": 9647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878343321737168e-06,
+      "loss": 1.012,
+      "step": 9648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.87636707881333e-06,
+      "loss": 0.834,
+      "step": 9649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.874391029916719e-06,
+      "loss": 0.9193,
+      "step": 9650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872415175140318e-06,
+      "loss": 0.9963,
+      "step": 9651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870439514577087e-06,
+      "loss": 0.9315,
+      "step": 9652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868464048319997e-06,
+      "loss": 0.9194,
+      "step": 9653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866488776461992e-06,
+      "loss": 1.006,
+      "step": 9654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8645136990960105e-06,
+      "loss": 1.0366,
+      "step": 9655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.862538816314991e-06,
+      "loss": 0.9489,
+      "step": 9656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8605641282118545e-06,
+      "loss": 0.9376,
+      "step": 9657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8585896348795145e-06,
+      "loss": 0.933,
+      "step": 9658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85661533641088e-06,
+      "loss": 0.9292,
+      "step": 9659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854641232898837e-06,
+      "loss": 0.925,
+      "step": 9660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852667324436283e-06,
+      "loss": 0.9605,
+      "step": 9661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850693611116088e-06,
+      "loss": 1.0117,
+      "step": 9662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.848720093031122e-06,
+      "loss": 0.9991,
+      "step": 9663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846746770274244e-06,
+      "loss": 1.0238,
+      "step": 9664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.844773642938304e-06,
+      "loss": 0.9752,
+      "step": 9665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842800711116143e-06,
+      "loss": 0.9628,
+      "step": 9666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840827974900596e-06,
+      "loss": 0.9427,
+      "step": 9667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838855434384474e-06,
+      "loss": 0.9233,
+      "step": 9668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836883089660602e-06,
+      "loss": 0.9997,
+      "step": 9669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834910940821779e-06,
+      "loss": 1.0012,
+      "step": 9670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8329389879607985e-06,
+      "loss": 0.9667,
+      "step": 9671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8309672311704476e-06,
+      "loss": 0.9617,
+      "step": 9672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.828995670543501e-06,
+      "loss": 1.0648,
+      "step": 9673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827024306172726e-06,
+      "loss": 0.8759,
+      "step": 9674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8250531381508765e-06,
+      "loss": 0.9367,
+      "step": 9675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82308216657071e-06,
+      "loss": 0.9478,
+      "step": 9676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821111391524958e-06,
+      "loss": 1.0456,
+      "step": 9677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.819140813106354e-06,
+      "loss": 0.9526,
+      "step": 9678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.817170431407617e-06,
+      "loss": 0.9371,
+      "step": 9679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815200246521456e-06,
+      "loss": 0.9684,
+      "step": 9680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.813230258540583e-06,
+      "loss": 0.8974,
+      "step": 9681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811260467557679e-06,
+      "loss": 0.942,
+      "step": 9682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809290873665429e-06,
+      "loss": 0.9148,
+      "step": 9683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.807321476956514e-06,
+      "loss": 1.032,
+      "step": 9684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805352277523597e-06,
+      "loss": 0.943,
+      "step": 9685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803383275459331e-06,
+      "loss": 0.97,
+      "step": 9686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801414470856363e-06,
+      "loss": 0.9437,
+      "step": 9687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799445863807329e-06,
+      "loss": 0.8861,
+      "step": 9688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797477454404867e-06,
+      "loss": 0.9446,
+      "step": 9689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.795509242741579e-06,
+      "loss": 0.9904,
+      "step": 9690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793541228910087e-06,
+      "loss": 0.9862,
+      "step": 9691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7915734130029855e-06,
+      "loss": 0.9188,
+      "step": 9692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789605795112867e-06,
+      "loss": 0.8996,
+      "step": 9693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7876383753323125e-06,
+      "loss": 0.9835,
+      "step": 9694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78567115375389e-06,
+      "loss": 0.9433,
+      "step": 9695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783704130470169e-06,
+      "loss": 0.9498,
+      "step": 9696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.781737305573705e-06,
+      "loss": 1.0228,
+      "step": 9697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779770679157028e-06,
+      "loss": 0.9297,
+      "step": 9698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777804251312687e-06,
+      "loss": 0.9344,
+      "step": 9699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7758380221332e-06,
+      "loss": 0.9742,
+      "step": 9700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773871991711084e-06,
+      "loss": 0.9832,
+      "step": 9701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771906160138848e-06,
+      "loss": 0.9457,
+      "step": 9702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7699405275089835e-06,
+      "loss": 0.9021,
+      "step": 9703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7679750939139865e-06,
+      "loss": 0.9245,
+      "step": 9704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.766009859446338e-06,
+      "loss": 1.0111,
+      "step": 9705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.764044824198489e-06,
+      "loss": 0.9833,
+      "step": 9706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.762079988262918e-06,
+      "loss": 0.9912,
+      "step": 9707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.760115351732067e-06,
+      "loss": 1.0078,
+      "step": 9708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.75815091469838e-06,
+      "loss": 0.9669,
+      "step": 9709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756186677254288e-06,
+      "loss": 0.8915,
+      "step": 9710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754222639492207e-06,
+      "loss": 0.9227,
+      "step": 9711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752258801504565e-06,
+      "loss": 0.9875,
+      "step": 9712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750295163383747e-06,
+      "loss": 0.9357,
+      "step": 9713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74833172522216e-06,
+      "loss": 0.9619,
+      "step": 9714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7463684871121864e-06,
+      "loss": 0.8968,
+      "step": 9715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744405449146197e-06,
+      "loss": 1.0039,
+      "step": 9716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.742442611416563e-06,
+      "loss": 0.9395,
+      "step": 9717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740479974015634e-06,
+      "loss": 0.9689,
+      "step": 9718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.738517537035765e-06,
+      "loss": 0.9939,
+      "step": 9719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736555300569293e-06,
+      "loss": 0.9227,
+      "step": 9720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7345932647085365e-06,
+      "loss": 0.9515,
+      "step": 9721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732631429545824e-06,
+      "loss": 0.9322,
+      "step": 9722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.730669795173458e-06,
+      "loss": 1.0428,
+      "step": 9723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.728708361683748e-06,
+      "loss": 0.9361,
+      "step": 9724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.726747129168976e-06,
+      "loss": 0.9496,
+      "step": 9725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.72478609772142e-06,
+      "loss": 0.9442,
+      "step": 9726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722825267433361e-06,
+      "loss": 0.9963,
+      "step": 9727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.720864638397055e-06,
+      "loss": 0.9362,
+      "step": 9728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.718904210704756e-06,
+      "loss": 0.9856,
+      "step": 9729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.716943984448709e-06,
+      "loss": 0.8947,
+      "step": 9730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.714983959721143e-06,
+      "loss": 0.9037,
+      "step": 9731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713024136614287e-06,
+      "loss": 0.9113,
+      "step": 9732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711064515220346e-06,
+      "loss": 0.947,
+      "step": 9733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709105095631537e-06,
+      "loss": 0.8718,
+      "step": 9734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707145877940053e-06,
+      "loss": 0.907,
+      "step": 9735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7051868622380756e-06,
+      "loss": 0.9828,
+      "step": 9736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7032280486177835e-06,
+      "loss": 0.945,
+      "step": 9737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701269437171345e-06,
+      "loss": 0.9437,
+      "step": 9738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.699311027990916e-06,
+      "loss": 0.9375,
+      "step": 9739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697352821168645e-06,
+      "loss": 0.979,
+      "step": 9740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695394816796666e-06,
+      "loss": 0.9767,
+      "step": 9741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.693437014967118e-06,
+      "loss": 0.9707,
+      "step": 9742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691479415772115e-06,
+      "loss": 0.9926,
+      "step": 9743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6895220193037645e-06,
+      "loss": 0.9856,
+      "step": 9744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687564825654171e-06,
+      "loss": 0.9207,
+      "step": 9745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685607834915421e-06,
+      "loss": 0.9839,
+      "step": 9746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6836510471796035e-06,
+      "loss": 0.9585,
+      "step": 9747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681694462538784e-06,
+      "loss": 0.9218,
+      "step": 9748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.67973808108502e-06,
+      "loss": 0.8988,
+      "step": 9749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677781902910376e-06,
+      "loss": 0.919,
+      "step": 9750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675825928106889e-06,
+      "loss": 0.914,
+      "step": 9751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673870156766591e-06,
+      "loss": 0.898,
+      "step": 9752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671914588981509e-06,
+      "loss": 1.0265,
+      "step": 9753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669959224843651e-06,
+      "loss": 0.874,
+      "step": 9754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668004064445034e-06,
+      "loss": 0.9728,
+      "step": 9755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6660491078776405e-06,
+      "loss": 0.9432,
+      "step": 9756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.664094355233462e-06,
+      "loss": 0.9198,
+      "step": 9757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.662139806604477e-06,
+      "loss": 0.856,
+      "step": 9758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660185462082648e-06,
+      "loss": 0.9585,
+      "step": 9759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658231321759932e-06,
+      "loss": 1.0098,
+      "step": 9760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656277385728273e-06,
+      "loss": 0.9178,
+      "step": 9761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654323654079617e-06,
+      "loss": 0.9045,
+      "step": 9762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6523701269058904e-06,
+      "loss": 0.9369,
+      "step": 9763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6504168042990015e-06,
+      "loss": 0.8869,
+      "step": 9764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.648463686350869e-06,
+      "loss": 0.9713,
+      "step": 9765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646510773153389e-06,
+      "loss": 0.9331,
+      "step": 9766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644558064798452e-06,
+      "loss": 0.9087,
+      "step": 9767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642605561377936e-06,
+      "loss": 0.8995,
+      "step": 9768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640653262983709e-06,
+      "loss": 0.9358,
+      "step": 9769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638701169707641e-06,
+      "loss": 0.843,
+      "step": 9770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.636749281641569e-06,
+      "loss": 1.0102,
+      "step": 9771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6347975988773455e-06,
+      "loss": 0.9891,
+      "step": 9772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632846121506796e-06,
+      "loss": 0.9704,
+      "step": 9773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630894849621748e-06,
+      "loss": 0.9497,
+      "step": 9774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628943783314007e-06,
+      "loss": 1.0164,
+      "step": 9775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.626992922675381e-06,
+      "loss": 0.9237,
+      "step": 9776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625042267797655e-06,
+      "loss": 0.9609,
+      "step": 9777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6230918187726255e-06,
+      "loss": 0.9559,
+      "step": 9778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6211415756920505e-06,
+      "loss": 0.9449,
+      "step": 9779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6191915386477065e-06,
+      "loss": 0.9448,
+      "step": 9780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617241707731343e-06,
+      "loss": 0.9356,
+      "step": 9781
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6152920830347025e-06,
+      "loss": 0.9374,
+      "step": 9782
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.613342664649522e-06,
+      "loss": 0.8403,
+      "step": 9783
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611393452667522e-06,
+      "loss": 1.0014,
+      "step": 9784
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609444447180426e-06,
+      "loss": 0.9737,
+      "step": 9785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.607495648279937e-06,
+      "loss": 0.9563,
+      "step": 9786
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605547056057743e-06,
+      "loss": 0.9216,
+      "step": 9787
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603598670605539e-06,
+      "loss": 0.9391,
+      "step": 9788
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601650492014998e-06,
+      "loss": 0.9406,
+      "step": 9789
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.59970252037779e-06,
+      "loss": 0.9123,
+      "step": 9790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5977547557855675e-06,
+      "loss": 0.892,
+      "step": 9791
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595807198329974e-06,
+      "loss": 0.9291,
+      "step": 9792
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593859848102657e-06,
+      "loss": 0.9597,
+      "step": 9793
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.591912705195241e-06,
+      "loss": 0.9046,
+      "step": 9794
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589965769699341e-06,
+      "loss": 0.9713,
+      "step": 9795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588019041706567e-06,
+      "loss": 0.9773,
+      "step": 9796
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586072521308517e-06,
+      "loss": 0.9368,
+      "step": 9797
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5841262085967786e-06,
+      "loss": 0.9408,
+      "step": 9798
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582180103662929e-06,
+      "loss": 1.0079,
+      "step": 9799
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5802342065985424e-06,
+      "loss": 0.9005,
+      "step": 9800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.578288517495177e-06,
+      "loss": 0.9819,
+      "step": 9801
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5763430364443816e-06,
+      "loss": 0.8069,
+      "step": 9802
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.574397763537695e-06,
+      "loss": 0.9153,
+      "step": 9803
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572452698866642e-06,
+      "loss": 0.9417,
+      "step": 9804
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570507842522759e-06,
+      "loss": 0.936,
+      "step": 9805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568563194597539e-06,
+      "loss": 0.9228,
+      "step": 9806
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566618755182487e-06,
+      "loss": 0.9192,
+      "step": 9807
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564674524369098e-06,
+      "loss": 0.9547,
+      "step": 9808
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562730502248852e-06,
+      "loss": 0.9419,
+      "step": 9809
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.56078668891322e-06,
+      "loss": 1.0194,
+      "step": 9810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.558843084453662e-06,
+      "loss": 0.9718,
+      "step": 9811
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556899688961624e-06,
+      "loss": 0.9953,
+      "step": 9812
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554956502528561e-06,
+      "loss": 0.9726,
+      "step": 9813
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.553013525245895e-06,
+      "loss": 0.9212,
+      "step": 9814
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5510707572050444e-06,
+      "loss": 0.9542,
+      "step": 9815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549128198497432e-06,
+      "loss": 0.9427,
+      "step": 9816
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547185849214455e-06,
+      "loss": 0.9599,
+      "step": 9817
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.545243709447505e-06,
+      "loss": 0.9154,
+      "step": 9818
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5433017792879655e-06,
+      "loss": 0.9555,
+      "step": 9819
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5413600588272045e-06,
+      "loss": 0.9471,
+      "step": 9820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539418548156597e-06,
+      "loss": 0.8603,
+      "step": 9821
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53747724736748e-06,
+      "loss": 1.0454,
+      "step": 9822
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.535536156551209e-06,
+      "loss": 1.0169,
+      "step": 9823
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533595275799113e-06,
+      "loss": 0.9014,
+      "step": 9824
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531654605202516e-06,
+      "loss": 0.9327,
+      "step": 9825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5297141448527295e-06,
+      "loss": 0.9446,
+      "step": 9826
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.527773894841053e-06,
+      "loss": 0.9519,
+      "step": 9827
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.525833855258794e-06,
+      "loss": 0.9542,
+      "step": 9828
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523894026197223e-06,
+      "loss": 0.9918,
+      "step": 9829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.521954407747615e-06,
+      "loss": 0.9665,
+      "step": 9830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520015000001241e-06,
+      "loss": 0.9104,
+      "step": 9831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51807580304935e-06,
+      "loss": 0.9604,
+      "step": 9832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5161368169831886e-06,
+      "loss": 0.9294,
+      "step": 9833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.514198041893988e-06,
+      "loss": 1.0235,
+      "step": 9834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51225947787297e-06,
+      "loss": 0.9488,
+      "step": 9835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510321125011361e-06,
+      "loss": 1.0025,
+      "step": 9836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508382983400349e-06,
+      "loss": 0.9473,
+      "step": 9837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5064450531311395e-06,
+      "loss": 0.9316,
+      "step": 9838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504507334294915e-06,
+      "loss": 1.0128,
+      "step": 9839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.502569826982849e-06,
+      "loss": 0.9343,
+      "step": 9840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5006325312861056e-06,
+      "loss": 1.0307,
+      "step": 9841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498695447295836e-06,
+      "loss": 0.9445,
+      "step": 9842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.496758575103193e-06,
+      "loss": 0.9668,
+      "step": 9843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49482191479931e-06,
+      "loss": 0.9981,
+      "step": 9844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492885466475301e-06,
+      "loss": 1.0374,
+      "step": 9845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490949230222291e-06,
+      "loss": 0.8794,
+      "step": 9846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.489013206131384e-06,
+      "loss": 1.0016,
+      "step": 9847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487077394293672e-06,
+      "loss": 0.9562,
+      "step": 9848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.48514179480024e-06,
+      "loss": 0.9264,
+      "step": 9849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483206407742161e-06,
+      "loss": 1.0161,
+      "step": 9850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481271233210506e-06,
+      "loss": 0.955,
+      "step": 9851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.479336271296328e-06,
+      "loss": 0.8964,
+      "step": 9852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.477401522090664e-06,
+      "loss": 0.9422,
+      "step": 9853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475466985684556e-06,
+      "loss": 0.9338,
+      "step": 9854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.47353266216903e-06,
+      "loss": 0.9358,
+      "step": 9855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471598551635097e-06,
+      "loss": 0.9278,
+      "step": 9856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.469664654173763e-06,
+      "loss": 0.9699,
+      "step": 9857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.467730969876019e-06,
+      "loss": 0.8975,
+      "step": 9858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4657974988328564e-06,
+      "loss": 0.9371,
+      "step": 9859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463864241135246e-06,
+      "loss": 0.9684,
+      "step": 9860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.461931196874154e-06,
+      "loss": 0.9568,
+      "step": 9861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459998366140534e-06,
+      "loss": 1.0351,
+      "step": 9862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458065749025329e-06,
+      "loss": 0.9981,
+      "step": 9863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456133345619479e-06,
+      "loss": 0.8779,
+      "step": 9864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.454201156013896e-06,
+      "loss": 0.9599,
+      "step": 9865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45226918029951e-06,
+      "loss": 0.9046,
+      "step": 9866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45033741856722e-06,
+      "loss": 0.9491,
+      "step": 9867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4484058709079114e-06,
+      "loss": 0.9077,
+      "step": 9868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4464745374124785e-06,
+      "loss": 0.9317,
+      "step": 9869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.444543418171788e-06,
+      "loss": 0.992,
+      "step": 9870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442612513276716e-06,
+      "loss": 0.9265,
+      "step": 9871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440681822818103e-06,
+      "loss": 0.8856,
+      "step": 9872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438751346886797e-06,
+      "loss": 1.0145,
+      "step": 9873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436821085573636e-06,
+      "loss": 0.864,
+      "step": 9874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434891038969439e-06,
+      "loss": 0.8836,
+      "step": 9875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432961207165021e-06,
+      "loss": 0.9785,
+      "step": 9876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431031590251188e-06,
+      "loss": 0.9031,
+      "step": 9877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.429102188318728e-06,
+      "loss": 0.9269,
+      "step": 9878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.427173001458428e-06,
+      "loss": 0.8979,
+      "step": 9879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425244029761055e-06,
+      "loss": 1.0099,
+      "step": 9880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4233152733173825e-06,
+      "loss": 1.0026,
+      "step": 9881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421386732218157e-06,
+      "loss": 1.0025,
+      "step": 9882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419458406554121e-06,
+      "loss": 0.9291,
+      "step": 9883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417530296416008e-06,
+      "loss": 0.9052,
+      "step": 9884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415602401894542e-06,
+      "loss": 1.0301,
+      "step": 9885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413674723080431e-06,
+      "loss": 0.8905,
+      "step": 9886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411747260064381e-06,
+      "loss": 0.9576,
+      "step": 9887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409820012937079e-06,
+      "loss": 0.9086,
+      "step": 9888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407892981789214e-06,
+      "loss": 0.966,
+      "step": 9889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405966166711454e-06,
+      "loss": 0.9499,
+      "step": 9890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.404039567794459e-06,
+      "loss": 0.968,
+      "step": 9891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402113185128883e-06,
+      "loss": 1.0164,
+      "step": 9892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400187018805363e-06,
+      "loss": 1.0714,
+      "step": 9893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.398261068914538e-06,
+      "loss": 0.9056,
+      "step": 9894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.396335335547023e-06,
+      "loss": 0.9229,
+      "step": 9895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.394409818793424e-06,
+      "loss": 0.9693,
+      "step": 9896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39248451874435e-06,
+      "loss": 0.9758,
+      "step": 9897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390559435490386e-06,
+      "loss": 0.9499,
+      "step": 9898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388634569122117e-06,
+      "loss": 0.9532,
+      "step": 9899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3867099197301065e-06,
+      "loss": 0.8959,
+      "step": 9900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384785487404914e-06,
+      "loss": 0.8916,
+      "step": 9901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382861272237101e-06,
+      "loss": 0.9571,
+      "step": 9902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.380937274317188e-06,
+      "loss": 0.9436,
+      "step": 9903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379013493735719e-06,
+      "loss": 0.9104,
+      "step": 9904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3770899305832045e-06,
+      "loss": 0.9774,
+      "step": 9905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.375166584950158e-06,
+      "loss": 0.9538,
+      "step": 9906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373243456927073e-06,
+      "loss": 1.0017,
+      "step": 9907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371320546604437e-06,
+      "loss": 0.934,
+      "step": 9908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3693978540727346e-06,
+      "loss": 1.032,
+      "step": 9909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3674753794224336e-06,
+      "loss": 0.9603,
+      "step": 9910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365553122743976e-06,
+      "loss": 0.9372,
+      "step": 9911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.363631084127826e-06,
+      "loss": 1.0322,
+      "step": 9912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3617092636644144e-06,
+      "loss": 0.9287,
+      "step": 9913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3597876614441645e-06,
+      "loss": 0.9428,
+      "step": 9914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357866277557497e-06,
+      "loss": 0.9699,
+      "step": 9915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355945112094812e-06,
+      "loss": 0.9071,
+      "step": 9916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3540241651465115e-06,
+      "loss": 0.9632,
+      "step": 9917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35210343680298e-06,
+      "loss": 0.982,
+      "step": 9918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35018292715459e-06,
+      "loss": 0.9806,
+      "step": 9919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348262636291708e-06,
+      "loss": 0.9193,
+      "step": 9920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346342564304688e-06,
+      "loss": 1.0292,
+      "step": 9921
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344422711283872e-06,
+      "loss": 0.9527,
+      "step": 9922
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.342503077319597e-06,
+      "loss": 0.9664,
+      "step": 9923
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340583662502181e-06,
+      "loss": 0.8999,
+      "step": 9924
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.33866446692195e-06,
+      "loss": 0.8932,
+      "step": 9925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.33674549066919e-06,
+      "loss": 0.9476,
+      "step": 9926
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334826733834206e-06,
+      "loss": 0.9824,
+      "step": 9927
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332908196507277e-06,
+      "loss": 1.0578,
+      "step": 9928
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.330989878778674e-06,
+      "loss": 0.9248,
+      "step": 9929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329071780738659e-06,
+      "loss": 0.9676,
+      "step": 9930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.327153902477481e-06,
+      "loss": 0.9622,
+      "step": 9931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325236244085385e-06,
+      "loss": 0.904,
+      "step": 9932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3233188056526065e-06,
+      "loss": 0.8685,
+      "step": 9933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32140158726935e-06,
+      "loss": 0.9946,
+      "step": 9934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319484589025838e-06,
+      "loss": 0.9987,
+      "step": 9935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3175678110122675e-06,
+      "loss": 0.9548,
+      "step": 9936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3156512533188265e-06,
+      "loss": 0.9137,
+      "step": 9937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3137349160356945e-06,
+      "loss": 0.9573,
+      "step": 9938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311818799253035e-06,
+      "loss": 1.0071,
+      "step": 9939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.309902903061015e-06,
+      "loss": 0.8964,
+      "step": 9940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307987227549779e-06,
+      "loss": 0.934,
+      "step": 9941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306071772809462e-06,
+      "loss": 0.9551,
+      "step": 9942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.304156538930193e-06,
+      "loss": 0.9789,
+      "step": 9943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.302241526002088e-06,
+      "loss": 0.998,
+      "step": 9944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300326734115252e-06,
+      "loss": 0.9241,
+      "step": 9945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298412163359778e-06,
+      "loss": 0.9724,
+      "step": 9946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2964978138257585e-06,
+      "loss": 0.9099,
+      "step": 9947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2945836856032654e-06,
+      "loss": 0.8624,
+      "step": 9948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292669778782363e-06,
+      "loss": 0.8984,
+      "step": 9949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290756093453104e-06,
+      "loss": 0.8977,
+      "step": 9950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288842629705531e-06,
+      "loss": 0.913,
+      "step": 9951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286929387629687e-06,
+      "loss": 0.97,
+      "step": 9952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.285016367315582e-06,
+      "loss": 0.8633,
+      "step": 9953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2831035688532295e-06,
+      "loss": 0.9074,
+      "step": 9954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281190992332641e-06,
+      "loss": 0.9858,
+      "step": 9955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2792786378438014e-06,
+      "loss": 0.9333,
+      "step": 9956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277366505476693e-06,
+      "loss": 1.0489,
+      "step": 9957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2754545953212875e-06,
+      "loss": 0.9585,
+      "step": 9958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27354290746754e-06,
+      "loss": 0.9622,
+      "step": 9959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271631442005409e-06,
+      "loss": 0.9876,
+      "step": 9960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.269720199024828e-06,
+      "loss": 0.9827,
+      "step": 9961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267809178615722e-06,
+      "loss": 0.8755,
+      "step": 9962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.265898380868018e-06,
+      "loss": 0.9409,
+      "step": 9963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.263987805871618e-06,
+      "loss": 0.99,
+      "step": 9964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262077453716421e-06,
+      "loss": 0.9863,
+      "step": 9965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2601673244923156e-06,
+      "loss": 0.967,
+      "step": 9966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258257418289172e-06,
+      "loss": 1.0207,
+      "step": 9967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256347735196867e-06,
+      "loss": 0.9681,
+      "step": 9968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.254438275305243e-06,
+      "loss": 0.9925,
+      "step": 9969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.252529038704155e-06,
+      "loss": 0.9948,
+      "step": 9970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2506200254834325e-06,
+      "loss": 0.9934,
+      "step": 9971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248711235732901e-06,
+      "loss": 0.9579,
+      "step": 9972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.246802669542373e-06,
+      "loss": 0.9949,
+      "step": 9973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244894327001648e-06,
+      "loss": 0.9401,
+      "step": 9974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242986208200529e-06,
+      "loss": 0.9824,
+      "step": 9975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241078313228788e-06,
+      "loss": 0.9657,
+      "step": 9976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2391706421761945e-06,
+      "loss": 1.0094,
+      "step": 9977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237263195132516e-06,
+      "loss": 1.0182,
+      "step": 9978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235355972187502e-06,
+      "loss": 0.9581,
+      "step": 9979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.23344897343089e-06,
+      "loss": 0.9351,
+      "step": 9980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231542198952409e-06,
+      "loss": 0.95,
+      "step": 9981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229635648841773e-06,
+      "loss": 0.8947,
+      "step": 9982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227729323188704e-06,
+      "loss": 1.0227,
+      "step": 9983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.225823222082881e-06,
+      "loss": 0.9208,
+      "step": 9984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223917345614005e-06,
+      "loss": 1.0321,
+      "step": 9985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222011693871746e-06,
+      "loss": 0.9418,
+      "step": 9986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2201062669457714e-06,
+      "loss": 0.9303,
+      "step": 9987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218201064925737e-06,
+      "loss": 0.9509,
+      "step": 9988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2162960879012806e-06,
+      "loss": 1.0192,
+      "step": 9989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214391335962046e-06,
+      "loss": 1.01,
+      "step": 9990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2124868091976555e-06,
+      "loss": 0.9303,
+      "step": 9991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210582507697711e-06,
+      "loss": 0.8843,
+      "step": 9992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208678431551826e-06,
+      "loss": 0.9812,
+      "step": 9993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.20677458084959e-06,
+      "loss": 0.9665,
+      "step": 9994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204870955680579e-06,
+      "loss": 0.998,
+      "step": 9995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202967556134369e-06,
+      "loss": 0.8967,
+      "step": 9996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201064382300511e-06,
+      "loss": 0.9615,
+      "step": 9997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199161434268566e-06,
+      "loss": 0.944,
+      "step": 9998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.19725871212807e-06,
+      "loss": 0.8887,
+      "step": 9999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1953562159685385e-06,
+      "loss": 1.059,
+      "step": 10000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193453945879503e-06,
+      "loss": 0.9461,
+      "step": 10001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1915519019504644e-06,
+      "loss": 0.9592,
+      "step": 10002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1896500842709195e-06,
+      "loss": 1.0186,
+      "step": 10003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1877484929303535e-06,
+      "loss": 0.9764,
+      "step": 10004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.185847128018235e-06,
+      "loss": 0.7999,
+      "step": 10005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183945989624038e-06,
+      "loss": 0.9889,
+      "step": 10006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182045077837212e-06,
+      "loss": 0.9501,
+      "step": 10007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180144392747198e-06,
+      "loss": 0.9912,
+      "step": 10008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1782439344434295e-06,
+      "loss": 1.021,
+      "step": 10009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176343703015328e-06,
+      "loss": 1.0347,
+      "step": 10010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174443698552303e-06,
+      "loss": 0.9525,
+      "step": 10011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.17254392114375e-06,
+      "loss": 0.8723,
+      "step": 10012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170644370879068e-06,
+      "loss": 1.0095,
+      "step": 10013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16874504784763e-06,
+      "loss": 0.9652,
+      "step": 10014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.166845952138805e-06,
+      "loss": 0.8878,
+      "step": 10015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.164947083841949e-06,
+      "loss": 0.9252,
+      "step": 10016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.163048443046406e-06,
+      "loss": 0.8847,
+      "step": 10017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161150029841524e-06,
+      "loss": 0.9391,
+      "step": 10018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159251844316615e-06,
+      "loss": 0.9504,
+      "step": 10019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157353886560991e-06,
+      "loss": 0.8886,
+      "step": 10020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1554561566639694e-06,
+      "loss": 0.9809,
+      "step": 10021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.153558654714835e-06,
+      "loss": 0.964,
+      "step": 10022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151661380802871e-06,
+      "loss": 0.8857,
+      "step": 10023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149764335017349e-06,
+      "loss": 0.946,
+      "step": 10024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147867517447525e-06,
+      "loss": 0.924,
+      "step": 10025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145970928182661e-06,
+      "loss": 0.9482,
+      "step": 10026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.144074567311982e-06,
+      "loss": 1.0026,
+      "step": 10027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142178434924727e-06,
+      "loss": 0.9046,
+      "step": 10028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14028253111011e-06,
+      "loss": 0.9417,
+      "step": 10029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.138386855957338e-06,
+      "loss": 0.9572,
+      "step": 10030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1364914095556065e-06,
+      "loss": 0.9315,
+      "step": 10031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134596191994099e-06,
+      "loss": 1.0196,
+      "step": 10032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132701203362e-06,
+      "loss": 0.8064,
+      "step": 10033
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1308064437484636e-06,
+      "loss": 0.9397,
+      "step": 10034
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.128911913242642e-06,
+      "loss": 0.8971,
+      "step": 10035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127017611933687e-06,
+      "loss": 0.9549,
+      "step": 10036
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1251235399107216e-06,
+      "loss": 0.951,
+      "step": 10037
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1232296972628716e-06,
+      "loss": 0.9253,
+      "step": 10038
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121336084079245e-06,
+      "loss": 0.8682,
+      "step": 10039
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.119442700448936e-06,
+      "loss": 0.8603,
+      "step": 10040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117549546461047e-06,
+      "loss": 0.937,
+      "step": 10041
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1156566222046436e-06,
+      "loss": 0.9158,
+      "step": 10042
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.113763927768791e-06,
+      "loss": 1.0216,
+      "step": 10043
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1118714632425535e-06,
+      "loss": 0.9573,
+      "step": 10044
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109979228714974e-06,
+      "loss": 0.9646,
+      "step": 10045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108087224275085e-06,
+      "loss": 0.9282,
+      "step": 10046
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.106195450011911e-06,
+      "loss": 0.9579,
+      "step": 10047
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1043039060144605e-06,
+      "loss": 0.9125,
+      "step": 10048
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102412592371746e-06,
+      "loss": 0.8694,
+      "step": 10049
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100521509172745e-06,
+      "loss": 0.9698,
+      "step": 10050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0986306565064495e-06,
+      "loss": 1.0124,
+      "step": 10051
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096740034461823e-06,
+      "loss": 0.9154,
+      "step": 10052
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094849643127824e-06,
+      "loss": 0.9132,
+      "step": 10053
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092959482593401e-06,
+      "loss": 0.8776,
+      "step": 10054
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091069552947488e-06,
+      "loss": 1.0071,
+      "step": 10055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089179854279017e-06,
+      "loss": 0.9143,
+      "step": 10056
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087290386676904e-06,
+      "loss": 0.9872,
+      "step": 10057
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.08540115023004e-06,
+      "loss": 0.9399,
+      "step": 10058
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083512145027332e-06,
+      "loss": 0.9836,
+      "step": 10059
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081623371157658e-06,
+      "loss": 0.9753,
+      "step": 10060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079734828709888e-06,
+      "loss": 1.0141,
+      "step": 10061
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0778465177728845e-06,
+      "loss": 0.978,
+      "step": 10062
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075958438435492e-06,
+      "loss": 1.0056,
+      "step": 10063
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074070590786559e-06,
+      "loss": 0.9131,
+      "step": 10064
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.072182974914906e-06,
+      "loss": 0.9632,
+      "step": 10065
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.070295590909354e-06,
+      "loss": 0.955,
+      "step": 10066
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0684084388587075e-06,
+      "loss": 0.9488,
+      "step": 10067
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.066521518851761e-06,
+      "loss": 0.9036,
+      "step": 10068
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0646348309773e-06,
+      "loss": 0.9112,
+      "step": 10069
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062748375324094e-06,
+      "loss": 0.9572,
+      "step": 10070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060862151980912e-06,
+      "loss": 0.9373,
+      "step": 10071
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058976161036506e-06,
+      "loss": 0.8925,
+      "step": 10072
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057090402579605e-06,
+      "loss": 0.9325,
+      "step": 10073
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055204876698951e-06,
+      "loss": 0.9527,
+      "step": 10074
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0533195834832585e-06,
+      "loss": 0.9731,
+      "step": 10075
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051434523021235e-06,
+      "loss": 0.948,
+      "step": 10076
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049549695401577e-06,
+      "loss": 0.9583,
+      "step": 10077
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.047665100712966e-06,
+      "loss": 0.8768,
+      "step": 10078
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045780739044086e-06,
+      "loss": 1.038,
+      "step": 10079
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.043896610483601e-06,
+      "loss": 0.9628,
+      "step": 10080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0420127151201514e-06,
+      "loss": 0.9684,
+      "step": 10081
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040129053042391e-06,
+      "loss": 0.9536,
+      "step": 10082
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.038245624338943e-06,
+      "loss": 0.8902,
+      "step": 10083
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036362429098439e-06,
+      "loss": 0.9115,
+      "step": 10084
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034479467409477e-06,
+      "loss": 0.9433,
+      "step": 10085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032596739360655e-06,
+      "loss": 0.8487,
+      "step": 10086
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030714245040567e-06,
+      "loss": 0.9611,
+      "step": 10087
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0288319845377855e-06,
+      "loss": 0.9085,
+      "step": 10088
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0269499579408765e-06,
+      "loss": 1.0116,
+      "step": 10089
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025068165338391e-06,
+      "loss": 0.993,
+      "step": 10090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023186606818877e-06,
+      "loss": 0.9867,
+      "step": 10091
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.021305282470863e-06,
+      "loss": 0.9611,
+      "step": 10092
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019424192382867e-06,
+      "loss": 0.8717,
+      "step": 10093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017543336643406e-06,
+      "loss": 0.9946,
+      "step": 10094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015662715340976e-06,
+      "loss": 0.9636,
+      "step": 10095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.013782328564065e-06,
+      "loss": 0.9324,
+      "step": 10096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0119021764011485e-06,
+      "loss": 0.9158,
+      "step": 10097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010022258940692e-06,
+      "loss": 0.8757,
+      "step": 10098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008142576271158e-06,
+      "loss": 0.9407,
+      "step": 10099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006263128480981e-06,
+      "loss": 0.933,
+      "step": 10100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.004383915658592e-06,
+      "loss": 0.9093,
+      "step": 10101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002504937892421e-06,
+      "loss": 0.9415,
+      "step": 10102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0006261952708755e-06,
+      "loss": 1.0235,
+      "step": 10103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998747687882355e-06,
+      "loss": 0.9154,
+      "step": 10104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996869415815247e-06,
+      "loss": 0.9693,
+      "step": 10105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9949913791579265e-06,
+      "loss": 0.8766,
+      "step": 10106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99311357799877e-06,
+      "loss": 0.9206,
+      "step": 10107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991236012426116e-06,
+      "loss": 0.9961,
+      "step": 10108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989358682528324e-06,
+      "loss": 0.9572,
+      "step": 10109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98748158839372e-06,
+      "loss": 1.0005,
+      "step": 10110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.985604730110628e-06,
+      "loss": 0.9186,
+      "step": 10111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9837281077673575e-06,
+      "loss": 0.9739,
+      "step": 10112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.981851721452209e-06,
+      "loss": 0.8989,
+      "step": 10113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979975571253467e-06,
+      "loss": 0.9658,
+      "step": 10114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9780996572594205e-06,
+      "loss": 0.8806,
+      "step": 10115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9762239795583215e-06,
+      "loss": 0.8592,
+      "step": 10116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974348538238435e-06,
+      "loss": 0.948,
+      "step": 10117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972473333388003e-06,
+      "loss": 0.9851,
+      "step": 10118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9705983650952575e-06,
+      "loss": 0.9374,
+      "step": 10119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968723633448421e-06,
+      "loss": 0.9432,
+      "step": 10120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966849138535701e-06,
+      "loss": 0.8995,
+      "step": 10121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964974880445305e-06,
+      "loss": 0.945,
+      "step": 10122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963100859265419e-06,
+      "loss": 0.9827,
+      "step": 10123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961227075084212e-06,
+      "loss": 0.9069,
+      "step": 10124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95935352798986e-06,
+      "loss": 0.9849,
+      "step": 10125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957480218070515e-06,
+      "loss": 0.9587,
+      "step": 10126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95560714541432e-06,
+      "loss": 0.8921,
+      "step": 10127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9537343101094085e-06,
+      "loss": 0.9366,
+      "step": 10128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951861712243897e-06,
+      "loss": 0.938,
+      "step": 10129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949989351905909e-06,
+      "loss": 0.9758,
+      "step": 10130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948117229183529e-06,
+      "loss": 0.977,
+      "step": 10131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.946245344164855e-06,
+      "loss": 0.9831,
+      "step": 10132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.944373696937959e-06,
+      "loss": 0.9572,
+      "step": 10133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94250228759091e-06,
+      "loss": 0.9606,
+      "step": 10134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9406311162117595e-06,
+      "loss": 0.9826,
+      "step": 10135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938760182888549e-06,
+      "loss": 0.9057,
+      "step": 10136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936889487709316e-06,
+      "loss": 0.9456,
+      "step": 10137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935019030762082e-06,
+      "loss": 0.9248,
+      "step": 10138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933148812134848e-06,
+      "loss": 0.9631,
+      "step": 10139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931278831915621e-06,
+      "loss": 0.8385,
+      "step": 10140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929409090192381e-06,
+      "loss": 0.9589,
+      "step": 10141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927539587053116e-06,
+      "loss": 0.874,
+      "step": 10142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.925670322585779e-06,
+      "loss": 0.9482,
+      "step": 10143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.923801296878325e-06,
+      "loss": 0.975,
+      "step": 10144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.921932510018701e-06,
+      "loss": 0.9538,
+      "step": 10145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920063962094837e-06,
+      "loss": 0.9134,
+      "step": 10146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9181956531946515e-06,
+      "loss": 0.9451,
+      "step": 10147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9163275834060545e-06,
+      "loss": 0.9608,
+      "step": 10148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91445975281694e-06,
+      "loss": 0.9937,
+      "step": 10149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9125921615151985e-06,
+      "loss": 0.999,
+      "step": 10150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910724809588701e-06,
+      "loss": 0.9899,
+      "step": 10151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908857697125311e-06,
+      "loss": 1.0041,
+      "step": 10152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906990824212885e-06,
+      "loss": 0.9754,
+      "step": 10153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905124190939262e-06,
+      "loss": 0.9525,
+      "step": 10154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903257797392271e-06,
+      "loss": 0.9539,
+      "step": 10155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901391643659732e-06,
+      "loss": 1.0072,
+      "step": 10156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899525729829449e-06,
+      "loss": 0.9097,
+      "step": 10157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897660055989222e-06,
+      "loss": 0.9067,
+      "step": 10158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8957946222268295e-06,
+      "loss": 0.9387,
+      "step": 10159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.893929428630052e-06,
+      "loss": 0.9711,
+      "step": 10160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892064475286649e-06,
+      "loss": 0.9612,
+      "step": 10161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89019976228437e-06,
+      "loss": 0.9186,
+      "step": 10162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888335289710954e-06,
+      "loss": 0.952,
+      "step": 10163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886471057654129e-06,
+      "loss": 0.959,
+      "step": 10164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884607066201619e-06,
+      "loss": 1.0344,
+      "step": 10165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.88274331544112e-06,
+      "loss": 0.9564,
+      "step": 10166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.880879805460325e-06,
+      "loss": 0.8877,
+      "step": 10167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879016536346925e-06,
+      "loss": 0.9596,
+      "step": 10168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877153508188588e-06,
+      "loss": 0.949,
+      "step": 10169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.875290721072975e-06,
+      "loss": 0.9729,
+      "step": 10170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873428175087732e-06,
+      "loss": 0.8992,
+      "step": 10171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8715658703204945e-06,
+      "loss": 0.9507,
+      "step": 10172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8697038068589005e-06,
+      "loss": 0.9586,
+      "step": 10173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.867841984790549e-06,
+      "loss": 0.9674,
+      "step": 10174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865980404203054e-06,
+      "loss": 0.8956,
+      "step": 10175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864119065184005e-06,
+      "loss": 0.9483,
+      "step": 10176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.862257967820981e-06,
+      "loss": 0.9492,
+      "step": 10177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.860397112201554e-06,
+      "loss": 1.0223,
+      "step": 10178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858536498413274e-06,
+      "loss": 1.0003,
+      "step": 10179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856676126543699e-06,
+      "loss": 0.9658,
+      "step": 10180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854815996680362e-06,
+      "loss": 0.9628,
+      "step": 10181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852956108910777e-06,
+      "loss": 1.0105,
+      "step": 10182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851096463322466e-06,
+      "loss": 0.9427,
+      "step": 10183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849237060002927e-06,
+      "loss": 0.9493,
+      "step": 10184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8473778990396495e-06,
+      "loss": 0.9647,
+      "step": 10185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845518980520111e-06,
+      "loss": 0.9953,
+      "step": 10186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843660304531776e-06,
+      "loss": 0.9197,
+      "step": 10187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84180187116211e-06,
+      "loss": 0.9602,
+      "step": 10188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839943680498546e-06,
+      "loss": 0.8872,
+      "step": 10189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838085732628516e-06,
+      "loss": 0.9091,
+      "step": 10190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836228027639449e-06,
+      "loss": 0.9626,
+      "step": 10191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834370565618752e-06,
+      "loss": 0.9511,
+      "step": 10192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.832513346653822e-06,
+      "loss": 0.9374,
+      "step": 10193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.830656370832046e-06,
+      "loss": 1.0041,
+      "step": 10194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8287996382407945e-06,
+      "loss": 0.9706,
+      "step": 10195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826943148967444e-06,
+      "loss": 0.9981,
+      "step": 10196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.825086903099332e-06,
+      "loss": 0.9765,
+      "step": 10197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823230900723811e-06,
+      "loss": 0.8927,
+      "step": 10198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8213751419282054e-06,
+      "loss": 0.9669,
+      "step": 10199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819519626799835e-06,
+      "loss": 0.8996,
+      "step": 10200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817664355426005e-06,
+      "loss": 1.0053,
+      "step": 10201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815809327894006e-06,
+      "loss": 0.9968,
+      "step": 10202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.813954544291133e-06,
+      "loss": 0.9404,
+      "step": 10203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812100004704654e-06,
+      "loss": 1.0017,
+      "step": 10204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810245709221822e-06,
+      "loss": 0.9092,
+      "step": 10205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808391657929894e-06,
+      "loss": 0.9446,
+      "step": 10206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806537850916106e-06,
+      "loss": 0.9651,
+      "step": 10207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804684288267683e-06,
+      "loss": 0.8906,
+      "step": 10208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.802830970071842e-06,
+      "loss": 0.894,
+      "step": 10209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800977896415782e-06,
+      "loss": 0.8774,
+      "step": 10210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799125067386701e-06,
+      "loss": 0.9486,
+      "step": 10211
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797272483071774e-06,
+      "loss": 0.8978,
+      "step": 10212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795420143558174e-06,
+      "loss": 0.9433,
+      "step": 10213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.793568048933054e-06,
+      "loss": 0.9261,
+      "step": 10214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.791716199283562e-06,
+      "loss": 0.9783,
+      "step": 10215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789864594696833e-06,
+      "loss": 0.9827,
+      "step": 10216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788013235259983e-06,
+      "loss": 0.9829,
+      "step": 10217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786162121060133e-06,
+      "loss": 0.9584,
+      "step": 10218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.784311252184377e-06,
+      "loss": 0.9114,
+      "step": 10219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782460628719804e-06,
+      "loss": 0.9409,
+      "step": 10220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780610250753491e-06,
+      "loss": 0.9202,
+      "step": 10221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778760118372498e-06,
+      "loss": 0.914,
+      "step": 10222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.77691023166389e-06,
+      "loss": 0.914,
+      "step": 10223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775060590714698e-06,
+      "loss": 0.9784,
+      "step": 10224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773211195611953e-06,
+      "loss": 0.968,
+      "step": 10225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771362046442679e-06,
+      "loss": 0.9415,
+      "step": 10226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769513143293884e-06,
+      "loss": 1.0122,
+      "step": 10227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7676644862525524e-06,
+      "loss": 0.9323,
+      "step": 10228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.765816075405679e-06,
+      "loss": 0.9642,
+      "step": 10229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.763967910840231e-06,
+      "loss": 0.9484,
+      "step": 10230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762119992643176e-06,
+      "loss": 0.9559,
+      "step": 10231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760272320901456e-06,
+      "loss": 0.8803,
+      "step": 10232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758424895702005e-06,
+      "loss": 0.8908,
+      "step": 10233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756577717131761e-06,
+      "loss": 0.9451,
+      "step": 10234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75473078527763e-06,
+      "loss": 0.9836,
+      "step": 10235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752884100226517e-06,
+      "loss": 0.9536,
+      "step": 10236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751037662065314e-06,
+      "loss": 0.9502,
+      "step": 10237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749191470880898e-06,
+      "loss": 0.9787,
+      "step": 10238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.74734552676014e-06,
+      "loss": 0.9182,
+      "step": 10239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745499829789889e-06,
+      "loss": 1.0315,
+      "step": 10240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743654380057001e-06,
+      "loss": 0.9606,
+      "step": 10241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741809177648302e-06,
+      "loss": 0.9069,
+      "step": 10242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739964222650615e-06,
+      "loss": 0.9328,
+      "step": 10243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.73811951515075e-06,
+      "loss": 0.9898,
+      "step": 10244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736275055235499e-06,
+      "loss": 0.8524,
+      "step": 10245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734430842991663e-06,
+      "loss": 1.0261,
+      "step": 10246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.732586878506003e-06,
+      "loss": 0.9015,
+      "step": 10247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7307431618652844e-06,
+      "loss": 0.9911,
+      "step": 10248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728899693156265e-06,
+      "loss": 0.9781,
+      "step": 10249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72705647246568e-06,
+      "loss": 0.8888,
+      "step": 10250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.725213499880258e-06,
+      "loss": 0.9729,
+      "step": 10251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.723370775486716e-06,
+      "loss": 0.8876,
+      "step": 10252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721528299371755e-06,
+      "loss": 0.9668,
+      "step": 10253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7196860716220795e-06,
+      "loss": 1.0044,
+      "step": 10254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717844092324355e-06,
+      "loss": 1.0216,
+      "step": 10255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716002361565263e-06,
+      "loss": 0.839,
+      "step": 10256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714160879431459e-06,
+      "loss": 0.9445,
+      "step": 10257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.712319646009587e-06,
+      "loss": 0.9697,
+      "step": 10258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710478661386283e-06,
+      "loss": 0.9059,
+      "step": 10259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7086379256481695e-06,
+      "loss": 0.982,
+      "step": 10260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706797438881853e-06,
+      "loss": 0.9437,
+      "step": 10261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704957201173948e-06,
+      "loss": 0.9616,
+      "step": 10262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703117212611023e-06,
+      "loss": 0.958,
+      "step": 10263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701277473279668e-06,
+      "loss": 0.9206,
+      "step": 10264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.699437983266442e-06,
+      "loss": 0.8337,
+      "step": 10265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697598742657899e-06,
+      "loss": 0.9407,
+      "step": 10266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.695759751540578e-06,
+      "loss": 0.946,
+      "step": 10267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693921010001007e-06,
+      "loss": 0.9348,
+      "step": 10268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692082518125708e-06,
+      "loss": 0.9402,
+      "step": 10269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690244276001188e-06,
+      "loss": 0.9192,
+      "step": 10270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.68840628371393e-06,
+      "loss": 0.9617,
+      "step": 10271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686568541350427e-06,
+      "loss": 0.9442,
+      "step": 10272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.684731048997145e-06,
+      "loss": 0.9761,
+      "step": 10273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682893806740544e-06,
+      "loss": 0.9987,
+      "step": 10274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.68105681466707e-06,
+      "loss": 0.9269,
+      "step": 10275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679220072863153e-06,
+      "loss": 0.9308,
+      "step": 10276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6773835814152265e-06,
+      "loss": 0.9571,
+      "step": 10277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675547340409696e-06,
+      "loss": 0.9548,
+      "step": 10278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.673711349932963e-06,
+      "loss": 0.985,
+      "step": 10279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6718756100714146e-06,
+      "loss": 0.9997,
+      "step": 10280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670040120911427e-06,
+      "loss": 0.9531,
+      "step": 10281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668204882539364e-06,
+      "loss": 0.8885,
+      "step": 10282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6663698950415746e-06,
+      "loss": 1.0086,
+      "step": 10283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664535158504407e-06,
+      "loss": 0.8946,
+      "step": 10284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662700673014191e-06,
+      "loss": 0.9929,
+      "step": 10285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6608664386572325e-06,
+      "loss": 0.907,
+      "step": 10286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6590324555198464e-06,
+      "loss": 0.9873,
+      "step": 10287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65719872368832e-06,
+      "loss": 0.9033,
+      "step": 10288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.655365243248946e-06,
+      "loss": 0.8901,
+      "step": 10289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6535320142879835e-06,
+      "loss": 0.9106,
+      "step": 10290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6516990368916896e-06,
+      "loss": 0.9687,
+      "step": 10291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.649866311146319e-06,
+      "loss": 0.8717,
+      "step": 10292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6480338371381e-06,
+      "loss": 0.9597,
+      "step": 10293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646201614953258e-06,
+      "loss": 0.97,
+      "step": 10294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.644369644678002e-06,
+      "loss": 1.0018,
+      "step": 10295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6425379263985304e-06,
+      "loss": 0.9227,
+      "step": 10296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640706460201032e-06,
+      "loss": 0.9304,
+      "step": 10297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63887524617168e-06,
+      "loss": 0.9859,
+      "step": 10298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637044284396635e-06,
+      "loss": 0.969,
+      "step": 10299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6352135749620545e-06,
+      "loss": 0.9136,
+      "step": 10300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633383117954076e-06,
+      "loss": 0.9218,
+      "step": 10301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631552913458825e-06,
+      "loss": 0.932,
+      "step": 10302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6297229615624185e-06,
+      "loss": 1.0195,
+      "step": 10303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627893262350961e-06,
+      "loss": 0.9331,
+      "step": 10304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626063815910543e-06,
+      "loss": 1.0007,
+      "step": 10305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62423462232724e-06,
+      "loss": 0.9746,
+      "step": 10306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.622405681687131e-06,
+      "loss": 0.9232,
+      "step": 10307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620576994076266e-06,
+      "loss": 0.9771,
+      "step": 10308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618748559580689e-06,
+      "loss": 0.817,
+      "step": 10309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616920378286433e-06,
+      "loss": 0.9799,
+      "step": 10310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615092450279516e-06,
+      "loss": 0.912,
+      "step": 10311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613264775645956e-06,
+      "loss": 1.0126,
+      "step": 10312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.611437354471738e-06,
+      "loss": 0.925,
+      "step": 10313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60961018684285e-06,
+      "loss": 0.9061,
+      "step": 10314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607783272845269e-06,
+      "loss": 0.9368,
+      "step": 10315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605956612564953e-06,
+      "loss": 1.0109,
+      "step": 10316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.604130206087851e-06,
+      "loss": 0.9211,
+      "step": 10317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602304053499902e-06,
+      "loss": 0.9575,
+      "step": 10318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6004781548870235e-06,
+      "loss": 0.9027,
+      "step": 10319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598652510335143e-06,
+      "loss": 0.9512,
+      "step": 10320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596827119930145e-06,
+      "loss": 0.9518,
+      "step": 10321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5950019837579295e-06,
+      "loss": 0.932,
+      "step": 10322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593177101904372e-06,
+      "loss": 1.0122,
+      "step": 10323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.591352474455336e-06,
+      "loss": 0.9618,
+      "step": 10324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5895281014966764e-06,
+      "loss": 0.9267,
+      "step": 10325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587703983114229e-06,
+      "loss": 0.9828,
+      "step": 10326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5858801193938305e-06,
+      "loss": 0.9206,
+      "step": 10327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5840565104213e-06,
+      "loss": 0.9537,
+      "step": 10328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582233156282431e-06,
+      "loss": 0.9961,
+      "step": 10329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5804100570630275e-06,
+      "loss": 0.857,
+      "step": 10330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578587212848866e-06,
+      "loss": 0.9738,
+      "step": 10331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576764623725719e-06,
+      "loss": 0.9264,
+      "step": 10332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.574942289779342e-06,
+      "loss": 0.9266,
+      "step": 10333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573120211095477e-06,
+      "loss": 0.9511,
+      "step": 10334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571298387759867e-06,
+      "loss": 0.9425,
+      "step": 10335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569476819858225e-06,
+      "loss": 0.8881,
+      "step": 10336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567655507476259e-06,
+      "loss": 0.9283,
+      "step": 10337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565834450699672e-06,
+      "loss": 0.9519,
+      "step": 10338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.564013649614148e-06,
+      "loss": 0.9367,
+      "step": 10339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562193104305359e-06,
+      "loss": 0.8602,
+      "step": 10340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560372814858966e-06,
+      "loss": 0.9465,
+      "step": 10341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558552781360617e-06,
+      "loss": 0.8559,
+      "step": 10342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556733003895957e-06,
+      "loss": 0.9357,
+      "step": 10343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554913482550598e-06,
+      "loss": 0.8693,
+      "step": 10344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553094217410163e-06,
+      "loss": 0.8813,
+      "step": 10345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551275208560251e-06,
+      "loss": 0.996,
+      "step": 10346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549456456086449e-06,
+      "loss": 0.9623,
+      "step": 10347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.547637960074334e-06,
+      "loss": 0.9961,
+      "step": 10348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5458197206094674e-06,
+      "loss": 0.9492,
+      "step": 10349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544001737777412e-06,
+      "loss": 0.8607,
+      "step": 10350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542184011663705e-06,
+      "loss": 0.9912,
+      "step": 10351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540366542353865e-06,
+      "loss": 0.9105,
+      "step": 10352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538549329933418e-06,
+      "loss": 0.9691,
+      "step": 10353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536732374487868e-06,
+      "loss": 0.9441,
+      "step": 10354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5349156761027045e-06,
+      "loss": 0.9436,
+      "step": 10355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.53309923486341e-06,
+      "loss": 0.9208,
+      "step": 10356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.531283050855446e-06,
+      "loss": 0.8256,
+      "step": 10357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529467124164279e-06,
+      "loss": 0.9335,
+      "step": 10358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527651454875347e-06,
+      "loss": 0.9073,
+      "step": 10359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525836043074082e-06,
+      "loss": 0.9712,
+      "step": 10360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5240208888459045e-06,
+      "loss": 0.9389,
+      "step": 10361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522205992276222e-06,
+      "loss": 1.0123,
+      "step": 10362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520391353450429e-06,
+      "loss": 0.8829,
+      "step": 10363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.518576972453906e-06,
+      "loss": 0.973,
+      "step": 10364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51676284937203e-06,
+      "loss": 0.9843,
+      "step": 10365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514948984290158e-06,
+      "loss": 0.9253,
+      "step": 10366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.513135377293638e-06,
+      "loss": 0.974,
+      "step": 10367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511322028467801e-06,
+      "loss": 0.9819,
+      "step": 10368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50950893789797e-06,
+      "loss": 0.8621,
+      "step": 10369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507696105669463e-06,
+      "loss": 0.9581,
+      "step": 10370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50588353186757e-06,
+      "loss": 0.8741,
+      "step": 10371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.504071216577576e-06,
+      "loss": 0.9077,
+      "step": 10372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502259159884761e-06,
+      "loss": 0.9412,
+      "step": 10373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500447361874385e-06,
+      "loss": 0.8999,
+      "step": 10374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498635822631696e-06,
+      "loss": 0.8731,
+      "step": 10375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496824542241934e-06,
+      "loss": 1.0344,
+      "step": 10376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495013520790318e-06,
+      "loss": 0.965,
+      "step": 10377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493202758362074e-06,
+      "loss": 0.9297,
+      "step": 10378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491392255042391e-06,
+      "loss": 0.8748,
+      "step": 10379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4895820109164565e-06,
+      "loss": 0.9029,
+      "step": 10380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4877720260694554e-06,
+      "loss": 0.9755,
+      "step": 10381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485962300586549e-06,
+      "loss": 1.0123,
+      "step": 10382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48415283455289e-06,
+      "loss": 0.8777,
+      "step": 10383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482343628053616e-06,
+      "loss": 0.908,
+      "step": 10384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480534681173852e-06,
+      "loss": 0.9143,
+      "step": 10385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.478725993998726e-06,
+      "loss": 0.9174,
+      "step": 10386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476917566613324e-06,
+      "loss": 0.9374,
+      "step": 10387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47510939910275e-06,
+      "loss": 0.9361,
+      "step": 10388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473301491552078e-06,
+      "loss": 0.8403,
+      "step": 10389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4714938440463755e-06,
+      "loss": 0.9331,
+      "step": 10390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469686456670696e-06,
+      "loss": 0.9529,
+      "step": 10391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467879329510076e-06,
+      "loss": 0.9773,
+      "step": 10392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.466072462649561e-06,
+      "loss": 0.906,
+      "step": 10393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.464265856174155e-06,
+      "loss": 0.9355,
+      "step": 10394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4624595101688636e-06,
+      "loss": 1.0429,
+      "step": 10395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460653424718687e-06,
+      "loss": 0.9957,
+      "step": 10396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458847599908602e-06,
+      "loss": 0.9695,
+      "step": 10397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.457042035823578e-06,
+      "loss": 0.9168,
+      "step": 10398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4552367325485715e-06,
+      "loss": 0.9113,
+      "step": 10399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453431690168521e-06,
+      "loss": 0.8675,
+      "step": 10400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451626908768372e-06,
+      "loss": 0.9807,
+      "step": 10401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4498223884330274e-06,
+      "loss": 0.9535,
+      "step": 10402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.448018129247405e-06,
+      "loss": 0.8936,
+      "step": 10403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446214131296398e-06,
+      "loss": 0.9423,
+      "step": 10404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444410394664887e-06,
+      "loss": 0.9212,
+      "step": 10405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442606919437744e-06,
+      "loss": 0.8572,
+      "step": 10406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440803705699822e-06,
+      "loss": 0.9682,
+      "step": 10407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439000753535976e-06,
+      "loss": 0.8854,
+      "step": 10408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437198063031036e-06,
+      "loss": 0.9788,
+      "step": 10409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4353956342698155e-06,
+      "loss": 0.9636,
+      "step": 10410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433593467337133e-06,
+      "loss": 0.9282,
+      "step": 10411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431791562317781e-06,
+      "loss": 0.9509,
+      "step": 10412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429989919296543e-06,
+      "loss": 0.9195,
+      "step": 10413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.428188538358192e-06,
+      "loss": 0.9239,
+      "step": 10414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426387419587483e-06,
+      "loss": 0.9079,
+      "step": 10415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424586563069172e-06,
+      "loss": 0.9589,
+      "step": 10416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422785968887992e-06,
+      "loss": 0.8644,
+      "step": 10417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420985637128657e-06,
+      "loss": 0.8631,
+      "step": 10418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4191855678758836e-06,
+      "loss": 0.9059,
+      "step": 10419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41738576121437e-06,
+      "loss": 0.8915,
+      "step": 10420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415586217228799e-06,
+      "loss": 0.8894,
+      "step": 10421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4137869360038456e-06,
+      "loss": 0.9761,
+      "step": 10422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411987917624166e-06,
+      "loss": 0.9194,
+      "step": 10423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.410189162174415e-06,
+      "loss": 0.9588,
+      "step": 10424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.408390669739228e-06,
+      "loss": 0.9775,
+      "step": 10425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.406592440403223e-06,
+      "loss": 0.9194,
+      "step": 10426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404794474251017e-06,
+      "loss": 0.9302,
+      "step": 10427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402996771367205e-06,
+      "loss": 0.8648,
+      "step": 10428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401199331836375e-06,
+      "loss": 0.9171,
+      "step": 10429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.399402155743098e-06,
+      "loss": 0.9275,
+      "step": 10430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397605243171941e-06,
+      "loss": 0.959,
+      "step": 10431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3958085942074536e-06,
+      "loss": 0.9737,
+      "step": 10432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.394012208934163e-06,
+      "loss": 0.9112,
+      "step": 10433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.392216087436603e-06,
+      "loss": 0.9694,
+      "step": 10434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3904202297992806e-06,
+      "loss": 0.8821,
+      "step": 10435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.388624636106703e-06,
+      "loss": 0.9386,
+      "step": 10436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38682930644335e-06,
+      "loss": 0.9728,
+      "step": 10437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385034240893693e-06,
+      "loss": 1.0592,
+      "step": 10438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.383239439542204e-06,
+      "loss": 0.9063,
+      "step": 10439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3814449024733285e-06,
+      "loss": 0.8916,
+      "step": 10440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379650629771505e-06,
+      "loss": 0.9338,
+      "step": 10441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.377856621521156e-06,
+      "loss": 0.9376,
+      "step": 10442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3760628778066916e-06,
+      "loss": 0.9701,
+      "step": 10443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374269398712524e-06,
+      "loss": 0.9432,
+      "step": 10444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3724761843230246e-06,
+      "loss": 0.936,
+      "step": 10445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.370683234722581e-06,
+      "loss": 0.9857,
+      "step": 10446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.368890549995551e-06,
+      "loss": 0.9209,
+      "step": 10447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367098130226287e-06,
+      "loss": 0.887,
+      "step": 10448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.365305975499125e-06,
+      "loss": 0.9234,
+      "step": 10449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.363514085898389e-06,
+      "loss": 0.9016,
+      "step": 10450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.361722461508395e-06,
+      "loss": 0.9255,
+      "step": 10451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.359931102413443e-06,
+      "loss": 0.8836,
+      "step": 10452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358140008697815e-06,
+      "loss": 0.9397,
+      "step": 10453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3563491804457945e-06,
+      "loss": 1.0308,
+      "step": 10454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3545586177416425e-06,
+      "loss": 0.9555,
+      "step": 10455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352768320669607e-06,
+      "loss": 0.9504,
+      "step": 10456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.350978289313929e-06,
+      "loss": 0.9215,
+      "step": 10457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349188523758827e-06,
+      "loss": 0.9594,
+      "step": 10458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347399024088527e-06,
+      "loss": 0.9905,
+      "step": 10459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.345609790387217e-06,
+      "loss": 0.9594,
+      "step": 10460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343820822739086e-06,
+      "loss": 1.0093,
+      "step": 10461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.342032121228318e-06,
+      "loss": 0.9124,
+      "step": 10462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.340243685939069e-06,
+      "loss": 0.9548,
+      "step": 10463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338455516955492e-06,
+      "loss": 0.9554,
+      "step": 10464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3366676143617245e-06,
+      "loss": 0.9638,
+      "step": 10465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334879978241886e-06,
+      "loss": 0.9097,
+      "step": 10466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333092608680103e-06,
+      "loss": 0.9476,
+      "step": 10467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331305505760459e-06,
+      "loss": 0.9503,
+      "step": 10468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3295186695670535e-06,
+      "loss": 1.0005,
+      "step": 10469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327732100183958e-06,
+      "loss": 0.9348,
+      "step": 10470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325945797695236e-06,
+      "loss": 0.8969,
+      "step": 10471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324159762184934e-06,
+      "loss": 1.0084,
+      "step": 10472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.32237399373709e-06,
+      "loss": 0.954,
+      "step": 10473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320588492435734e-06,
+      "loss": 0.9328,
+      "step": 10474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.318803258364879e-06,
+      "loss": 0.8861,
+      "step": 10475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.317018291608513e-06,
+      "loss": 0.9343,
+      "step": 10476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.315233592250634e-06,
+      "loss": 0.8816,
+      "step": 10477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313449160375214e-06,
+      "loss": 1.0209,
+      "step": 10478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3116649960662145e-06,
+      "loss": 0.8662,
+      "step": 10479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.309881099407585e-06,
+      "loss": 0.8956,
+      "step": 10480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3080974704832575e-06,
+      "loss": 0.9633,
+      "step": 10481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306314109377165e-06,
+      "loss": 0.962,
+      "step": 10482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.304531016173216e-06,
+      "loss": 0.857,
+      "step": 10483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.302748190955307e-06,
+      "loss": 0.8986,
+      "step": 10484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300965633807326e-06,
+      "loss": 0.9631,
+      "step": 10485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299183344813146e-06,
+      "loss": 0.8644,
+      "step": 10486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297401324056629e-06,
+      "loss": 0.9566,
+      "step": 10487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.295619571621624e-06,
+      "loss": 0.9371,
+      "step": 10488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293838087591962e-06,
+      "loss": 0.905,
+      "step": 10489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.292056872051478e-06,
+      "loss": 0.8991,
+      "step": 10490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2902759250839664e-06,
+      "loss": 1.0164,
+      "step": 10491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288495246773239e-06,
+      "loss": 0.9993,
+      "step": 10492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.286714837203076e-06,
+      "loss": 0.9122,
+      "step": 10493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2849346964572495e-06,
+      "loss": 0.9166,
+      "step": 10494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.283154824619519e-06,
+      "loss": 1.0015,
+      "step": 10495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281375221773632e-06,
+      "loss": 0.9249,
+      "step": 10496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2795958880033254e-06,
+      "loss": 0.9259,
+      "step": 10497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277816823392326e-06,
+      "loss": 0.8667,
+      "step": 10498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276038028024328e-06,
+      "loss": 1.0548,
+      "step": 10499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274259501983042e-06,
+      "loss": 0.9236,
+      "step": 10500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272481245352146e-06,
+      "loss": 0.8667,
+      "step": 10501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2707032582153185e-06,
+      "loss": 0.9135,
+      "step": 10502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268925540656209e-06,
+      "loss": 1.0373,
+      "step": 10503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267148092758465e-06,
+      "loss": 1.0197,
+      "step": 10504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.265370914605725e-06,
+      "loss": 0.9172,
+      "step": 10505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.263594006281609e-06,
+      "loss": 0.9423,
+      "step": 10506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26181736786972e-06,
+      "loss": 0.8812,
+      "step": 10507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260040999453657e-06,
+      "loss": 0.9145,
+      "step": 10508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258264901117003e-06,
+      "loss": 0.8938,
+      "step": 10509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256489072943326e-06,
+      "loss": 0.8799,
+      "step": 10510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25471351501618e-06,
+      "loss": 0.9907,
+      "step": 10511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.252938227419118e-06,
+      "loss": 0.9355,
+      "step": 10512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251163210235665e-06,
+      "loss": 0.8856,
+      "step": 10513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249388463549343e-06,
+      "loss": 0.9955,
+      "step": 10514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247613987443657e-06,
+      "loss": 0.9857,
+      "step": 10515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.245839782002098e-06,
+      "loss": 0.8491,
+      "step": 10516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244065847308156e-06,
+      "loss": 0.9745,
+      "step": 10517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.242292183445289e-06,
+      "loss": 0.9313,
+      "step": 10518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240518790496953e-06,
+      "loss": 0.9821,
+      "step": 10519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238745668546595e-06,
+      "loss": 0.9662,
+      "step": 10520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236972817677644e-06,
+      "loss": 0.9738,
+      "step": 10521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.235200237973518e-06,
+      "loss": 0.9088,
+      "step": 10522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.233427929517618e-06,
+      "loss": 0.903,
+      "step": 10523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231655892393335e-06,
+      "loss": 0.9564,
+      "step": 10524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229884126684056e-06,
+      "loss": 0.9804,
+      "step": 10525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228112632473138e-06,
+      "loss": 0.897,
+      "step": 10526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.226341409843934e-06,
+      "loss": 0.983,
+      "step": 10527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224570458879791e-06,
+      "loss": 0.8899,
+      "step": 10528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222799779664032e-06,
+      "loss": 0.859,
+      "step": 10529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221029372279974e-06,
+      "loss": 0.9281,
+      "step": 10530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219259236810918e-06,
+      "loss": 0.9183,
+      "step": 10531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.217489373340151e-06,
+      "loss": 0.9031,
+      "step": 10532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.215719781950959e-06,
+      "loss": 0.9204,
+      "step": 10533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2139504627265915e-06,
+      "loss": 0.9355,
+      "step": 10534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21218141575031e-06,
+      "loss": 0.9585,
+      "step": 10535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210412641105349e-06,
+      "loss": 0.9348,
+      "step": 10536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.208644138874935e-06,
+      "loss": 0.9102,
+      "step": 10537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.206875909142279e-06,
+      "loss": 0.9283,
+      "step": 10538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2051079519905766e-06,
+      "loss": 0.9466,
+      "step": 10539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203340267503023e-06,
+      "loss": 0.9096,
+      "step": 10540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201572855762792e-06,
+      "loss": 0.8223,
+      "step": 10541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199805716853035e-06,
+      "loss": 0.9502,
+      "step": 10542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198038850856908e-06,
+      "loss": 0.9217,
+      "step": 10543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196272257857545e-06,
+      "loss": 0.964,
+      "step": 10544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19450593793807e-06,
+      "loss": 0.938,
+      "step": 10545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19273989118159e-06,
+      "loss": 0.9482,
+      "step": 10546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.190974117671199e-06,
+      "loss": 0.8848,
+      "step": 10547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.189208617489992e-06,
+      "loss": 0.8763,
+      "step": 10548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1874433907210275e-06,
+      "loss": 0.9609,
+      "step": 10549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185678437447372e-06,
+      "loss": 1.0073,
+      "step": 10550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.183913757752069e-06,
+      "loss": 0.9203,
+      "step": 10551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.182149351718151e-06,
+      "loss": 0.9363,
+      "step": 10552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180385219428637e-06,
+      "loss": 0.9281,
+      "step": 10553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178621360966531e-06,
+      "loss": 1.0491,
+      "step": 10554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1768577764148335e-06,
+      "loss": 0.9834,
+      "step": 10555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.175094465856526e-06,
+      "loss": 0.9073,
+      "step": 10556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1733314293745655e-06,
+      "loss": 0.9287,
+      "step": 10557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171568667051919e-06,
+      "loss": 0.9215,
+      "step": 10558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169806178971523e-06,
+      "loss": 0.9834,
+      "step": 10559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.168043965216308e-06,
+      "loss": 0.9717,
+      "step": 10560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166282025869193e-06,
+      "loss": 0.9213,
+      "step": 10561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164520361013075e-06,
+      "loss": 0.9711,
+      "step": 10562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.162758970730853e-06,
+      "loss": 1.0266,
+      "step": 10563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160997855105406e-06,
+      "loss": 0.9317,
+      "step": 10564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1592370142195845e-06,
+      "loss": 1.0183,
+      "step": 10565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.157476448156256e-06,
+      "loss": 0.9535,
+      "step": 10566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155716156998253e-06,
+      "loss": 0.9859,
+      "step": 10567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.153956140828401e-06,
+      "loss": 1.0104,
+      "step": 10568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.152196399729516e-06,
+      "loss": 0.985,
+      "step": 10569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.150436933784392e-06,
+      "loss": 0.9784,
+      "step": 10570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.148677743075824e-06,
+      "loss": 0.9359,
+      "step": 10571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1469188276865844e-06,
+      "loss": 0.9594,
+      "step": 10572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145160187699433e-06,
+      "loss": 1.025,
+      "step": 10573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.143401823197118e-06,
+      "loss": 0.8838,
+      "step": 10574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.141643734262375e-06,
+      "loss": 0.9008,
+      "step": 10575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139885920977927e-06,
+      "loss": 0.9547,
+      "step": 10576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.138128383426481e-06,
+      "loss": 0.8843,
+      "step": 10577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.136371121690739e-06,
+      "loss": 0.9256,
+      "step": 10578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134614135853381e-06,
+      "loss": 0.9955,
+      "step": 10579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.132857425997078e-06,
+      "loss": 0.9134,
+      "step": 10580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131100992204488e-06,
+      "loss": 0.9074,
+      "step": 10581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129344834558251e-06,
+      "loss": 0.9247,
+      "step": 10582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12758895314101e-06,
+      "loss": 0.9501,
+      "step": 10583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125833348035373e-06,
+      "loss": 0.947,
+      "step": 10584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124078019323946e-06,
+      "loss": 0.9305,
+      "step": 10585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1223229670893276e-06,
+      "loss": 0.8737,
+      "step": 10586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1205681914140936e-06,
+      "loss": 0.89,
+      "step": 10587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118813692380812e-06,
+      "loss": 0.9036,
+      "step": 10588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1170594700720364e-06,
+      "loss": 0.8756,
+      "step": 10589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1153055245703026e-06,
+      "loss": 0.9096,
+      "step": 10590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.113551855958149e-06,
+      "loss": 0.9415,
+      "step": 10591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111798464318077e-06,
+      "loss": 0.9886,
+      "step": 10592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110045349732598e-06,
+      "loss": 0.9051,
+      "step": 10593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108292512284196e-06,
+      "loss": 0.9573,
+      "step": 10594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1065399520553465e-06,
+      "loss": 0.948,
+      "step": 10595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.104787669128514e-06,
+      "loss": 0.8986,
+      "step": 10596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1030356635861415e-06,
+      "loss": 0.9715,
+      "step": 10597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1012839355106735e-06,
+      "loss": 0.938,
+      "step": 10598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099532484984534e-06,
+      "loss": 1.0014,
+      "step": 10599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097781312090121e-06,
+      "loss": 0.9119,
+      "step": 10600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0960304169098435e-06,
+      "loss": 1.0735,
+      "step": 10601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.094279799526081e-06,
+      "loss": 0.9253,
+      "step": 10602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.092529460021205e-06,
+      "loss": 0.8985,
+      "step": 10603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.090779398477573e-06,
+      "loss": 0.9252,
+      "step": 10604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.089029614977526e-06,
+      "loss": 0.9496,
+      "step": 10605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087280109603407e-06,
+      "loss": 0.9191,
+      "step": 10606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085530882437524e-06,
+      "loss": 0.9457,
+      "step": 10607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.083781933562182e-06,
+      "loss": 0.9143,
+      "step": 10608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.08203326305968e-06,
+      "loss": 0.8921,
+      "step": 10609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080284871012294e-06,
+      "loss": 0.9648,
+      "step": 10610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0785367575022925e-06,
+      "loss": 0.9047,
+      "step": 10611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076788922611925e-06,
+      "loss": 0.87,
+      "step": 10612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.07504136642343e-06,
+      "loss": 0.9853,
+      "step": 10613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073294089019045e-06,
+      "loss": 0.8849,
+      "step": 10614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.071547090480969e-06,
+      "loss": 0.9196,
+      "step": 10615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069800370891415e-06,
+      "loss": 0.9782,
+      "step": 10616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.068053930332564e-06,
+      "loss": 0.8852,
+      "step": 10617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066307768886592e-06,
+      "loss": 0.9177,
+      "step": 10618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064561886635661e-06,
+      "loss": 0.9453,
+      "step": 10619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.062816283661913e-06,
+      "loss": 0.927,
+      "step": 10620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061070960047493e-06,
+      "loss": 0.8686,
+      "step": 10621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059325915874522e-06,
+      "loss": 1.0133,
+      "step": 10622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.057581151225097e-06,
+      "loss": 0.8831,
+      "step": 10623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.055836666181325e-06,
+      "loss": 0.8728,
+      "step": 10624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.054092460825285e-06,
+      "loss": 0.9399,
+      "step": 10625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052348535239047e-06,
+      "loss": 0.9061,
+      "step": 10626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.050604889504665e-06,
+      "loss": 0.9405,
+      "step": 10627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.048861523704178e-06,
+      "loss": 0.9463,
+      "step": 10628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047118437919626e-06,
+      "loss": 0.9663,
+      "step": 10629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045375632233019e-06,
+      "loss": 1.0304,
+      "step": 10630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.043633106726362e-06,
+      "loss": 0.964,
+      "step": 10631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041890861481644e-06,
+      "loss": 0.974,
+      "step": 10632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0401488965808435e-06,
+      "loss": 0.9262,
+      "step": 10633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038407212105924e-06,
+      "loss": 0.9387,
+      "step": 10634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.036665808138834e-06,
+      "loss": 0.9781,
+      "step": 10635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03492468476151e-06,
+      "loss": 0.988,
+      "step": 10636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.033183842055881e-06,
+      "loss": 0.9574,
+      "step": 10637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031443280103858e-06,
+      "loss": 0.9528,
+      "step": 10638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.029702998987334e-06,
+      "loss": 0.9299,
+      "step": 10639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027962998788197e-06,
+      "loss": 0.935,
+      "step": 10640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.026223279588317e-06,
+      "loss": 0.9949,
+      "step": 10641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024483841469553e-06,
+      "loss": 0.9703,
+      "step": 10642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.022744684513746e-06,
+      "loss": 0.9689,
+      "step": 10643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021005808802735e-06,
+      "loss": 0.9339,
+      "step": 10644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0192672144183365e-06,
+      "loss": 0.9111,
+      "step": 10645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017528901442348e-06,
+      "loss": 1.0221,
+      "step": 10646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.015790869956571e-06,
+      "loss": 0.9175,
+      "step": 10647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014053120042776e-06,
+      "loss": 0.9217,
+      "step": 10648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0123156517827396e-06,
+      "loss": 1.0383,
+      "step": 10649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010578465258206e-06,
+      "loss": 0.9975,
+      "step": 10650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.00884156055091e-06,
+      "loss": 0.8934,
+      "step": 10651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007104937742587e-06,
+      "loss": 1.0221,
+      "step": 10652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0053685969149445e-06,
+      "loss": 0.9391,
+      "step": 10653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.003632538149684e-06,
+      "loss": 0.8445,
+      "step": 10654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.001896761528489e-06,
+      "loss": 0.9659,
+      "step": 10655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000161267133033e-06,
+      "loss": 0.9253,
+      "step": 10656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998426055044974e-06,
+      "loss": 0.9615,
+      "step": 10657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.996691125345957e-06,
+      "loss": 0.9535,
+      "step": 10658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994956478117621e-06,
+      "loss": 0.8955,
+      "step": 10659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9932221134415805e-06,
+      "loss": 0.8732,
+      "step": 10660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991488031399444e-06,
+      "loss": 0.9627,
+      "step": 10661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989754232072804e-06,
+      "loss": 0.9552,
+      "step": 10662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988020715543234e-06,
+      "loss": 0.9146,
+      "step": 10663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986287481892313e-06,
+      "loss": 0.9286,
+      "step": 10664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984554531201583e-06,
+      "loss": 0.8424,
+      "step": 10665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9828218635525836e-06,
+      "loss": 0.9546,
+      "step": 10666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.981089479026849e-06,
+      "loss": 0.9551,
+      "step": 10667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979357377705887e-06,
+      "loss": 0.878,
+      "step": 10668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.977625559671199e-06,
+      "loss": 0.9477,
+      "step": 10669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97589402500427e-06,
+      "loss": 0.9868,
+      "step": 10670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97416277378657e-06,
+      "loss": 0.9599,
+      "step": 10671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97243180609957e-06,
+      "loss": 0.9503,
+      "step": 10672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970701122024706e-06,
+      "loss": 0.9754,
+      "step": 10673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968970721643409e-06,
+      "loss": 0.9574,
+      "step": 10674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967240605037107e-06,
+      "loss": 0.9031,
+      "step": 10675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965510772287202e-06,
+      "loss": 1.0169,
+      "step": 10676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963781223475089e-06,
+      "loss": 1.0026,
+      "step": 10677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.962051958682146e-06,
+      "loss": 0.8518,
+      "step": 10678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.960322977989736e-06,
+      "loss": 0.946,
+      "step": 10679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.958594281479221e-06,
+      "loss": 0.8946,
+      "step": 10680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.956865869231929e-06,
+      "loss": 1.0426,
+      "step": 10681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955137741329194e-06,
+      "loss": 0.9369,
+      "step": 10682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953409897852329e-06,
+      "loss": 0.9795,
+      "step": 10683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.951682338882629e-06,
+      "loss": 0.9944,
+      "step": 10684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.949955064501383e-06,
+      "loss": 0.9189,
+      "step": 10685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.94822807478986e-06,
+      "loss": 0.8743,
+      "step": 10686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946501369829323e-06,
+      "loss": 0.987,
+      "step": 10687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.944774949701021e-06,
+      "loss": 0.8963,
+      "step": 10688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9430488144861745e-06,
+      "loss": 0.9042,
+      "step": 10689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.941322964266015e-06,
+      "loss": 0.8833,
+      "step": 10690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.939597399121741e-06,
+      "loss": 0.8447,
+      "step": 10691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937872119134547e-06,
+      "loss": 0.9264,
+      "step": 10692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936147124385611e-06,
+      "loss": 0.9968,
+      "step": 10693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.934422414956096e-06,
+      "loss": 0.9021,
+      "step": 10694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932697990927159e-06,
+      "loss": 0.9485,
+      "step": 10695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.930973852379936e-06,
+      "loss": 0.9909,
+      "step": 10696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929249999395554e-06,
+      "loss": 0.9333,
+      "step": 10697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92752643205512e-06,
+      "loss": 0.9266,
+      "step": 10698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925803150439735e-06,
+      "loss": 0.9395,
+      "step": 10699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924080154630485e-06,
+      "loss": 0.91,
+      "step": 10700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922357444708436e-06,
+      "loss": 0.8757,
+      "step": 10701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920635020754652e-06,
+      "loss": 0.9917,
+      "step": 10702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918912882850178e-06,
+      "loss": 0.9105,
+      "step": 10703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.917191031076035e-06,
+      "loss": 0.9563,
+      "step": 10704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.915469465513251e-06,
+      "loss": 0.9084,
+      "step": 10705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.913748186242823e-06,
+      "loss": 0.9192,
+      "step": 10706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912027193345751e-06,
+      "loss": 0.981,
+      "step": 10707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.910306486903002e-06,
+      "loss": 0.8274,
+      "step": 10708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90858606699554e-06,
+      "loss": 0.8899,
+      "step": 10709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906865933704321e-06,
+      "loss": 0.9751,
+      "step": 10710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.905146087110283e-06,
+      "loss": 0.9405,
+      "step": 10711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.903426527294338e-06,
+      "loss": 0.9395,
+      "step": 10712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901707254337406e-06,
+      "loss": 0.9147,
+      "step": 10713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.899988268320379e-06,
+      "loss": 0.9672,
+      "step": 10714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8982695693241405e-06,
+      "loss": 0.9928,
+      "step": 10715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.896551157429559e-06,
+      "loss": 0.9136,
+      "step": 10716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.894833032717485e-06,
+      "loss": 0.9238,
+      "step": 10717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.893115195268771e-06,
+      "loss": 0.9558,
+      "step": 10718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891397645164241e-06,
+      "loss": 0.9667,
+      "step": 10719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889680382484708e-06,
+      "loss": 0.9695,
+      "step": 10720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887963407310976e-06,
+      "loss": 0.895,
+      "step": 10721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886246719723832e-06,
+      "loss": 0.8746,
+      "step": 10722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88453031980405e-06,
+      "loss": 0.862,
+      "step": 10723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882814207632387e-06,
+      "loss": 0.9351,
+      "step": 10724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881098383289598e-06,
+      "loss": 0.94,
+      "step": 10725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879382846856414e-06,
+      "loss": 0.8778,
+      "step": 10726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877667598413555e-06,
+      "loss": 0.8613,
+      "step": 10727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875952638041726e-06,
+      "loss": 0.9297,
+      "step": 10728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874237965821619e-06,
+      "loss": 0.949,
+      "step": 10729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.872523581833922e-06,
+      "loss": 0.8599,
+      "step": 10730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870809486159294e-06,
+      "loss": 0.9028,
+      "step": 10731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869095678878383e-06,
+      "loss": 0.8625,
+      "step": 10732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867382160071838e-06,
+      "loss": 0.9327,
+      "step": 10733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.86566892982028e-06,
+      "loss": 0.9343,
+      "step": 10734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.86395598820432e-06,
+      "loss": 0.9002,
+      "step": 10735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862243335304558e-06,
+      "loss": 1.0024,
+      "step": 10736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860530971201573e-06,
+      "loss": 0.9166,
+      "step": 10737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.858818895975948e-06,
+      "loss": 0.9321,
+      "step": 10738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857107109708226e-06,
+      "loss": 0.8945,
+      "step": 10739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855395612478961e-06,
+      "loss": 0.9646,
+      "step": 10740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853684404368679e-06,
+      "loss": 0.999,
+      "step": 10741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851973485457899e-06,
+      "loss": 0.936,
+      "step": 10742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.850262855827122e-06,
+      "loss": 0.9418,
+      "step": 10743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848552515556835e-06,
+      "loss": 0.9679,
+      "step": 10744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.846842464727519e-06,
+      "loss": 0.9365,
+      "step": 10745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845132703419638e-06,
+      "loss": 0.9265,
+      "step": 10746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.84342323171363e-06,
+      "loss": 0.9535,
+      "step": 10747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8417140496899384e-06,
+      "loss": 0.9265,
+      "step": 10748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8400051574289845e-06,
+      "loss": 0.8622,
+      "step": 10749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838296555011172e-06,
+      "loss": 0.9976,
+      "step": 10750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.836588242516898e-06,
+      "loss": 0.91,
+      "step": 10751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834880220026538e-06,
+      "loss": 0.9776,
+      "step": 10752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.833172487620469e-06,
+      "loss": 0.9937,
+      "step": 10753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8314650453790335e-06,
+      "loss": 0.9755,
+      "step": 10754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8297578933825715e-06,
+      "loss": 0.9371,
+      "step": 10755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828051031711418e-06,
+      "loss": 0.8603,
+      "step": 10756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8263444604458765e-06,
+      "loss": 0.938,
+      "step": 10757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824638179666249e-06,
+      "loss": 0.9119,
+      "step": 10758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.82293218945282e-06,
+      "loss": 0.9996,
+      "step": 10759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.821226489885856e-06,
+      "loss": 0.9407,
+      "step": 10760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.819521081045625e-06,
+      "loss": 0.9782,
+      "step": 10761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817815963012358e-06,
+      "loss": 0.9602,
+      "step": 10762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.816111135866295e-06,
+      "loss": 0.9659,
+      "step": 10763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8144065996876478e-06,
+      "loss": 0.9972,
+      "step": 10764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.812702354556621e-06,
+      "loss": 0.9367,
+      "step": 10765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810998400553402e-06,
+      "loss": 0.8979,
+      "step": 10766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8092947377581635e-06,
+      "loss": 0.8986,
+      "step": 10767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8075913662510733e-06,
+      "loss": 0.9474,
+      "step": 10768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8058882861122792e-06,
+      "loss": 0.9341,
+      "step": 10769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.804185497421906e-06,
+      "loss": 0.9023,
+      "step": 10770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.802483000260084e-06,
+      "loss": 0.9236,
+      "step": 10771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8007807947069177e-06,
+      "loss": 0.9362,
+      "step": 10772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7990788808424973e-06,
+      "loss": 0.9071,
+      "step": 10773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7973772587469036e-06,
+      "loss": 0.8827,
+      "step": 10774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7956759285002e-06,
+      "loss": 0.9814,
+      "step": 10775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7939748901824435e-06,
+      "loss": 0.9511,
+      "step": 10776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.79227414387367e-06,
+      "loss": 0.96,
+      "step": 10777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7905736896539026e-06,
+      "loss": 0.8632,
+      "step": 10778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.788873527603152e-06,
+      "loss": 0.8717,
+      "step": 10779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7871736578014164e-06,
+      "loss": 0.9668,
+      "step": 10780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.785474080328677e-06,
+      "loss": 0.9218,
+      "step": 10781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7837747952649018e-06,
+      "loss": 0.9116,
+      "step": 10782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7820758026900526e-06,
+      "loss": 0.9524,
+      "step": 10783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7803771026840676e-06,
+      "loss": 0.8815,
+      "step": 10784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.778678695326874e-06,
+      "loss": 0.9297,
+      "step": 10785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.776980580698386e-06,
+      "loss": 0.9008,
+      "step": 10786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.775282758878507e-06,
+      "loss": 0.883,
+      "step": 10787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7735852299471208e-06,
+      "loss": 0.9258,
+      "step": 10788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7718879939841013e-06,
+      "loss": 0.9002,
+      "step": 10789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7701910510693053e-06,
+      "loss": 0.9578,
+      "step": 10790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7684944012825832e-06,
+      "loss": 0.9798,
+      "step": 10791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7667980447037634e-06,
+      "loss": 0.9002,
+      "step": 10792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7651019814126656e-06,
+      "loss": 0.9854,
+      "step": 10793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7634062114890925e-06,
+      "loss": 0.9159,
+      "step": 10794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.76171073501283e-06,
+      "loss": 0.9528,
+      "step": 10795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7600155520636673e-06,
+      "loss": 0.9785,
+      "step": 10796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7583206627213543e-06,
+      "loss": 0.895,
+      "step": 10797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7566260670656408e-06,
+      "loss": 0.9628,
+      "step": 10798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7549317651762686e-06,
+      "loss": 0.917,
+      "step": 10799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.753237757132955e-06,
+      "loss": 0.9534,
+      "step": 10800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.751544043015407e-06,
+      "loss": 0.9324,
+      "step": 10801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7498506229033193e-06,
+      "loss": 0.9023,
+      "step": 10802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7481574968763678e-06,
+      "loss": 0.9537,
+      "step": 10803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7464646650142277e-06,
+      "loss": 0.9774,
+      "step": 10804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.744772127396539e-06,
+      "loss": 0.9744,
+      "step": 10805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7430798841029468e-06,
+      "loss": 0.9263,
+      "step": 10806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7413879352130756e-06,
+      "loss": 0.9442,
+      "step": 10807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.739696280806534e-06,
+      "loss": 0.9689,
+      "step": 10808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.73800492096292e-06,
+      "loss": 0.9575,
+      "step": 10809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7363138557618097e-06,
+      "loss": 0.9849,
+      "step": 10810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7346230852827857e-06,
+      "loss": 0.8896,
+      "step": 10811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732932609605391e-06,
+      "loss": 0.992,
+      "step": 10812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7312424288091665e-06,
+      "loss": 1.0474,
+      "step": 10813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.729552542973647e-06,
+      "loss": 0.9656,
+      "step": 10814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7278629521783426e-06,
+      "loss": 0.9343,
+      "step": 10815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726173656502752e-06,
+      "loss": 0.9183,
+      "step": 10816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7244846560263604e-06,
+      "loss": 0.9464,
+      "step": 10817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.722795950828638e-06,
+      "loss": 1.0349,
+      "step": 10818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.721107540989051e-06,
+      "loss": 1.0023,
+      "step": 10819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7194194265870297e-06,
+      "loss": 0.9004,
+      "step": 10820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7177316077020167e-06,
+      "loss": 0.8258,
+      "step": 10821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7160440844134214e-06,
+      "loss": 0.9399,
+      "step": 10822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7143568568006483e-06,
+      "loss": 0.945,
+      "step": 10823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712669924943084e-06,
+      "loss": 0.8909,
+      "step": 10824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.710983288920105e-06,
+      "loss": 1.0002,
+      "step": 10825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.709296948811065e-06,
+      "loss": 0.8956,
+      "step": 10826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7076109046953247e-06,
+      "loss": 0.9309,
+      "step": 10827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7059251566522003e-06,
+      "loss": 0.9671,
+      "step": 10828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.704239704761021e-06,
+      "loss": 0.9418,
+      "step": 10829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7025545491010905e-06,
+      "loss": 0.9349,
+      "step": 10830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.700869689751696e-06,
+      "loss": 0.9999,
+      "step": 10831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6991851267921165e-06,
+      "loss": 0.9045,
+      "step": 10832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6975008603016114e-06,
+      "loss": 0.8766,
+      "step": 10833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6958168903594362e-06,
+      "loss": 0.9456,
+      "step": 10834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.694133217044825e-06,
+      "loss": 0.8981,
+      "step": 10835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.69244984043699e-06,
+      "loss": 1.017,
+      "step": 10836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6907667606151475e-06,
+      "loss": 0.9489,
+      "step": 10837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6890839776584887e-06,
+      "loss": 0.8657,
+      "step": 10838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.687401491646191e-06,
+      "loss": 0.9449,
+      "step": 10839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6857193026574188e-06,
+      "loss": 0.963,
+      "step": 10840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6840374107713228e-06,
+      "loss": 0.9329,
+      "step": 10841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.682355816067046e-06,
+      "loss": 0.8943,
+      "step": 10842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.680674518623707e-06,
+      "loss": 0.8371,
+      "step": 10843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.678993518520416e-06,
+      "loss": 1.0403,
+      "step": 10844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.677312815836268e-06,
+      "loss": 0.9566,
+      "step": 10845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.675632410650344e-06,
+      "loss": 0.9545,
+      "step": 10846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.673952303041711e-06,
+      "loss": 0.8709,
+      "step": 10847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6722724930894205e-06,
+      "loss": 0.9946,
+      "step": 10848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.670592980872518e-06,
+      "loss": 0.9517,
+      "step": 10849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.668913766470027e-06,
+      "loss": 0.9606,
+      "step": 10850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66723484996095e-06,
+      "loss": 0.9132,
+      "step": 10851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6655562314242956e-06,
+      "loss": 0.9701,
+      "step": 10852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6638779109390367e-06,
+      "loss": 0.8749,
+      "step": 10853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6621998885841557e-06,
+      "loss": 0.9121,
+      "step": 10854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.660522164438597e-06,
+      "loss": 0.96,
+      "step": 10855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.658844738581301e-06,
+      "loss": 0.9295,
+      "step": 10856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.657167611091201e-06,
+      "loss": 0.9722,
+      "step": 10857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6554907820472083e-06,
+      "loss": 0.9444,
+      "step": 10858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6538142515282214e-06,
+      "loss": 0.8834,
+      "step": 10859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.652138019613125e-06,
+      "loss": 0.8404,
+      "step": 10860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.650462086380786e-06,
+      "loss": 0.945,
+      "step": 10861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6487864519100723e-06,
+      "loss": 0.9885,
+      "step": 10862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6471111162798155e-06,
+      "loss": 0.9407,
+      "step": 10863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645436079568846e-06,
+      "loss": 0.9209,
+      "step": 10864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6437613418559836e-06,
+      "loss": 0.9783,
+      "step": 10865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.642086903220027e-06,
+      "loss": 0.9545,
+      "step": 10866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6404127637397615e-06,
+      "loss": 0.8785,
+      "step": 10867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.63873892349396e-06,
+      "loss": 0.938,
+      "step": 10868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.63706538256138e-06,
+      "loss": 0.9259,
+      "step": 10869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6353921410207673e-06,
+      "loss": 1.0304,
+      "step": 10870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.633719198950847e-06,
+      "loss": 0.9815,
+      "step": 10871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.632046556430344e-06,
+      "loss": 0.9739,
+      "step": 10872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.630374213537956e-06,
+      "loss": 0.9847,
+      "step": 10873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6287021703523705e-06,
+      "loss": 0.9587,
+      "step": 10874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6270304269522618e-06,
+      "loss": 0.8914,
+      "step": 10875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.625358983416285e-06,
+      "loss": 0.9424,
+      "step": 10876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.623687839823098e-06,
+      "loss": 0.961,
+      "step": 10877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6220169962513206e-06,
+      "loss": 0.9808,
+      "step": 10878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.620346452779571e-06,
+      "loss": 0.9293,
+      "step": 10879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.618676209486458e-06,
+      "loss": 0.9464,
+      "step": 10880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.617006266450569e-06,
+      "loss": 0.9307,
+      "step": 10881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.615336623750477e-06,
+      "loss": 0.9362,
+      "step": 10882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6136672814647446e-06,
+      "loss": 0.9185,
+      "step": 10883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6119982396719144e-06,
+      "loss": 0.9401,
+      "step": 10884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.610329498450529e-06,
+      "loss": 0.9832,
+      "step": 10885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6086610578790935e-06,
+      "loss": 0.9091,
+      "step": 10886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606992918036123e-06,
+      "loss": 0.8342,
+      "step": 10887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.605325079000104e-06,
+      "loss": 0.9149,
+      "step": 10888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.603657540849511e-06,
+      "loss": 0.9709,
+      "step": 10889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6019903036628067e-06,
+      "loss": 0.9921,
+      "step": 10890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.600323367518437e-06,
+      "loss": 0.9527,
+      "step": 10891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5986567324948406e-06,
+      "loss": 0.9507,
+      "step": 10892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5969903986704356e-06,
+      "loss": 0.9617,
+      "step": 10893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5953243661236204e-06,
+      "loss": 0.8996,
+      "step": 10894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.593658634932793e-06,
+      "loss": 0.9471,
+      "step": 10895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5919932051763284e-06,
+      "loss": 0.9268,
+      "step": 10896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.59032807693259e-06,
+      "loss": 1.0294,
+      "step": 10897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5886632502799236e-06,
+      "loss": 0.8668,
+      "step": 10898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.586998725296662e-06,
+      "loss": 1.0218,
+      "step": 10899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.585334502061132e-06,
+      "loss": 0.9601,
+      "step": 10900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.583670580651639e-06,
+      "loss": 0.8875,
+      "step": 10901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5820069611464646e-06,
+      "loss": 0.9578,
+      "step": 10902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.580343643623896e-06,
+      "loss": 0.9448,
+      "step": 10903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5786806281621943e-06,
+      "loss": 0.9488,
+      "step": 10904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5770179148396068e-06,
+      "loss": 0.9656,
+      "step": 10905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.57535550373437e-06,
+      "loss": 0.9021,
+      "step": 10906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5736933949247e-06,
+      "loss": 0.9697,
+      "step": 10907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5720315884888135e-06,
+      "loss": 0.8715,
+      "step": 10908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5703700845048893e-06,
+      "loss": 0.9306,
+      "step": 10909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.568708883051115e-06,
+      "loss": 0.8884,
+      "step": 10910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5670479842056527e-06,
+      "loss": 0.9564,
+      "step": 10911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5653873880466493e-06,
+      "loss": 0.9143,
+      "step": 10912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5637270946522405e-06,
+      "loss": 1.0069,
+      "step": 10913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.562067104100546e-06,
+      "loss": 0.966,
+      "step": 10914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5604074164696777e-06,
+      "loss": 0.9616,
+      "step": 10915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5587480318377276e-06,
+      "loss": 0.9855,
+      "step": 10916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5570889502827643e-06,
+      "loss": 0.9614,
+      "step": 10917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.555430171882861e-06,
+      "loss": 0.9286,
+      "step": 10918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5537716967160664e-06,
+      "loss": 0.9863,
+      "step": 10919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5521135248604133e-06,
+      "loss": 0.9226,
+      "step": 10920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.550455656393924e-06,
+      "loss": 0.8616,
+      "step": 10921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5487980913946028e-06,
+      "loss": 0.9368,
+      "step": 10922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5471408299404476e-06,
+      "loss": 0.9451,
+      "step": 10923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.545483872109433e-06,
+      "loss": 0.9541,
+      "step": 10924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.543827217979524e-06,
+      "loss": 0.9109,
+      "step": 10925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5421708676286704e-06,
+      "loss": 0.9729,
+      "step": 10926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.540514821134807e-06,
+      "loss": 0.9129,
+      "step": 10927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.538859078575856e-06,
+      "loss": 0.8649,
+      "step": 10928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5372036400297205e-06,
+      "loss": 0.9367,
+      "step": 10929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5355485055742976e-06,
+      "loss": 0.9024,
+      "step": 10930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5338936752874654e-06,
+      "loss": 0.9646,
+      "step": 10931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5322391492470853e-06,
+      "loss": 0.9341,
+      "step": 10932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.530584927531009e-06,
+      "loss": 0.9622,
+      "step": 10933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528931010217066e-06,
+      "loss": 1.012,
+      "step": 10934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.52727739738309e-06,
+      "loss": 0.95,
+      "step": 10935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.525624089106875e-06,
+      "loss": 0.9003,
+      "step": 10936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5239710854662147e-06,
+      "loss": 0.9358,
+      "step": 10937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5223183865388944e-06,
+      "loss": 0.9303,
+      "step": 10938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.520665992402672e-06,
+      "loss": 0.9349,
+      "step": 10939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.519013903135299e-06,
+      "loss": 1.0044,
+      "step": 10940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.517362118814509e-06,
+      "loss": 1.0294,
+      "step": 10941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5157106395180206e-06,
+      "loss": 0.991,
+      "step": 10942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5140594653235494e-06,
+      "loss": 0.8765,
+      "step": 10943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.512408596308776e-06,
+      "loss": 0.9419,
+      "step": 10944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510758032551379e-06,
+      "loss": 1.0656,
+      "step": 10945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.509107774129029e-06,
+      "loss": 0.9688,
+      "step": 10946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507457821119371e-06,
+      "loss": 0.9148,
+      "step": 10947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5058081736000384e-06,
+      "loss": 0.9405,
+      "step": 10948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.504158831648652e-06,
+      "loss": 0.9639,
+      "step": 10949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5025097953428133e-06,
+      "loss": 0.9123,
+      "step": 10950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.500861064760125e-06,
+      "loss": 0.9486,
+      "step": 10951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.49921263997815e-06,
+      "loss": 0.9484,
+      "step": 10952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.497564521074461e-06,
+      "loss": 1.0098,
+      "step": 10953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.495916708126602e-06,
+      "loss": 1.0115,
+      "step": 10954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.494269201212107e-06,
+      "loss": 0.9816,
+      "step": 10955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4926220004084965e-06,
+      "loss": 0.9675,
+      "step": 10956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4909751057932704e-06,
+      "loss": 0.9108,
+      "step": 10957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.489328517443927e-06,
+      "loss": 0.9649,
+      "step": 10958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4876822354379424e-06,
+      "loss": 0.9229,
+      "step": 10959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4860362598527674e-06,
+      "loss": 0.9444,
+      "step": 10960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4843905907658605e-06,
+      "loss": 0.9459,
+      "step": 10961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4827452282546493e-06,
+      "loss": 0.9163,
+      "step": 10962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4811001723965544e-06,
+      "loss": 0.8787,
+      "step": 10963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.479455423268978e-06,
+      "loss": 0.946,
+      "step": 10964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4778109809493075e-06,
+      "loss": 0.9221,
+      "step": 10965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4761668455149276e-06,
+      "loss": 0.9194,
+      "step": 10966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4745230170431854e-06,
+      "loss": 0.9792,
+      "step": 10967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4728794956114376e-06,
+      "loss": 0.9206,
+      "step": 10968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4712362812970122e-06,
+      "loss": 0.9084,
+      "step": 10969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.469593374177227e-06,
+      "loss": 0.9308,
+      "step": 10970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.467950774329385e-06,
+      "loss": 0.9192,
+      "step": 10971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4663084818307702e-06,
+      "loss": 0.9051,
+      "step": 10972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.464666496758664e-06,
+      "loss": 0.9427,
+      "step": 10973
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.463024819190327e-06,
+      "loss": 0.8781,
+      "step": 10974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.461383449202992e-06,
+      "loss": 0.9362,
+      "step": 10975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.459742386873901e-06,
+      "loss": 0.9281,
+      "step": 10976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4581016322802675e-06,
+      "loss": 0.9346,
+      "step": 10977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4564611854992903e-06,
+      "loss": 0.9538,
+      "step": 10978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.454821046608159e-06,
+      "loss": 0.9012,
+      "step": 10979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4531812156840426e-06,
+      "loss": 0.9256,
+      "step": 10980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4515416928041044e-06,
+      "loss": 0.9849,
+      "step": 10981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4499024780454904e-06,
+      "loss": 0.9333,
+      "step": 10982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4482635714853175e-06,
+      "loss": 0.958,
+      "step": 10983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4466249732007117e-06,
+      "loss": 0.9573,
+      "step": 10984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4449866832687696e-06,
+      "loss": 0.8883,
+      "step": 10985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4433487017665755e-06,
+      "loss": 0.9393,
+      "step": 10986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.441711028771203e-06,
+      "loss": 0.8941,
+      "step": 10987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.440073664359703e-06,
+      "loss": 0.8791,
+      "step": 10988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4384366086091257e-06,
+      "loss": 0.9339,
+      "step": 10989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.436799861596496e-06,
+      "loss": 0.8743,
+      "step": 10990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.435163423398824e-06,
+      "loss": 0.9717,
+      "step": 10991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.433527294093111e-06,
+      "loss": 0.9254,
+      "step": 10992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4318914737563403e-06,
+      "loss": 0.9643,
+      "step": 10993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4302559624654807e-06,
+      "loss": 0.9205,
+      "step": 10994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4286207602974843e-06,
+      "loss": 0.9789,
+      "step": 10995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4269858673292977e-06,
+      "loss": 0.9052,
+      "step": 10996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.425351283637843e-06,
+      "loss": 0.8854,
+      "step": 10997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4237170093000327e-06,
+      "loss": 0.9279,
+      "step": 10998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4220830443927634e-06,
+      "loss": 0.9857,
+      "step": 10999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4204493889929123e-06,
+      "loss": 0.9365,
+      "step": 11000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4188160431773586e-06,
+      "loss": 0.9209,
+      "step": 11001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.417183007022944e-06,
+      "loss": 0.9412,
+      "step": 11002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4155502806065086e-06,
+      "loss": 0.979,
+      "step": 11003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.413917864004881e-06,
+      "loss": 0.9547,
+      "step": 11004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4122857572948685e-06,
+      "loss": 0.9338,
+      "step": 11005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.410653960553265e-06,
+      "loss": 0.9653,
+      "step": 11006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.409022473856852e-06,
+      "loss": 0.9271,
+      "step": 11007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4073912972823896e-06,
+      "loss": 0.9295,
+      "step": 11008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4057604309066407e-06,
+      "loss": 0.9568,
+      "step": 11009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4041298748063266e-06,
+      "loss": 0.9322,
+      "step": 11010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40249962905818e-06,
+      "loss": 0.8531,
+      "step": 11011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4008696937389054e-06,
+      "loss": 0.9129,
+      "step": 11012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.399240068925194e-06,
+      "loss": 0.9047,
+      "step": 11013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3976107546937244e-06,
+      "loss": 0.938,
+      "step": 11014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.395981751121159e-06,
+      "loss": 0.9645,
+      "step": 11015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3943530582841476e-06,
+      "loss": 0.9664,
+      "step": 11016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3927246762593245e-06,
+      "loss": 0.9932,
+      "step": 11017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.391096605123304e-06,
+      "loss": 0.9397,
+      "step": 11018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.389468844952699e-06,
+      "loss": 0.9084,
+      "step": 11019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3878413958240973e-06,
+      "loss": 0.9691,
+      "step": 11020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.386214257814072e-06,
+      "loss": 0.9913,
+      "step": 11021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3845874309991856e-06,
+      "loss": 0.8079,
+      "step": 11022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3829609154559805e-06,
+      "loss": 0.9257,
+      "step": 11023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3813347112609996e-06,
+      "loss": 0.9333,
+      "step": 11024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3797088184907476e-06,
+      "loss": 0.9469,
+      "step": 11025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3780832372217278e-06,
+      "loss": 0.9236,
+      "step": 11026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3764579675304356e-06,
+      "loss": 0.9167,
+      "step": 11027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374833009493339e-06,
+      "loss": 0.9366,
+      "step": 11028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.373208363186897e-06,
+      "loss": 0.9467,
+      "step": 11029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3715840286875546e-06,
+      "loss": 0.9105,
+      "step": 11030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.369960006071735e-06,
+      "loss": 0.9308,
+      "step": 11031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.368336295415865e-06,
+      "loss": 0.9623,
+      "step": 11032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.366712896796329e-06,
+      "loss": 0.844,
+      "step": 11033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3650898102895225e-06,
+      "loss": 0.9353,
+      "step": 11034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3634670359718124e-06,
+      "loss": 0.9771,
+      "step": 11035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.361844573919555e-06,
+      "loss": 0.8939,
+      "step": 11036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3602224242090897e-06,
+      "loss": 0.8598,
+      "step": 11037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3586005869167413e-06,
+      "loss": 0.9865,
+      "step": 11038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356979062118827e-06,
+      "loss": 0.9346,
+      "step": 11039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3553578498916428e-06,
+      "loss": 0.8944,
+      "step": 11040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3537369503114627e-06,
+      "loss": 0.9439,
+      "step": 11041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3521163634545617e-06,
+      "loss": 0.9501,
+      "step": 11042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3504960893971903e-06,
+      "loss": 0.8544,
+      "step": 11043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348876128215588e-06,
+      "loss": 0.8918,
+      "step": 11044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.347256479985975e-06,
+      "loss": 0.9221,
+      "step": 11045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.345637144784558e-06,
+      "loss": 0.8881,
+      "step": 11046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344018122687539e-06,
+      "loss": 0.9209,
+      "step": 11047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3423994137710938e-06,
+      "loss": 0.8834,
+      "step": 11048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3407810181113787e-06,
+      "loss": 1.0173,
+      "step": 11049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3391629357845533e-06,
+      "loss": 0.9979,
+      "step": 11050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.337545166866748e-06,
+      "loss": 0.9059,
+      "step": 11051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.335927711434085e-06,
+      "loss": 0.9239,
+      "step": 11052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3343105695626667e-06,
+      "loss": 0.9324,
+      "step": 11053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3326937413285822e-06,
+      "loss": 0.9059,
+      "step": 11054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.331077226807914e-06,
+      "loss": 0.9377,
+      "step": 11055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.329461026076719e-06,
+      "loss": 0.9585,
+      "step": 11056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3278451392110434e-06,
+      "loss": 0.9826,
+      "step": 11057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.326229566286919e-06,
+      "loss": 0.969,
+      "step": 11058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3246143073803638e-06,
+      "loss": 0.9707,
+      "step": 11059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3229993625673785e-06,
+      "loss": 0.9172,
+      "step": 11060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3213847319239468e-06,
+      "loss": 0.9564,
+      "step": 11061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.319770415526048e-06,
+      "loss": 1.0255,
+      "step": 11062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3181564134496403e-06,
+      "loss": 0.9302,
+      "step": 11063
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316542725770656e-06,
+      "loss": 0.8353,
+      "step": 11064
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3149293525650328e-06,
+      "loss": 0.8862,
+      "step": 11065
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.313316293908678e-06,
+      "loss": 0.919,
+      "step": 11066
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3117035498774995e-06,
+      "loss": 0.9383,
+      "step": 11067
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3100911205473707e-06,
+      "loss": 0.927,
+      "step": 11068
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.308479005994162e-06,
+      "loss": 0.9547,
+      "step": 11069
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3068672062937314e-06,
+      "loss": 0.9417,
+      "step": 11070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3052557215219173e-06,
+      "loss": 0.8744,
+      "step": 11071
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3036445517545437e-06,
+      "loss": 0.9237,
+      "step": 11072
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3020336970674184e-06,
+      "loss": 0.9259,
+      "step": 11073
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3004231575363377e-06,
+      "loss": 0.9611,
+      "step": 11074
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.298812933237081e-06,
+      "loss": 0.986,
+      "step": 11075
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.29720302424541e-06,
+      "loss": 0.8671,
+      "step": 11076
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.295593430637083e-06,
+      "loss": 0.9169,
+      "step": 11077
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.293984152487829e-06,
+      "loss": 0.9478,
+      "step": 11078
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.292375189873371e-06,
+      "loss": 0.9439,
+      "step": 11079
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.290766542869415e-06,
+      "loss": 0.9139,
+      "step": 11080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2891582115516462e-06,
+      "loss": 0.9957,
+      "step": 11081
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2875501959957532e-06,
+      "loss": 0.9056,
+      "step": 11082
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2859424962773857e-06,
+      "loss": 0.906,
+      "step": 11083
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.28433511247219e-06,
+      "loss": 0.9529,
+      "step": 11084
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2827280446558043e-06,
+      "loss": 0.9949,
+      "step": 11085
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2811212929038426e-06,
+      "loss": 0.9349,
+      "step": 11086
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2795148572919055e-06,
+      "loss": 0.9672,
+      "step": 11087
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.27790873789558e-06,
+      "loss": 0.8648,
+      "step": 11088
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.276302934790435e-06,
+      "loss": 0.9978,
+      "step": 11089
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2746974480520377e-06,
+      "loss": 1.0275,
+      "step": 11090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2730922777559205e-06,
+      "loss": 0.9316,
+      "step": 11091
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2714874239776096e-06,
+      "loss": 0.9838,
+      "step": 11092
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.269882886792625e-06,
+      "loss": 0.9645,
+      "step": 11093
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2682786662764597e-06,
+      "loss": 0.8887,
+      "step": 11094
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266674762504598e-06,
+      "loss": 0.8753,
+      "step": 11095
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2650711755525065e-06,
+      "loss": 0.9484,
+      "step": 11096
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263467905495635e-06,
+      "loss": 0.8534,
+      "step": 11097
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2618649524094303e-06,
+      "loss": 0.9027,
+      "step": 11098
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.260262316369304e-06,
+      "loss": 0.8483,
+      "step": 11099
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2586599974506726e-06,
+      "loss": 1.0077,
+      "step": 11100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257057995728926e-06,
+      "loss": 0.8594,
+      "step": 11101
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2554563112794434e-06,
+      "loss": 0.952,
+      "step": 11102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2538549441775867e-06,
+      "loss": 0.9459,
+      "step": 11103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2522538944987012e-06,
+      "loss": 0.8889,
+      "step": 11104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.250653162318128e-06,
+      "loss": 0.9576,
+      "step": 11105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2490527477111855e-06,
+      "loss": 0.9032,
+      "step": 11106
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.247452650753166e-06,
+      "loss": 0.909,
+      "step": 11107
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.245852871519367e-06,
+      "loss": 0.947,
+      "step": 11108
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.244253410085062e-06,
+      "loss": 0.8944,
+      "step": 11109
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.242654266525508e-06,
+      "loss": 0.8941,
+      "step": 11110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241055440915948e-06,
+      "loss": 0.9734,
+      "step": 11111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2394569333316083e-06,
+      "loss": 0.9483,
+      "step": 11112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2378587438477126e-06,
+      "loss": 0.9328,
+      "step": 11113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2362608725394463e-06,
+      "loss": 0.9128,
+      "step": 11114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.234663319482003e-06,
+      "loss": 0.9911,
+      "step": 11115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.233066084750549e-06,
+      "loss": 0.9315,
+      "step": 11116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.231469168420237e-06,
+      "loss": 0.9824,
+      "step": 11117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2298725705662072e-06,
+      "loss": 0.9332,
+      "step": 11118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.228276291263579e-06,
+      "loss": 0.9267,
+      "step": 11119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.226680330587468e-06,
+      "loss": 0.9489,
+      "step": 11120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2250846886129704e-06,
+      "loss": 0.9191,
+      "step": 11121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.223489365415152e-06,
+      "loss": 0.8586,
+      "step": 11122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.221894361069088e-06,
+      "loss": 0.895,
+      "step": 11123
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.220299675649825e-06,
+      "loss": 0.8869,
+      "step": 11124
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.218705309232395e-06,
+      "loss": 1.0099,
+      "step": 11125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2171112618918187e-06,
+      "loss": 0.9686,
+      "step": 11126
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2155175337030953e-06,
+      "loss": 1.0297,
+      "step": 11127
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.213924124741222e-06,
+      "loss": 0.9091,
+      "step": 11128
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.212331035081171e-06,
+      "loss": 0.9472,
+      "step": 11129
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2107382647978914e-06,
+      "loss": 0.8542,
+      "step": 11130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2091458139663388e-06,
+      "loss": 0.9616,
+      "step": 11131
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.207553682661436e-06,
+      "loss": 0.9783,
+      "step": 11132
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2059618709580987e-06,
+      "loss": 0.9598,
+      "step": 11133
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204370378931225e-06,
+      "loss": 0.9923,
+      "step": 11134
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2027792066556952e-06,
+      "loss": 0.9344,
+      "step": 11135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2011883542063837e-06,
+      "loss": 0.8759,
+      "step": 11136
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1995978216581423e-06,
+      "loss": 0.9261,
+      "step": 11137
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.198007609085809e-06,
+      "loss": 0.9214,
+      "step": 11138
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1964177165642073e-06,
+      "loss": 0.9303,
+      "step": 11139
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.194828144168146e-06,
+      "loss": 0.9676,
+      "step": 11140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1932388919724166e-06,
+      "loss": 0.9263,
+      "step": 11141
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.191649960051797e-06,
+      "loss": 0.9242,
+      "step": 11142
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.190061348481054e-06,
+      "loss": 0.9912,
+      "step": 11143
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1884730573349342e-06,
+      "loss": 0.9806,
+      "step": 11144
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1868850866881705e-06,
+      "loss": 0.9129,
+      "step": 11145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185297436615481e-06,
+      "loss": 0.917,
+      "step": 11146
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.183710107191564e-06,
+      "loss": 0.9767,
+      "step": 11147
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.18212309849112e-06,
+      "loss": 0.9853,
+      "step": 11148
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.180536410588809e-06,
+      "loss": 0.9082,
+      "step": 11149
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17895004355929e-06,
+      "loss": 0.8853,
+      "step": 11150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1773639974772108e-06,
+      "loss": 0.9313,
+      "step": 11151
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.175778272417198e-06,
+      "loss": 0.9766,
+      "step": 11152
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1741928684538614e-06,
+      "loss": 0.933,
+      "step": 11153
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1726077856617996e-06,
+      "loss": 0.9656,
+      "step": 11154
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1710230241155915e-06,
+      "loss": 0.9766,
+      "step": 11155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1694385838898136e-06,
+      "loss": 0.8911,
+      "step": 11156
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1678544650590036e-06,
+      "loss": 0.9201,
+      "step": 11157
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1662706676977104e-06,
+      "loss": 0.9073,
+      "step": 11158
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.16468719188045e-06,
+      "loss": 0.8936,
+      "step": 11159
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.16310403768173e-06,
+      "loss": 0.962,
+      "step": 11160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1615212051760425e-06,
+      "loss": 0.9105,
+      "step": 11161
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.159938694437863e-06,
+      "loss": 0.9648,
+      "step": 11162
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1583565055416477e-06,
+      "loss": 0.9512,
+      "step": 11163
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.156774638561855e-06,
+      "loss": 1.0565,
+      "step": 11164
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1551930935729014e-06,
+      "loss": 0.8875,
+      "step": 11165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1536118706492125e-06,
+      "loss": 1.003,
+      "step": 11166
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1520309698651853e-06,
+      "loss": 0.91,
+      "step": 11167
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1504503912952044e-06,
+      "loss": 0.8048,
+      "step": 11168
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.148870135013642e-06,
+      "loss": 0.8581,
+      "step": 11169
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.147290201094848e-06,
+      "loss": 0.9312,
+      "step": 11170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1457105896131735e-06,
+      "loss": 0.9706,
+      "step": 11171
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.144131300642932e-06,
+      "loss": 0.9089,
+      "step": 11172
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1425523342584343e-06,
+      "loss": 0.8936,
+      "step": 11173
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1409736905339795e-06,
+      "loss": 0.8565,
+      "step": 11174
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.139395369543845e-06,
+      "loss": 0.9617,
+      "step": 11175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.137817371362294e-06,
+      "loss": 0.9785,
+      "step": 11176
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1362396960635755e-06,
+      "loss": 0.8729,
+      "step": 11177
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1346623437219204e-06,
+      "loss": 0.8431,
+      "step": 11178
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1330853144115547e-06,
+      "loss": 0.9461,
+      "step": 11179
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1315086082066713e-06,
+      "loss": 0.8266,
+      "step": 11180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.129932225181467e-06,
+      "loss": 0.8815,
+      "step": 11181
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.12835616541011e-06,
+      "loss": 0.9279,
+      "step": 11182
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1267804289667602e-06,
+      "loss": 1.009,
+      "step": 11183
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1252050159255574e-06,
+      "loss": 0.9612,
+      "step": 11184
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.123629926360627e-06,
+      "loss": 0.931,
+      "step": 11185
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1220551603460868e-06,
+      "loss": 0.867,
+      "step": 11186
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.120480717956035e-06,
+      "loss": 0.9709,
+      "step": 11187
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1189065992645405e-06,
+      "loss": 0.964,
+      "step": 11188
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117332804345682e-06,
+      "loss": 0.9238,
+      "step": 11189
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.115759333273505e-06,
+      "loss": 0.9035,
+      "step": 11190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1141861861220467e-06,
+      "loss": 0.9386,
+      "step": 11191
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1126133629653266e-06,
+      "loss": 0.8682,
+      "step": 11192
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1110408638773483e-06,
+      "loss": 0.9995,
+      "step": 11193
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1094686889321068e-06,
+      "loss": 0.8547,
+      "step": 11194
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1078968382035747e-06,
+      "loss": 0.8905,
+      "step": 11195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.10632531176571e-06,
+      "loss": 0.9037,
+      "step": 11196
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104754109692458e-06,
+      "loss": 0.9154,
+      "step": 11197
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.103183232057748e-06,
+      "loss": 0.9356,
+      "step": 11198
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.101612678935493e-06,
+      "loss": 1.0085,
+      "step": 11199
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.100042450399592e-06,
+      "loss": 0.9627,
+      "step": 11200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.098472546523924e-06,
+      "loss": 0.9422,
+      "step": 11201
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0969029673823636e-06,
+      "loss": 0.944,
+      "step": 11202
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0953337130487616e-06,
+      "loss": 0.9582,
+      "step": 11203
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0937647835969543e-06,
+      "loss": 1.019,
+      "step": 11204
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.092196179100764e-06,
+      "loss": 0.9757,
+      "step": 11205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0906278996339955e-06,
+      "loss": 0.7984,
+      "step": 11206
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089059945270444e-06,
+      "loss": 0.8509,
+      "step": 11207
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.087492316083879e-06,
+      "loss": 0.9279,
+      "step": 11208
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.08592501214807e-06,
+      "loss": 0.8955,
+      "step": 11209
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0843580335367616e-06,
+      "loss": 0.9524,
+      "step": 11210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0827913803236754e-06,
+      "loss": 0.9328,
+      "step": 11211
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.081225052582534e-06,
+      "loss": 0.9286,
+      "step": 11212
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.079659050387033e-06,
+      "loss": 0.9741,
+      "step": 11213
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0780933738108644e-06,
+      "loss": 0.9321,
+      "step": 11214
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.076528022927688e-06,
+      "loss": 0.932,
+      "step": 11215
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0749629978111583e-06,
+      "loss": 0.8708,
+      "step": 11216
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0733982985349196e-06,
+      "loss": 0.9184,
+      "step": 11217
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.071833925172593e-06,
+      "loss": 0.9297,
+      "step": 11218
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0702698777977826e-06,
+      "loss": 0.8837,
+      "step": 11219
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068706156484084e-06,
+      "loss": 0.9254,
+      "step": 11220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0671427613050698e-06,
+      "loss": 0.9606,
+      "step": 11221
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0655796923343116e-06,
+      "loss": 0.905,
+      "step": 11222
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.064016949645342e-06,
+      "loss": 0.9357,
+      "step": 11223
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0624545333117027e-06,
+      "loss": 0.856,
+      "step": 11224
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0608924434069066e-06,
+      "loss": 0.8448,
+      "step": 11225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.059330680004452e-06,
+      "loss": 1.0383,
+      "step": 11226
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.057769243177825e-06,
+      "loss": 0.9208,
+      "step": 11227
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.056208133000491e-06,
+      "loss": 0.8291,
+      "step": 11228
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.054647349545915e-06,
+      "loss": 0.9954,
+      "step": 11229
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0530868928875257e-06,
+      "loss": 0.9558,
+      "step": 11230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.051526763098747e-06,
+      "loss": 0.9536,
+      "step": 11231
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0499669602529914e-06,
+      "loss": 0.8798,
+      "step": 11232
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.048407484423649e-06,
+      "loss": 0.9236,
+      "step": 11233
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0468483356840993e-06,
+      "loss": 0.9056,
+      "step": 11234
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.045289514107701e-06,
+      "loss": 0.9593,
+      "step": 11235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0437310197677995e-06,
+      "loss": 0.9843,
+      "step": 11236
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.042172852737735e-06,
+      "loss": 0.9243,
+      "step": 11237
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.040615013090813e-06,
+      "loss": 0.9675,
+      "step": 11238
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0390575009003344e-06,
+      "loss": 1.0008,
+      "step": 11239
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037500316239589e-06,
+      "loss": 0.9667,
+      "step": 11240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0359434591818446e-06,
+      "loss": 0.9981,
+      "step": 11241
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0343869298003547e-06,
+      "loss": 0.9803,
+      "step": 11242
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0328307281683576e-06,
+      "loss": 0.9131,
+      "step": 11243
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0312748543590744e-06,
+      "loss": 0.8493,
+      "step": 11244
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.029719308445721e-06,
+      "loss": 0.8704,
+      "step": 11245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0281640905014787e-06,
+      "loss": 0.868,
+      "step": 11246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.026609200599532e-06,
+      "loss": 0.8536,
+      "step": 11247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.02505463881304e-06,
+      "loss": 0.9025,
+      "step": 11248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0235004052151485e-06,
+      "loss": 0.9668,
+      "step": 11249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021946499878988e-06,
+      "loss": 0.8542,
+      "step": 11250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.020392922877671e-06,
+      "loss": 0.8987,
+      "step": 11251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0188396742843027e-06,
+      "loss": 0.9619,
+      "step": 11252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0172867541719696e-06,
+      "loss": 1.0211,
+      "step": 11253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.015734162613728e-06,
+      "loss": 0.9904,
+      "step": 11254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.014181899682642e-06,
+      "loss": 0.9759,
+      "step": 11255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0126299654517467e-06,
+      "loss": 1.0163,
+      "step": 11256
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.011078359994063e-06,
+      "loss": 0.893,
+      "step": 11257
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0095270833825995e-06,
+      "loss": 0.9164,
+      "step": 11258
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0079761356903447e-06,
+      "loss": 0.9924,
+      "step": 11259
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.006425516990279e-06,
+      "loss": 0.852,
+      "step": 11260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0048752273553607e-06,
+      "loss": 0.9567,
+      "step": 11261
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0033252668585345e-06,
+      "loss": 0.952,
+      "step": 11262
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.001775635572731e-06,
+      "loss": 0.9268,
+      "step": 11263
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0002263335708627e-06,
+      "loss": 0.9783,
+      "step": 11264
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.99867736092583e-06,
+      "loss": 0.9499,
+      "step": 11265
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9971287177105102e-06,
+      "loss": 0.9088,
+      "step": 11266
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9955804039977797e-06,
+      "loss": 0.9296,
+      "step": 11267
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9940324198604898e-06,
+      "loss": 0.8757,
+      "step": 11268
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.992484765371466e-06,
+      "loss": 0.8962,
+      "step": 11269
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.990937440603541e-06,
+      "loss": 0.9321,
+      "step": 11270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.989390445629512e-06,
+      "loss": 0.9288,
+      "step": 11271
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9878437805221793e-06,
+      "loss": 0.9687,
+      "step": 11272
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.986297445354308e-06,
+      "loss": 0.8745,
+      "step": 11273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.984751440198658e-06,
+      "loss": 0.8933,
+      "step": 11274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.983205765127978e-06,
+      "loss": 0.9349,
+      "step": 11275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9816604202149945e-06,
+      "loss": 0.9357,
+      "step": 11276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9801154055324135e-06,
+      "loss": 0.9598,
+      "step": 11277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.978570721152938e-06,
+      "loss": 0.9512,
+      "step": 11278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.977026367149245e-06,
+      "loss": 0.9395,
+      "step": 11279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9754823435940095e-06,
+      "loss": 0.9799,
+      "step": 11280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9739386505598712e-06,
+      "loss": 0.9805,
+      "step": 11281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9723952881194663e-06,
+      "loss": 0.9486,
+      "step": 11282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.970852256345419e-06,
+      "loss": 0.9386,
+      "step": 11283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.969309555310329e-06,
+      "loss": 0.9124,
+      "step": 11284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9677671850867862e-06,
+      "loss": 1.0031,
+      "step": 11285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.966225145747361e-06,
+      "loss": 0.9238,
+      "step": 11286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964683437364612e-06,
+      "loss": 0.9346,
+      "step": 11287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.963142060011078e-06,
+      "loss": 0.884,
+      "step": 11288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9616010137592845e-06,
+      "loss": 0.9191,
+      "step": 11289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9600602986817463e-06,
+      "loss": 0.8759,
+      "step": 11290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.958519914850955e-06,
+      "loss": 0.8811,
+      "step": 11291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9569798623393886e-06,
+      "loss": 0.9779,
+      "step": 11292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.955440141219511e-06,
+      "loss": 0.8866,
+      "step": 11293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9539007515637674e-06,
+      "loss": 0.959,
+      "step": 11294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.952361693444599e-06,
+      "loss": 0.8997,
+      "step": 11295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9508229669344123e-06,
+      "loss": 0.9159,
+      "step": 11296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9492845721056096e-06,
+      "loss": 0.8691,
+      "step": 11297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.947746509030581e-06,
+      "loss": 0.8922,
+      "step": 11298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.946208777781694e-06,
+      "loss": 0.93,
+      "step": 11299
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.944671378431302e-06,
+      "loss": 0.9769,
+      "step": 11300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9431343110517443e-06,
+      "loss": 0.9693,
+      "step": 11301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9415975757153394e-06,
+      "loss": 1.0008,
+      "step": 11302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.940061172494406e-06,
+      "loss": 0.9081,
+      "step": 11303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9385251014612215e-06,
+      "loss": 1.0257,
+      "step": 11304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936989362688071e-06,
+      "loss": 0.918,
+      "step": 11305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.935453956247213e-06,
+      "loss": 0.9381,
+      "step": 11306
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.933918882210891e-06,
+      "loss": 0.9536,
+      "step": 11307
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9323841406513354e-06,
+      "loss": 0.9374,
+      "step": 11308
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.930849731640756e-06,
+      "loss": 0.949,
+      "step": 11309
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.929315655251356e-06,
+      "loss": 0.8894,
+      "step": 11310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9277819115553185e-06,
+      "loss": 0.9269,
+      "step": 11311
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9262485006247997e-06,
+      "loss": 0.9633,
+      "step": 11312
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.924715422531961e-06,
+      "loss": 0.8724,
+      "step": 11313
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9231826773489335e-06,
+      "loss": 0.9901,
+      "step": 11314
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9216502651478363e-06,
+      "loss": 0.9792,
+      "step": 11315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9201181860007745e-06,
+      "loss": 0.9813,
+      "step": 11316
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9185864399798324e-06,
+      "loss": 0.9461,
+      "step": 11317
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9170550271570886e-06,
+      "loss": 0.9491,
+      "step": 11318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9155239476046003e-06,
+      "loss": 0.9583,
+      "step": 11319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9139932013943995e-06,
+      "loss": 0.9636,
+      "step": 11320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.91246278859852e-06,
+      "loss": 0.9166,
+      "step": 11321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9109327092889694e-06,
+      "loss": 0.9892,
+      "step": 11322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.909402963537741e-06,
+      "loss": 0.9253,
+      "step": 11323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907873551416813e-06,
+      "loss": 0.9113,
+      "step": 11324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.906344472998146e-06,
+      "loss": 0.916,
+      "step": 11325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.904815728353696e-06,
+      "loss": 0.8825,
+      "step": 11326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.903287317555381e-06,
+      "loss": 1.0153,
+      "step": 11327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901759240675127e-06,
+      "loss": 0.9098,
+      "step": 11328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9002314977848302e-06,
+      "loss": 0.9286,
+      "step": 11329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8987040889563733e-06,
+      "loss": 0.8705,
+      "step": 11330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8971770142616275e-06,
+      "loss": 0.9261,
+      "step": 11331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8956502737724403e-06,
+      "loss": 0.9154,
+      "step": 11332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894123867560655e-06,
+      "loss": 0.9398,
+      "step": 11333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8925977956980943e-06,
+      "loss": 0.9551,
+      "step": 11334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8910720582565534e-06,
+      "loss": 0.9198,
+      "step": 11335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8895466553078312e-06,
+      "loss": 0.9197,
+      "step": 11336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.888021586923698e-06,
+      "loss": 1.0153,
+      "step": 11337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.886496853175914e-06,
+      "loss": 0.928,
+      "step": 11338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.88497245413622e-06,
+      "loss": 0.9757,
+      "step": 11339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.883448389876341e-06,
+      "loss": 0.9293,
+      "step": 11340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.881924660467993e-06,
+      "loss": 0.9817,
+      "step": 11341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8804012659828685e-06,
+      "loss": 0.9496,
+      "step": 11342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8788782064926467e-06,
+      "loss": 0.9111,
+      "step": 11343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8773554820689932e-06,
+      "loss": 0.9202,
+      "step": 11344
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875833092783553e-06,
+      "loss": 0.9072,
+      "step": 11345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.874311038707961e-06,
+      "loss": 0.9823,
+      "step": 11346
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.872789319913829e-06,
+      "loss": 0.887,
+      "step": 11347
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8712679364727648e-06,
+      "loss": 0.883,
+      "step": 11348
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8697468884563497e-06,
+      "loss": 0.8899,
+      "step": 11349
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8682261759361528e-06,
+      "loss": 0.9563,
+      "step": 11350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8667057989837276e-06,
+      "loss": 0.9378,
+      "step": 11351
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.865185757670612e-06,
+      "loss": 0.9554,
+      "step": 11352
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8636660520683266e-06,
+      "loss": 0.8345,
+      "step": 11353
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.862146682248378e-06,
+      "loss": 0.9701,
+      "step": 11354
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8606276482822535e-06,
+      "loss": 0.9116,
+      "step": 11355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8591089502414327e-06,
+      "loss": 0.9696,
+      "step": 11356
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.857590588197373e-06,
+      "loss": 0.915,
+      "step": 11357
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8560725622215146e-06,
+      "loss": 0.929,
+      "step": 11358
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.854554872385287e-06,
+      "loss": 0.8612,
+      "step": 11359
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8530375187600956e-06,
+      "loss": 0.9253,
+      "step": 11360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8515205014173474e-06,
+      "loss": 0.8667,
+      "step": 11361
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8500038204284108e-06,
+      "loss": 0.9862,
+      "step": 11362
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.84848747586465e-06,
+      "loss": 1.0231,
+      "step": 11363
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8469714677974192e-06,
+      "loss": 0.9456,
+      "step": 11364
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8454557962980466e-06,
+      "loss": 0.8872,
+      "step": 11365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8439404614378497e-06,
+      "loss": 0.8831,
+      "step": 11366
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8424254632881265e-06,
+      "loss": 0.8911,
+      "step": 11367
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8409108019201614e-06,
+      "loss": 0.8692,
+      "step": 11368
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8393964774052296e-06,
+      "loss": 0.9147,
+      "step": 11369
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.837882489814573e-06,
+      "loss": 0.8826,
+      "step": 11370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8363688392194377e-06,
+      "loss": 1.0128,
+      "step": 11371
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8348555256910405e-06,
+      "loss": 0.9463,
+      "step": 11372
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.833342549300587e-06,
+      "loss": 1.0072,
+      "step": 11373
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831829910119268e-06,
+      "loss": 0.904,
+      "step": 11374
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8303176082182513e-06,
+      "loss": 0.9764,
+      "step": 11375
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.828805643668704e-06,
+      "loss": 0.9698,
+      "step": 11376
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.827294016541765e-06,
+      "loss": 0.9069,
+      "step": 11377
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.825782726908551e-06,
+      "loss": 0.9098,
+      "step": 11378
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8242717748401827e-06,
+      "loss": 0.9427,
+      "step": 11379
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.82276116040775e-06,
+      "loss": 0.9876,
+      "step": 11380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8212508836823314e-06,
+      "loss": 1.0077,
+      "step": 11381
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.81974094473499e-06,
+      "loss": 0.915,
+      "step": 11382
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8182313436367683e-06,
+      "loss": 0.9045,
+      "step": 11383
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.816722080458705e-06,
+      "loss": 0.9593,
+      "step": 11384
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.815213155271804e-06,
+      "loss": 0.9473,
+      "step": 11385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.813704568147073e-06,
+      "loss": 0.9319,
+      "step": 11386
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8121963191554914e-06,
+      "loss": 0.9719,
+      "step": 11387
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8106884083680264e-06,
+      "loss": 0.9247,
+      "step": 11388
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.809180835855627e-06,
+      "loss": 0.9149,
+      "step": 11389
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8076736016892312e-06,
+      "loss": 0.952,
+      "step": 11390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8061667059397544e-06,
+      "loss": 0.8736,
+      "step": 11391
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8046601486781078e-06,
+      "loss": 0.8492,
+      "step": 11392
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8031539299751663e-06,
+      "loss": 0.8666,
+      "step": 11393
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8016480499018115e-06,
+      "loss": 0.9121,
+      "step": 11394
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.800142508528897e-06,
+      "loss": 0.9004,
+      "step": 11395
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7986373059272587e-06,
+      "loss": 0.9048,
+      "step": 11396
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7971324421677227e-06,
+      "loss": 0.855,
+      "step": 11397
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.795627917321093e-06,
+      "loss": 0.9554,
+      "step": 11398
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.794123731458168e-06,
+      "loss": 0.9468,
+      "step": 11399
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.792619884649722e-06,
+      "loss": 0.9639,
+      "step": 11400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791116376966506e-06,
+      "loss": 0.9533,
+      "step": 11401
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789613208479275e-06,
+      "loss": 0.9115,
+      "step": 11402
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7881103792587505e-06,
+      "loss": 1.0027,
+      "step": 11403
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7866078893756467e-06,
+      "loss": 0.8998,
+      "step": 11404
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7851057389006587e-06,
+      "loss": 0.9035,
+      "step": 11405
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7836039279044625e-06,
+      "loss": 1.0077,
+      "step": 11406
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7821024564577294e-06,
+      "loss": 0.8827,
+      "step": 11407
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7806013246311046e-06,
+      "loss": 0.8621,
+      "step": 11408
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.779100532495219e-06,
+      "loss": 0.9379,
+      "step": 11409
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7776000801206892e-06,
+      "loss": 1.026,
+      "step": 11410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.776099967578115e-06,
+      "loss": 0.9513,
+      "step": 11411
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774600194938082e-06,
+      "loss": 0.981,
+      "step": 11412
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7731007622711524e-06,
+      "loss": 0.9197,
+      "step": 11413
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7716016696478865e-06,
+      "loss": 0.9515,
+      "step": 11414
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7701029171388163e-06,
+      "loss": 0.9149,
+      "step": 11415
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7686045048144614e-06,
+      "loss": 0.9721,
+      "step": 11416
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.767106432745327e-06,
+      "loss": 0.8553,
+      "step": 11417
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7656087010018973e-06,
+      "loss": 0.8632,
+      "step": 11418
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7641113096546546e-06,
+      "loss": 0.9685,
+      "step": 11419
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7626142587740444e-06,
+      "loss": 0.9303,
+      "step": 11420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7611175484305073e-06,
+      "loss": 0.9736,
+      "step": 11421
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7596211786944736e-06,
+      "loss": 0.8479,
+      "step": 11422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7581251496363504e-06,
+      "loss": 0.9339,
+      "step": 11423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.756629461326522e-06,
+      "loss": 0.9033,
+      "step": 11424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7551341138353715e-06,
+      "loss": 0.9107,
+      "step": 11425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7536391072332537e-06,
+      "loss": 0.9491,
+      "step": 11426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7521444415905218e-06,
+      "loss": 0.9222,
+      "step": 11427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7506501169774935e-06,
+      "loss": 0.9035,
+      "step": 11428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.749156133464481e-06,
+      "loss": 0.9669,
+      "step": 11429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7476624911217864e-06,
+      "loss": 0.9623,
+      "step": 11430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7461691900196854e-06,
+      "loss": 0.9104,
+      "step": 11431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7446762302284424e-06,
+      "loss": 0.9312,
+      "step": 11432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7431836118183043e-06,
+      "loss": 0.9115,
+      "step": 11433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.741691334859503e-06,
+      "loss": 0.9004,
+      "step": 11434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.740199399422253e-06,
+      "loss": 0.9302,
+      "step": 11435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7387078055767523e-06,
+      "loss": 0.8867,
+      "step": 11436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.737216553393188e-06,
+      "loss": 0.8857,
+      "step": 11437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7357256429417257e-06,
+      "loss": 0.9215,
+      "step": 11438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.734235074292516e-06,
+      "loss": 0.9255,
+      "step": 11439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.732744847515694e-06,
+      "loss": 0.8843,
+      "step": 11440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7312549626813756e-06,
+      "loss": 0.9221,
+      "step": 11441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7297654198596725e-06,
+      "loss": 0.9311,
+      "step": 11442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7282762191206624e-06,
+      "loss": 0.9037,
+      "step": 11443
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7267873605344153e-06,
+      "loss": 0.9803,
+      "step": 11444
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7252988441709927e-06,
+      "loss": 0.9119,
+      "step": 11445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7238106701004298e-06,
+      "loss": 0.9749,
+      "step": 11446
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.722322838392749e-06,
+      "loss": 0.8802,
+      "step": 11447
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.720835349117955e-06,
+      "loss": 0.9443,
+      "step": 11448
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.719348202346037e-06,
+      "loss": 0.9289,
+      "step": 11449
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7178613981469782e-06,
+      "loss": 0.9569,
+      "step": 11450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.716374936590722e-06,
+      "loss": 0.8988,
+      "step": 11451
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.714888817747221e-06,
+      "loss": 0.9494,
+      "step": 11452
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7134030416863966e-06,
+      "loss": 0.9486,
+      "step": 11453
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7119176084781596e-06,
+      "loss": 0.935,
+      "step": 11454
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.710432518192403e-06,
+      "loss": 0.9555,
+      "step": 11455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7089477708989996e-06,
+      "loss": 0.893,
+      "step": 11456
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7074633666678184e-06,
+      "loss": 0.9203,
+      "step": 11457
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7059793055687032e-06,
+      "loss": 0.9596,
+      "step": 11458
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.704495587671475e-06,
+      "loss": 0.9708,
+      "step": 11459
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7030122130459536e-06,
+      "loss": 0.917,
+      "step": 11460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.701529181761934e-06,
+      "loss": 0.9797,
+      "step": 11461
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7000464938891944e-06,
+      "loss": 0.9648,
+      "step": 11462
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6985641494975024e-06,
+      "loss": 0.9614,
+      "step": 11463
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6970821486566e-06,
+      "loss": 0.8898,
+      "step": 11464
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6956004914362265e-06,
+      "loss": 0.8622,
+      "step": 11465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.694119177906097e-06,
+      "loss": 0.8824,
+      "step": 11466
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6926382081359026e-06,
+      "loss": 0.8548,
+      "step": 11467
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.691157582195335e-06,
+      "loss": 0.9494,
+      "step": 11468
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.689677300154059e-06,
+      "loss": 0.9651,
+      "step": 11469
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6881973620817246e-06,
+      "loss": 0.948,
+      "step": 11470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6867177680479684e-06,
+      "loss": 0.8996,
+      "step": 11471
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.685238518122403e-06,
+      "loss": 1.0163,
+      "step": 11472
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6837596123746423e-06,
+      "loss": 0.9544,
+      "step": 11473
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68228105087426e-06,
+      "loss": 0.8958,
+      "step": 11474
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6808028336908344e-06,
+      "loss": 1.0128,
+      "step": 11475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6793249608939177e-06,
+      "loss": 0.9806,
+      "step": 11476
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6778474325530457e-06,
+      "loss": 0.8836,
+      "step": 11477
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6763702487377407e-06,
+      "loss": 0.9054,
+      "step": 11478
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.674893409517505e-06,
+      "loss": 0.9561,
+      "step": 11479
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6734169149618337e-06,
+      "loss": 0.9055,
+      "step": 11480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6719407651401985e-06,
+      "loss": 0.9273,
+      "step": 11481
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.670464960122049e-06,
+      "loss": 0.9763,
+      "step": 11482
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.668989499976833e-06,
+      "loss": 0.8961,
+      "step": 11483
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6675143847739683e-06,
+      "loss": 0.8941,
+      "step": 11484
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6660396145828716e-06,
+      "loss": 1.0176,
+      "step": 11485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6645651894729274e-06,
+      "loss": 0.9568,
+      "step": 11486
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6630911095135093e-06,
+      "loss": 0.9783,
+      "step": 11487
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6616173747739838e-06,
+      "loss": 0.9381,
+      "step": 11488
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.660143985323689e-06,
+      "loss": 0.9214,
+      "step": 11489
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.658670941231953e-06,
+      "loss": 0.9286,
+      "step": 11490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6571982425680863e-06,
+      "loss": 0.8249,
+      "step": 11491
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6557258894013814e-06,
+      "loss": 0.9235,
+      "step": 11492
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6542538818011187e-06,
+      "loss": 0.93,
+      "step": 11493
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.652782219836555e-06,
+      "loss": 0.9068,
+      "step": 11494
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6513109035769423e-06,
+      "loss": 0.8963,
+      "step": 11495
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6498399330915057e-06,
+      "loss": 0.926,
+      "step": 11496
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.64836930844946e-06,
+      "loss": 0.9341,
+      "step": 11497
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6468990297199994e-06,
+      "loss": 0.8931,
+      "step": 11498
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6454290969723064e-06,
+      "loss": 0.9996,
+      "step": 11499
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6439595102755444e-06,
+      "loss": 0.9881,
+      "step": 11500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6424902696988596e-06,
+      "loss": 0.9309,
+      "step": 11501
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6410213753113824e-06,
+      "loss": 0.9068,
+      "step": 11502
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6395528271822326e-06,
+      "loss": 0.9223,
+      "step": 11503
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.638084625380507e-06,
+      "loss": 0.9925,
+      "step": 11504
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6366167699752878e-06,
+      "loss": 0.9525,
+      "step": 11505
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.635149261035641e-06,
+      "loss": 0.9166,
+      "step": 11506
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6336820986306143e-06,
+      "loss": 0.9546,
+      "step": 11507
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.632215282829249e-06,
+      "loss": 0.9063,
+      "step": 11508
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6307488137005555e-06,
+      "loss": 0.9597,
+      "step": 11509
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6292826913135326e-06,
+      "loss": 0.9259,
+      "step": 11510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6278169157371725e-06,
+      "loss": 0.9189,
+      "step": 11511
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6263514870404406e-06,
+      "loss": 0.9329,
+      "step": 11512
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6248864052922875e-06,
+      "loss": 0.8715,
+      "step": 11513
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6234216705616512e-06,
+      "loss": 0.9027,
+      "step": 11514
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6219572829174454e-06,
+      "loss": 0.9481,
+      "step": 11515
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6204932424285847e-06,
+      "loss": 0.945,
+      "step": 11516
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.619029549163943e-06,
+      "loss": 1.0068,
+      "step": 11517
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.617566203192399e-06,
+      "loss": 0.8766,
+      "step": 11518
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.616103204582805e-06,
+      "loss": 0.9375,
+      "step": 11519
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6146405534039974e-06,
+      "loss": 0.9695,
+      "step": 11520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6131782497247993e-06,
+      "loss": 0.9371,
+      "step": 11521
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6117162936140117e-06,
+      "loss": 0.8622,
+      "step": 11522
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.610254685140429e-06,
+      "loss": 0.9978,
+      "step": 11523
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.608793424372823e-06,
+      "loss": 0.9383,
+      "step": 11524
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.607332511379943e-06,
+      "loss": 0.9416,
+      "step": 11525
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.605871946230535e-06,
+      "loss": 0.8268,
+      "step": 11526
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6044117289933212e-06,
+      "loss": 1.0361,
+      "step": 11527
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6029518597370075e-06,
+      "loss": 0.8976,
+      "step": 11528
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6014923385302836e-06,
+      "loss": 0.9086,
+      "step": 11529
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.600033165441823e-06,
+      "loss": 0.8933,
+      "step": 11530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5985743405402906e-06,
+      "loss": 1.0238,
+      "step": 11531
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597115863894316e-06,
+      "loss": 0.9703,
+      "step": 11532
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5956577355725334e-06,
+      "loss": 0.9314,
+      "step": 11533
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594199955643548e-06,
+      "loss": 0.9722,
+      "step": 11534
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.592742524175952e-06,
+      "loss": 0.9384,
+      "step": 11535
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5912854412383226e-06,
+      "loss": 0.9602,
+      "step": 11536
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.589828706899217e-06,
+      "loss": 0.9779,
+      "step": 11537
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.588372321227177e-06,
+      "loss": 0.8794,
+      "step": 11538
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.586916284290737e-06,
+      "loss": 0.9057,
+      "step": 11539
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5854605961583954e-06,
+      "loss": 0.8672,
+      "step": 11540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5840052568986564e-06,
+      "loss": 0.9657,
+      "step": 11541
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5825502665799917e-06,
+      "loss": 0.9029,
+      "step": 11542
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5810956252708653e-06,
+      "loss": 0.9744,
+      "step": 11543
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5796413330397184e-06,
+      "loss": 0.8999,
+      "step": 11544
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5781873899549793e-06,
+      "loss": 0.8784,
+      "step": 11545
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5767337960850625e-06,
+      "loss": 0.8927,
+      "step": 11546
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.575280551498366e-06,
+      "loss": 0.9028,
+      "step": 11547
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.573827656263258e-06,
+      "loss": 0.9442,
+      "step": 11548
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57237511044811e-06,
+      "loss": 0.9422,
+      "step": 11549
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5709229141212656e-06,
+      "loss": 0.9713,
+      "step": 11550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5694710673510535e-06,
+      "loss": 0.856,
+      "step": 11551
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5680195702057874e-06,
+      "loss": 0.9376,
+      "step": 11552
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5665684227537604e-06,
+      "loss": 0.943,
+      "step": 11553
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565117625063259e-06,
+      "loss": 0.834,
+      "step": 11554
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5636671772025433e-06,
+      "loss": 0.9277,
+      "step": 11555
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5622170792398603e-06,
+      "loss": 0.9195,
+      "step": 11556
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5607673312434422e-06,
+      "loss": 0.9296,
+      "step": 11557
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.559317933281502e-06,
+      "loss": 0.9387,
+      "step": 11558
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5578688854222377e-06,
+      "loss": 0.9013,
+      "step": 11559
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.556420187733827e-06,
+      "loss": 0.8759,
+      "step": 11560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5549718402844426e-06,
+      "loss": 0.9794,
+      "step": 11561
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5535238431422282e-06,
+      "loss": 0.876,
+      "step": 11562
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552076196375316e-06,
+      "loss": 0.9268,
+      "step": 11563
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5506289000518215e-06,
+      "loss": 0.9774,
+      "step": 11564
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5491819542398413e-06,
+      "loss": 0.9346,
+      "step": 11565
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.547735359007466e-06,
+      "loss": 0.9255,
+      "step": 11566
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546289114422752e-06,
+      "loss": 0.7906,
+      "step": 11567
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5448432205537497e-06,
+      "loss": 0.9958,
+      "step": 11568
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5433976774684986e-06,
+      "loss": 0.9577,
+      "step": 11569
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.541952485235011e-06,
+      "loss": 0.9635,
+      "step": 11570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.540507643921287e-06,
+      "loss": 0.9575,
+      "step": 11571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.539063153595309e-06,
+      "loss": 0.9538,
+      "step": 11572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5376190143250436e-06,
+      "loss": 0.9078,
+      "step": 11573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536175226178448e-06,
+      "loss": 0.9924,
+      "step": 11574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.534731789223447e-06,
+      "loss": 0.8906,
+      "step": 11575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5332887035279583e-06,
+      "loss": 0.9316,
+      "step": 11576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5318459691598896e-06,
+      "loss": 0.9026,
+      "step": 11577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.530403586187121e-06,
+      "loss": 0.9239,
+      "step": 11578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5289615546775217e-06,
+      "loss": 0.9684,
+      "step": 11579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5275198746989405e-06,
+      "loss": 0.9287,
+      "step": 11580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5260785463192104e-06,
+      "loss": 0.8972,
+      "step": 11581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5246375696061585e-06,
+      "loss": 0.9341,
+      "step": 11582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523196944627575e-06,
+      "loss": 0.8858,
+      "step": 11583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521756671451253e-06,
+      "loss": 0.957,
+      "step": 11584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5203167501449576e-06,
+      "loss": 0.9713,
+      "step": 11585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5188771807764424e-06,
+      "loss": 0.9101,
+      "step": 11586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5174379634134406e-06,
+      "loss": 0.981,
+      "step": 11587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5159990981236694e-06,
+      "loss": 0.9276,
+      "step": 11588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.514560584974839e-06,
+      "loss": 0.9437,
+      "step": 11589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5131224240346273e-06,
+      "loss": 0.8863,
+      "step": 11590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.511684615370703e-06,
+      "loss": 0.9525,
+      "step": 11591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5102471590507226e-06,
+      "loss": 0.9501,
+      "step": 11592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.508810055142322e-06,
+      "loss": 0.9217,
+      "step": 11593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507373303713119e-06,
+      "loss": 0.9697,
+      "step": 11594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.505936904830717e-06,
+      "loss": 0.9717,
+      "step": 11595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5045008585626982e-06,
+      "loss": 0.9044,
+      "step": 11596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5030651649766413e-06,
+      "loss": 0.9817,
+      "step": 11597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5016298241400883e-06,
+      "loss": 0.9348,
+      "step": 11598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.500194836120583e-06,
+      "loss": 0.8995,
+      "step": 11599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4987602009856437e-06,
+      "loss": 0.9218,
+      "step": 11600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4973259188027734e-06,
+      "loss": 0.8874,
+      "step": 11601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4958919896394575e-06,
+      "loss": 0.8864,
+      "step": 11602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4944584135631632e-06,
+      "loss": 0.9593,
+      "step": 11603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4930251906413515e-06,
+      "loss": 0.9386,
+      "step": 11604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4915923209414573e-06,
+      "loss": 0.9222,
+      "step": 11605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4901598045308916e-06,
+      "loss": 0.9169,
+      "step": 11606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.488727641477069e-06,
+      "loss": 0.9919,
+      "step": 11607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4872958318473706e-06,
+      "loss": 0.9154,
+      "step": 11608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4858643757091684e-06,
+      "loss": 0.939,
+      "step": 11609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.484433273129816e-06,
+      "loss": 0.8709,
+      "step": 11610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4830025241766466e-06,
+      "loss": 0.9059,
+      "step": 11611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.481572128916987e-06,
+      "loss": 0.8551,
+      "step": 11612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4801420874181404e-06,
+      "loss": 0.9279,
+      "step": 11613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.478712399747385e-06,
+      "loss": 0.9454,
+      "step": 11614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4772830659720002e-06,
+      "loss": 0.9551,
+      "step": 11615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.475854086159237e-06,
+      "loss": 0.924,
+      "step": 11616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4744254603763317e-06,
+      "loss": 1.0114,
+      "step": 11617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4729971886905058e-06,
+      "loss": 0.971,
+      "step": 11618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.47156927116896e-06,
+      "loss": 0.9678,
+      "step": 11619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4701417078788857e-06,
+      "loss": 0.8651,
+      "step": 11620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.468714498887452e-06,
+      "loss": 1.0168,
+      "step": 11621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4672876442618134e-06,
+      "loss": 0.8917,
+      "step": 11622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.465861144069104e-06,
+      "loss": 0.9505,
+      "step": 11623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.464434998376447e-06,
+      "loss": 0.9244,
+      "step": 11624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.463009207250945e-06,
+      "loss": 0.9204,
+      "step": 11625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.461583770759681e-06,
+      "loss": 0.9564,
+      "step": 11626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.460158688969734e-06,
+      "loss": 0.9065,
+      "step": 11627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4587339619481544e-06,
+      "loss": 0.9904,
+      "step": 11628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4573095897619736e-06,
+      "loss": 0.9089,
+      "step": 11629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.455885572478218e-06,
+      "loss": 0.92,
+      "step": 11630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4544619101638865e-06,
+      "loss": 0.9306,
+      "step": 11631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453038602885974e-06,
+      "loss": 0.9575,
+      "step": 11632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4516156507114443e-06,
+      "loss": 0.942,
+      "step": 11633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4501930537072483e-06,
+      "loss": 0.9158,
+      "step": 11634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.448770811940329e-06,
+      "loss": 0.9781,
+      "step": 11635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.447348925477604e-06,
+      "loss": 0.9128,
+      "step": 11636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.445927394385976e-06,
+      "loss": 0.9063,
+      "step": 11637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.444506218732332e-06,
+      "loss": 0.961,
+      "step": 11638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44308539858354e-06,
+      "loss": 0.9513,
+      "step": 11639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4416649340064602e-06,
+      "loss": 0.917,
+      "step": 11640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.440244825067918e-06,
+      "loss": 0.9466,
+      "step": 11641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438825071834742e-06,
+      "loss": 0.9498,
+      "step": 11642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.437405674373732e-06,
+      "loss": 0.9467,
+      "step": 11643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4359866327516747e-06,
+      "loss": 0.9839,
+      "step": 11644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.434567947035339e-06,
+      "loss": 0.9481,
+      "step": 11645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4331496172914737e-06,
+      "loss": 0.8878,
+      "step": 11646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4317316435868253e-06,
+      "loss": 0.8502,
+      "step": 11647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4303140259881044e-06,
+      "loss": 1.0167,
+      "step": 11648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.428896764562012e-06,
+      "loss": 0.9054,
+      "step": 11649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42747985937524e-06,
+      "loss": 0.9367,
+      "step": 11650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.426063310494455e-06,
+      "loss": 0.9099,
+      "step": 11651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4246471179863086e-06,
+      "loss": 0.939,
+      "step": 11652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4232312819174364e-06,
+      "loss": 0.8902,
+      "step": 11653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4218158023544547e-06,
+      "loss": 0.9035,
+      "step": 11654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4204006793639724e-06,
+      "loss": 1.0046,
+      "step": 11655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.418985913012568e-06,
+      "loss": 0.8484,
+      "step": 11656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.417571503366808e-06,
+      "loss": 1.0065,
+      "step": 11657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4161574504932504e-06,
+      "loss": 0.8368,
+      "step": 11658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4147437544584274e-06,
+      "loss": 0.9301,
+      "step": 11659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4133304153288573e-06,
+      "loss": 0.9575,
+      "step": 11660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4119174331710394e-06,
+      "loss": 0.9334,
+      "step": 11661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.410504808051457e-06,
+      "loss": 0.9301,
+      "step": 11662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.409092540036585e-06,
+      "loss": 0.8628,
+      "step": 11663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4076806291928632e-06,
+      "loss": 0.9496,
+      "step": 11664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4062690755867345e-06,
+      "loss": 0.932,
+      "step": 11665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4048578792846135e-06,
+      "loss": 0.8987,
+      "step": 11666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4034470403528986e-06,
+      "loss": 0.9736,
+      "step": 11667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4020365588579752e-06,
+      "loss": 1.0513,
+      "step": 11668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4006264348662054e-06,
+      "loss": 1.0141,
+      "step": 11669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3992166684439466e-06,
+      "loss": 0.9291,
+      "step": 11670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.397807259657532e-06,
+      "loss": 0.9811,
+      "step": 11671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.396398208573267e-06,
+      "loss": 0.9658,
+      "step": 11672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.394989515257461e-06,
+      "loss": 0.9584,
+      "step": 11673
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3935811797763932e-06,
+      "loss": 0.9536,
+      "step": 11674
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3921732021963305e-06,
+      "loss": 0.9022,
+      "step": 11675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.390765582583521e-06,
+      "loss": 0.9371,
+      "step": 11676
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3893583210041927e-06,
+      "loss": 0.956,
+      "step": 11677
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3879514175245676e-06,
+      "loss": 0.9554,
+      "step": 11678
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.386544872210841e-06,
+      "loss": 0.8965,
+      "step": 11679
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385138685129195e-06,
+      "loss": 0.9177,
+      "step": 11680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.383732856345794e-06,
+      "loss": 0.9084,
+      "step": 11681
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3823273859267847e-06,
+      "loss": 0.9328,
+      "step": 11682
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380922273938299e-06,
+      "loss": 0.9355,
+      "step": 11683
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379517520446447e-06,
+      "loss": 0.9627,
+      "step": 11684
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.378113125517333e-06,
+      "loss": 0.9073,
+      "step": 11685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3767090892170364e-06,
+      "loss": 0.8993,
+      "step": 11686
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3753054116116126e-06,
+      "loss": 0.9458,
+      "step": 11687
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3739020927671165e-06,
+      "loss": 0.9117,
+      "step": 11688
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.372499132749575e-06,
+      "loss": 0.9488,
+      "step": 11689
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.371096531625e-06,
+      "loss": 0.8958,
+      "step": 11690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3696942894593878e-06,
+      "loss": 0.8784,
+      "step": 11691
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.368292406318716e-06,
+      "loss": 0.8441,
+      "step": 11692
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3668908822689507e-06,
+      "loss": 0.9262,
+      "step": 11693
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3654897173760372e-06,
+      "loss": 0.8383,
+      "step": 11694
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3640889117058953e-06,
+      "loss": 0.9181,
+      "step": 11695
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3626884653244465e-06,
+      "loss": 0.9353,
+      "step": 11696
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3612883782975806e-06,
+      "loss": 0.8904,
+      "step": 11697
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3598886506911757e-06,
+      "loss": 0.9651,
+      "step": 11698
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3584892825710927e-06,
+      "loss": 0.9363,
+      "step": 11699
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3570902740031733e-06,
+      "loss": 0.9103,
+      "step": 11700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355691625053248e-06,
+      "loss": 0.947,
+      "step": 11701
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3542933357871254e-06,
+      "loss": 0.9787,
+      "step": 11702
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3528954062705987e-06,
+      "loss": 0.9118,
+      "step": 11703
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.351497836569443e-06,
+      "loss": 0.9362,
+      "step": 11704
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3501006267494184e-06,
+      "loss": 0.942,
+      "step": 11705
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.348703776876267e-06,
+      "loss": 0.9412,
+      "step": 11706
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.347307287015709e-06,
+      "loss": 1.0138,
+      "step": 11707
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3459111572334613e-06,
+      "loss": 0.9301,
+      "step": 11708
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344515387595212e-06,
+      "loss": 0.9261,
+      "step": 11709
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3431199781666346e-06,
+      "loss": 0.9418,
+      "step": 11710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.341724929013387e-06,
+      "loss": 0.8822,
+      "step": 11711
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3403302402011062e-06,
+      "loss": 0.8704,
+      "step": 11712
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338935911795426e-06,
+      "loss": 1.0176,
+      "step": 11713
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3375419438619427e-06,
+      "loss": 0.8942,
+      "step": 11714
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3361483364662476e-06,
+      "loss": 0.9042,
+      "step": 11715
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.334755089673917e-06,
+      "loss": 0.8583,
+      "step": 11716
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3333622035505064e-06,
+      "loss": 0.9372,
+      "step": 11717
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.331969678161552e-06,
+      "loss": 0.8699,
+      "step": 11718
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3305775135725784e-06,
+      "loss": 0.8734,
+      "step": 11719
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3291857098490843e-06,
+      "loss": 1.0046,
+      "step": 11720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.327794267056569e-06,
+      "loss": 0.864,
+      "step": 11721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.326403185260491e-06,
+      "loss": 0.9519,
+      "step": 11722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3250124645263108e-06,
+      "loss": 0.9606,
+      "step": 11723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.323622104919465e-06,
+      "loss": 0.9201,
+      "step": 11724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.322232106505373e-06,
+      "loss": 0.9277,
+      "step": 11725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3208424693494368e-06,
+      "loss": 0.8732,
+      "step": 11726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.319453193517044e-06,
+      "loss": 0.971,
+      "step": 11727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3180642790735576e-06,
+      "loss": 0.9692,
+      "step": 11728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3166757260843408e-06,
+      "loss": 0.9665,
+      "step": 11729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3152875346147164e-06,
+      "loss": 0.8298,
+      "step": 11730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3138997047300104e-06,
+      "loss": 0.891,
+      "step": 11731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.312512236495521e-06,
+      "loss": 0.9261,
+      "step": 11732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3111251299765326e-06,
+      "loss": 0.996,
+      "step": 11733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3097383852383105e-06,
+      "loss": 1.0371,
+      "step": 11734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3083520023461035e-06,
+      "loss": 0.8892,
+      "step": 11735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.306965981365149e-06,
+      "loss": 0.9281,
+      "step": 11736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.305580322360662e-06,
+      "loss": 0.9123,
+      "step": 11737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3041950253978352e-06,
+      "loss": 0.975,
+      "step": 11738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.302810090541856e-06,
+      "loss": 0.916,
+      "step": 11739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3014255178578883e-06,
+      "loss": 0.9437,
+      "step": 11740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3000413074110782e-06,
+      "loss": 0.8773,
+      "step": 11741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.298657459266558e-06,
+      "loss": 0.9512,
+      "step": 11742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2972739734894355e-06,
+      "loss": 0.9358,
+      "step": 11743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2958908501448185e-06,
+      "loss": 0.9031,
+      "step": 11744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2945080892977747e-06,
+      "loss": 0.9296,
+      "step": 11745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2931256910133736e-06,
+      "loss": 0.9028,
+      "step": 11746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2917436553566586e-06,
+      "loss": 1.0604,
+      "step": 11747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.290361982392657e-06,
+      "loss": 0.9569,
+      "step": 11748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2889806721863818e-06,
+      "loss": 0.961,
+      "step": 11749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2875997248028214e-06,
+      "loss": 0.8825,
+      "step": 11750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286219140306961e-06,
+      "loss": 0.9365,
+      "step": 11751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2848389187637608e-06,
+      "loss": 0.9594,
+      "step": 11752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2834590602381534e-06,
+      "loss": 0.9031,
+      "step": 11753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.282079564795073e-06,
+      "loss": 0.9603,
+      "step": 11754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2807004324994275e-06,
+      "loss": 0.922,
+      "step": 11755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2793216634161065e-06,
+      "loss": 0.9467,
+      "step": 11756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2779432576099856e-06,
+      "loss": 0.9003,
+      "step": 11757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2765652151459193e-06,
+      "loss": 0.9126,
+      "step": 11758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2751875360887533e-06,
+      "loss": 0.9337,
+      "step": 11759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2738102205033086e-06,
+      "loss": 0.9464,
+      "step": 11760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2724332684543904e-06,
+      "loss": 0.8991,
+      "step": 11761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.271056680006788e-06,
+      "loss": 0.9346,
+      "step": 11762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.269680455225274e-06,
+      "loss": 0.9561,
+      "step": 11763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2683045941746018e-06,
+      "loss": 0.9662,
+      "step": 11764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2669290969195114e-06,
+      "loss": 0.9756,
+      "step": 11765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265553963524718e-06,
+      "loss": 0.9964,
+      "step": 11766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.264179194054933e-06,
+      "loss": 0.9263,
+      "step": 11767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2628047885748382e-06,
+      "loss": 0.9521,
+      "step": 11768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2614307471491035e-06,
+      "loss": 0.8832,
+      "step": 11769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.260057069842381e-06,
+      "loss": 0.8799,
+      "step": 11770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2586837567193055e-06,
+      "loss": 0.8976,
+      "step": 11771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257310807844495e-06,
+      "loss": 0.9568,
+      "step": 11772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2559382232825466e-06,
+      "loss": 0.8437,
+      "step": 11773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.254566003098051e-06,
+      "loss": 1.0208,
+      "step": 11774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.253194147355571e-06,
+      "loss": 0.8668,
+      "step": 11775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2518226561196555e-06,
+      "loss": 0.9935,
+      "step": 11776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2504515294548356e-06,
+      "loss": 0.9447,
+      "step": 11777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2490807674256264e-06,
+      "loss": 0.929,
+      "step": 11778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2477103700965306e-06,
+      "loss": 0.9304,
+      "step": 11779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2463403375320236e-06,
+      "loss": 0.9308,
+      "step": 11780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2449706697965666e-06,
+      "loss": 0.9055,
+      "step": 11781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.243601366954613e-06,
+      "loss": 0.9313,
+      "step": 11782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2422324290705876e-06,
+      "loss": 0.8911,
+      "step": 11783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2408638562089035e-06,
+      "loss": 0.8925,
+      "step": 11784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239495648433956e-06,
+      "loss": 0.9191,
+      "step": 11785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2381278058101173e-06,
+      "loss": 0.9893,
+      "step": 11786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.236760328401759e-06,
+      "loss": 0.9093,
+      "step": 11787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2353932162732115e-06,
+      "loss": 0.9567,
+      "step": 11788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2340264694888103e-06,
+      "loss": 0.927,
+      "step": 11789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.232660088112861e-06,
+      "loss": 0.9264,
+      "step": 11790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2312940722096544e-06,
+      "loss": 0.8817,
+      "step": 11791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.229928421843467e-06,
+      "loss": 0.9158,
+      "step": 11792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.228563137078551e-06,
+      "loss": 0.9713,
+      "step": 11793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2271982179791564e-06,
+      "loss": 0.9106,
+      "step": 11794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.225833664609497e-06,
+      "loss": 0.8865,
+      "step": 11795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.224469477033778e-06,
+      "loss": 0.8823,
+      "step": 11796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2231056553161947e-06,
+      "loss": 0.9605,
+      "step": 11797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2217421995209144e-06,
+      "loss": 0.9753,
+      "step": 11798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.220379109712092e-06,
+      "loss": 0.9427,
+      "step": 11799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2190163859538627e-06,
+      "loss": 0.9119,
+      "step": 11800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2176540283103453e-06,
+      "loss": 0.925,
+      "step": 11801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.21629203684565e-06,
+      "loss": 0.8878,
+      "step": 11802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2149304116238525e-06,
+      "loss": 0.8657,
+      "step": 11803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.213569152709021e-06,
+      "loss": 0.8976,
+      "step": 11804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2122082601652117e-06,
+      "loss": 0.9535,
+      "step": 11805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2108477340564562e-06,
+      "loss": 0.9759,
+      "step": 11806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2094875744467703e-06,
+      "loss": 0.9675,
+      "step": 11807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.208127781400151e-06,
+      "loss": 0.9367,
+      "step": 11808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2067683549805808e-06,
+      "loss": 0.954,
+      "step": 11809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2054092952520288e-06,
+      "loss": 0.9119,
+      "step": 11810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2040506022784337e-06,
+      "loss": 0.903,
+      "step": 11811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2026922761237312e-06,
+      "loss": 0.9655,
+      "step": 11812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.201334316851833e-06,
+      "loss": 0.9318,
+      "step": 11813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1999767245266346e-06,
+      "loss": 0.9265,
+      "step": 11814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1986194992120137e-06,
+      "loss": 0.9826,
+      "step": 11815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1972626409718277e-06,
+      "loss": 0.8697,
+      "step": 11816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1959061498699276e-06,
+      "loss": 0.8789,
+      "step": 11817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.194550025970137e-06,
+      "loss": 0.9087,
+      "step": 11818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1931942693362584e-06,
+      "loss": 0.8765,
+      "step": 11819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.191838880032091e-06,
+      "loss": 0.9289,
+      "step": 11820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190483858121407e-06,
+      "loss": 0.9437,
+      "step": 11821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1891292036679643e-06,
+      "loss": 0.9731,
+      "step": 11822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1877749167355013e-06,
+      "loss": 0.9446,
+      "step": 11823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.186420997387738e-06,
+      "loss": 0.9641,
+      "step": 11824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.185067445688387e-06,
+      "loss": 0.9879,
+      "step": 11825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183714261701131e-06,
+      "loss": 0.9056,
+      "step": 11826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1823614454896425e-06,
+      "loss": 0.891,
+      "step": 11827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181008997117574e-06,
+      "loss": 0.9244,
+      "step": 11828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1796569166485613e-06,
+      "loss": 0.9513,
+      "step": 11829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178305204146224e-06,
+      "loss": 0.9247,
+      "step": 11830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1769538596741612e-06,
+      "loss": 0.8461,
+      "step": 11831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175602883295961e-06,
+      "loss": 0.923,
+      "step": 11832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1742522750751893e-06,
+      "loss": 0.9086,
+      "step": 11833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.172902035075394e-06,
+      "loss": 0.9203,
+      "step": 11834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1715521633601066e-06,
+      "loss": 0.9532,
+      "step": 11835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170202659992845e-06,
+      "loss": 0.8776,
+      "step": 11836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1688535250371047e-06,
+      "loss": 0.9768,
+      "step": 11837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.167504758556365e-06,
+      "loss": 0.8856,
+      "step": 11838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1661563606140877e-06,
+      "loss": 0.8689,
+      "step": 11839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.164808331273723e-06,
+      "loss": 0.9164,
+      "step": 11840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1634606705986994e-06,
+      "loss": 0.8952,
+      "step": 11841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1621133786524183e-06,
+      "loss": 0.9111,
+      "step": 11842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1607664554982833e-06,
+      "loss": 0.8987,
+      "step": 11843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.159419901199664e-06,
+      "loss": 0.9214,
+      "step": 11844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1580737158199272e-06,
+      "loss": 0.9323,
+      "step": 11845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156727899422406e-06,
+      "loss": 0.938,
+      "step": 11846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1553824520704257e-06,
+      "loss": 0.883,
+      "step": 11847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1540373738272966e-06,
+      "loss": 0.9227,
+      "step": 11848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152692664756306e-06,
+      "loss": 0.9485,
+      "step": 11849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.151348324920727e-06,
+      "loss": 0.9163,
+      "step": 11850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150004354383811e-06,
+      "loss": 0.8792,
+      "step": 11851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148660753208799e-06,
+      "loss": 1.0413,
+      "step": 11852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.147317521458908e-06,
+      "loss": 0.9142,
+      "step": 11853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1459746591973396e-06,
+      "loss": 0.9235,
+      "step": 11854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1446321664872826e-06,
+      "loss": 0.9459,
+      "step": 11855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1432900433919025e-06,
+      "loss": 0.9057,
+      "step": 11856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1419482899743504e-06,
+      "loss": 0.9271,
+      "step": 11857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1406069062977577e-06,
+      "loss": 0.9175,
+      "step": 11858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1392658924252384e-06,
+      "loss": 0.9336,
+      "step": 11859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.137925248419899e-06,
+      "loss": 0.9152,
+      "step": 11860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1365849743448097e-06,
+      "loss": 0.9193,
+      "step": 11861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1352450702630356e-06,
+      "loss": 0.937,
+      "step": 11862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133905536237628e-06,
+      "loss": 0.9153,
+      "step": 11863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1325663723316127e-06,
+      "loss": 1.0339,
+      "step": 11864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1312275786079993e-06,
+      "loss": 0.9738,
+      "step": 11865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.129889155129783e-06,
+      "loss": 0.8729,
+      "step": 11866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.128551101959936e-06,
+      "loss": 1.0181,
+      "step": 11867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1272134191614256e-06,
+      "loss": 0.9639,
+      "step": 11868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1258761067971836e-06,
+      "loss": 0.8926,
+      "step": 11869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1245391649301404e-06,
+      "loss": 0.9141,
+      "step": 11870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1232025936231992e-06,
+      "loss": 0.9316,
+      "step": 11871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.121866392939251e-06,
+      "loss": 0.9842,
+      "step": 11872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1205305629411667e-06,
+      "loss": 0.9277,
+      "step": 11873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.119195103691799e-06,
+      "loss": 0.878,
+      "step": 11874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.117860015253983e-06,
+      "loss": 0.9311,
+      "step": 11875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.116525297690547e-06,
+      "loss": 0.9587,
+      "step": 11876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.115190951064281e-06,
+      "loss": 0.9262,
+      "step": 11877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1138569754379767e-06,
+      "loss": 0.9304,
+      "step": 11878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1125233708743997e-06,
+      "loss": 0.9374,
+      "step": 11879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1111901374362985e-06,
+      "loss": 0.9198,
+      "step": 11880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1098572751864053e-06,
+      "loss": 0.9267,
+      "step": 11881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1085247841874322e-06,
+      "loss": 0.9478,
+      "step": 11882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1071926645020813e-06,
+      "loss": 0.833,
+      "step": 11883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.105860916193032e-06,
+      "loss": 0.948,
+      "step": 11884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1045295393229394e-06,
+      "loss": 0.9628,
+      "step": 11885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1031985339544546e-06,
+      "loss": 0.9481,
+      "step": 11886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.101867900150203e-06,
+      "loss": 0.918,
+      "step": 11887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1005376379727937e-06,
+      "loss": 0.9184,
+      "step": 11888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.099207747484819e-06,
+      "loss": 0.9641,
+      "step": 11889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0978782287488517e-06,
+      "loss": 0.9275,
+      "step": 11890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.096549081827457e-06,
+      "loss": 0.9834,
+      "step": 11891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095220306783162e-06,
+      "loss": 1.0178,
+      "step": 11892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0938919036784987e-06,
+      "loss": 0.8529,
+      "step": 11893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092563872575968e-06,
+      "loss": 0.9449,
+      "step": 11894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.091236213538058e-06,
+      "loss": 0.928,
+      "step": 11895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0899089266272386e-06,
+      "loss": 0.9018,
+      "step": 11896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0885820119059584e-06,
+      "loss": 0.9248,
+      "step": 11897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0872554694366574e-06,
+      "loss": 0.9557,
+      "step": 11898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0859292992817535e-06,
+      "loss": 1.0131,
+      "step": 11899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.084603501503638e-06,
+      "loss": 0.9152,
+      "step": 11900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0832780761646997e-06,
+      "loss": 0.896,
+      "step": 11901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0819530233273e-06,
+      "loss": 1.0241,
+      "step": 11902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0806283430537934e-06,
+      "loss": 0.9115,
+      "step": 11903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0793040354064997e-06,
+      "loss": 0.9642,
+      "step": 11904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.077980100447733e-06,
+      "loss": 0.9579,
+      "step": 11905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0766565382397917e-06,
+      "loss": 0.9808,
+      "step": 11906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075333348844951e-06,
+      "loss": 0.8953,
+      "step": 11907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07401053232547e-06,
+      "loss": 0.8922,
+      "step": 11908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0726880887435895e-06,
+      "loss": 0.8757,
+      "step": 11909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.071366018161536e-06,
+      "loss": 0.9353,
+      "step": 11910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0700443206415135e-06,
+      "loss": 0.9156,
+      "step": 11911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0687229962457134e-06,
+      "loss": 0.8725,
+      "step": 11912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.067402045036303e-06,
+      "loss": 0.8676,
+      "step": 11913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0660814670754425e-06,
+      "loss": 0.8942,
+      "step": 11914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0647612624252664e-06,
+      "loss": 0.9441,
+      "step": 11915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0634414311478924e-06,
+      "loss": 0.8981,
+      "step": 11916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.062121973305422e-06,
+      "loss": 1.0192,
+      "step": 11917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.060802888959941e-06,
+      "loss": 0.8898,
+      "step": 11918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.059484178173512e-06,
+      "loss": 0.8747,
+      "step": 11919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.058165841008184e-06,
+      "loss": 0.8504,
+      "step": 11920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0568478775259925e-06,
+      "loss": 0.9239,
+      "step": 11921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.055530287788947e-06,
+      "loss": 0.9811,
+      "step": 11922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.054213071859046e-06,
+      "loss": 0.9542,
+      "step": 11923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0528962297982647e-06,
+      "loss": 0.9943,
+      "step": 11924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051579761668564e-06,
+      "loss": 0.956,
+      "step": 11925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050263667531893e-06,
+      "loss": 0.8469,
+      "step": 11926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0489479474501704e-06,
+      "loss": 0.8331,
+      "step": 11927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0476326014853042e-06,
+      "loss": 0.8646,
+      "step": 11928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0463176296991894e-06,
+      "loss": 0.8608,
+      "step": 11929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0450030321536953e-06,
+      "loss": 0.9179,
+      "step": 11930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0436888089106787e-06,
+      "loss": 0.9166,
+      "step": 11931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0423749600319754e-06,
+      "loss": 0.9529,
+      "step": 11932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.041061485579404e-06,
+      "loss": 0.9603,
+      "step": 11933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.039748385614775e-06,
+      "loss": 0.8664,
+      "step": 11934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0384356601998612e-06,
+      "loss": 0.8587,
+      "step": 11935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.037123309396437e-06,
+      "loss": 0.9234,
+      "step": 11936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035811333266251e-06,
+      "loss": 0.8871,
+      "step": 11937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0344997318710334e-06,
+      "loss": 0.9666,
+      "step": 11938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0331885052725e-06,
+      "loss": 0.9586,
+      "step": 11939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031877653532344e-06,
+      "loss": 0.9334,
+      "step": 11940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.030567176712248e-06,
+      "loss": 0.9617,
+      "step": 11941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0292570748738747e-06,
+      "loss": 1.0031,
+      "step": 11942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.027947348078859e-06,
+      "loss": 0.8892,
+      "step": 11943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.026637996388836e-06,
+      "loss": 0.9155,
+      "step": 11944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0253290198654096e-06,
+      "loss": 0.9742,
+      "step": 11945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0240204185701728e-06,
+      "loss": 0.9804,
+      "step": 11946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.022712192564695e-06,
+      "loss": 0.9055,
+      "step": 11947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0214043419105323e-06,
+      "loss": 0.9787,
+      "step": 11948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020096866669229e-06,
+      "loss": 0.9313,
+      "step": 11949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.018789766902297e-06,
+      "loss": 0.9247,
+      "step": 11950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.017483042671238e-06,
+      "loss": 0.8569,
+      "step": 11951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016176694037543e-06,
+      "loss": 0.9464,
+      "step": 11952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.014870721062676e-06,
+      "loss": 0.974,
+      "step": 11953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0135651238080854e-06,
+      "loss": 0.8927,
+      "step": 11954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.012259902335204e-06,
+      "loss": 0.9094,
+      "step": 11955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0109550567054437e-06,
+      "loss": 0.883,
+      "step": 11956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009650586980206e-06,
+      "loss": 0.9293,
+      "step": 11957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.008346493220862e-06,
+      "loss": 0.8672,
+      "step": 11958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.007042775488779e-06,
+      "loss": 0.9505,
+      "step": 11959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0057394338452964e-06,
+      "loss": 0.9774,
+      "step": 11960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0044364683517414e-06,
+      "loss": 0.9788,
+      "step": 11961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0031338790694212e-06,
+      "loss": 0.9424,
+      "step": 11962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001831666059624e-06,
+      "loss": 0.8928,
+      "step": 11963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.000529829383626e-06,
+      "loss": 0.9549,
+      "step": 11964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9992283691026826e-06,
+      "loss": 0.8995,
+      "step": 11965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.997927285278023e-06,
+      "loss": 0.926,
+      "step": 11966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9966265779708736e-06,
+      "loss": 0.9353,
+      "step": 11967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.995326247242435e-06,
+      "loss": 0.9095,
+      "step": 11968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.99402629315389e-06,
+      "loss": 0.9371,
+      "step": 11969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.992726715766403e-06,
+      "loss": 0.8965,
+      "step": 11970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9914275151411234e-06,
+      "loss": 0.8854,
+      "step": 11971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9901286913391836e-06,
+      "loss": 0.9157,
+      "step": 11972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9888302444216957e-06,
+      "loss": 0.9757,
+      "step": 11973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9875321744497542e-06,
+      "loss": 0.8933,
+      "step": 11974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.986234481484437e-06,
+      "loss": 1.0388,
+      "step": 11975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.984937165586803e-06,
+      "loss": 0.9331,
+      "step": 11976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9836402268178954e-06,
+      "loss": 0.9553,
+      "step": 11977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982343665238735e-06,
+      "loss": 0.9388,
+      "step": 11978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9810474809103343e-06,
+      "loss": 0.9632,
+      "step": 11979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9797516738936774e-06,
+      "loss": 0.9189,
+      "step": 11980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.978456244249737e-06,
+      "loss": 0.8552,
+      "step": 11981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9771611920394664e-06,
+      "loss": 0.8699,
+      "step": 11982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975866517323798e-06,
+      "loss": 0.9126,
+      "step": 11983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9745722201636562e-06,
+      "loss": 0.9028,
+      "step": 11984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9732783006199352e-06,
+      "loss": 0.8597,
+      "step": 11985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9719847587535158e-06,
+      "loss": 0.871,
+      "step": 11986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.970691594625267e-06,
+      "loss": 0.9744,
+      "step": 11987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969398808296037e-06,
+      "loss": 0.9714,
+      "step": 11988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968106399826646e-06,
+      "loss": 0.9046,
+      "step": 11989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.966814369277913e-06,
+      "loss": 0.887,
+      "step": 11990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9655227167106262e-06,
+      "loss": 0.8569,
+      "step": 11991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.964231442185568e-06,
+      "loss": 0.8873,
+      "step": 11992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9629405457634887e-06,
+      "loss": 0.9714,
+      "step": 11993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.961650027505129e-06,
+      "loss": 0.9513,
+      "step": 11994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.960359887471216e-06,
+      "loss": 0.9321,
+      "step": 11995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9590701257224496e-06,
+      "loss": 1.0076,
+      "step": 11996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9577807423195194e-06,
+      "loss": 0.92,
+      "step": 11997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9564917373230916e-06,
+      "loss": 0.8532,
+      "step": 11998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9552031107938164e-06,
+      "loss": 0.9976,
+      "step": 11999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9539148627923332e-06,
+      "loss": 0.9187,
+      "step": 12000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9526269933792464e-06,
+      "loss": 0.8988,
+      "step": 12001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951339502615164e-06,
+      "loss": 0.9067,
+      "step": 12002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9500523905606592e-06,
+      "loss": 0.8799,
+      "step": 12003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9487656572762968e-06,
+      "loss": 0.8964,
+      "step": 12004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947479302822621e-06,
+      "loss": 0.9146,
+      "step": 12005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946193327260153e-06,
+      "loss": 0.8354,
+      "step": 12006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9449077306494114e-06,
+      "loss": 0.9686,
+      "step": 12007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9436225130508768e-06,
+      "loss": 0.8823,
+      "step": 12008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9423376745250233e-06,
+      "loss": 0.9059,
+      "step": 12009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9410532151323114e-06,
+      "loss": 0.9803,
+      "step": 12010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9397691349331738e-06,
+      "loss": 0.8839,
+      "step": 12011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.938485433988031e-06,
+      "loss": 0.9609,
+      "step": 12012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937202112357284e-06,
+      "loss": 0.9422,
+      "step": 12013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9359191701013126e-06,
+      "loss": 0.9063,
+      "step": 12014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9346366072804924e-06,
+      "loss": 0.8851,
+      "step": 12015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933354423955159e-06,
+      "loss": 0.931,
+      "step": 12016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.932072620185651e-06,
+      "loss": 0.9431,
+      "step": 12017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930791196032278e-06,
+      "loss": 0.9514,
+      "step": 12018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929510151555334e-06,
+      "loss": 1.0151,
+      "step": 12019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9282294868150943e-06,
+      "loss": 0.929,
+      "step": 12020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.926949201871816e-06,
+      "loss": 0.9684,
+      "step": 12021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925669296785745e-06,
+      "loss": 0.9984,
+      "step": 12022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9243897716171033e-06,
+      "loss": 0.951,
+      "step": 12023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.923110626426088e-06,
+      "loss": 0.9502,
+      "step": 12024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9218318612728948e-06,
+      "loss": 0.8947,
+      "step": 12025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9205534762176893e-06,
+      "loss": 0.9583,
+      "step": 12026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.919275471320622e-06,
+      "loss": 0.9555,
+      "step": 12027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.917997846641828e-06,
+      "loss": 0.9468,
+      "step": 12028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.91672060224142e-06,
+      "loss": 1.0145,
+      "step": 12029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.915443738179499e-06,
+      "loss": 0.886,
+      "step": 12030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.914167254516146e-06,
+      "loss": 0.903,
+      "step": 12031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9128911513114148e-06,
+      "loss": 0.8546,
+      "step": 12032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9116154286253565e-06,
+      "loss": 0.949,
+      "step": 12033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.910340086517994e-06,
+      "loss": 0.9612,
+      "step": 12034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9090651250493376e-06,
+      "loss": 0.9344,
+      "step": 12035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9077905442793754e-06,
+      "loss": 0.8879,
+      "step": 12036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9065163442680767e-06,
+      "loss": 0.8805,
+      "step": 12037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9052425250754037e-06,
+      "loss": 0.8735,
+      "step": 12038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.903969086761287e-06,
+      "loss": 0.8777,
+      "step": 12039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9026960293856466e-06,
+      "loss": 0.8257,
+      "step": 12040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9014233530083825e-06,
+      "loss": 0.8929,
+      "step": 12041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.900151057689379e-06,
+      "loss": 0.9436,
+      "step": 12042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8988791434884979e-06,
+      "loss": 0.8752,
+      "step": 12043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.897607610465585e-06,
+      "loss": 0.9272,
+      "step": 12044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8963364586804744e-06,
+      "loss": 0.91,
+      "step": 12045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8950656881929775e-06,
+      "loss": 0.956,
+      "step": 12046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8937952990628771e-06,
+      "loss": 0.9423,
+      "step": 12047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8925252913499591e-06,
+      "loss": 0.9446,
+      "step": 12048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8912556651139723e-06,
+      "loss": 0.8841,
+      "step": 12049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8899864204146656e-06,
+      "loss": 0.89,
+      "step": 12050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8887175573117522e-06,
+      "loss": 0.9618,
+      "step": 12051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8874490758649345e-06,
+      "loss": 0.991,
+      "step": 12052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8861809761339035e-06,
+      "loss": 0.9015,
+      "step": 12053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8849132581783237e-06,
+      "loss": 0.9811,
+      "step": 12054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.883645922057844e-06,
+      "loss": 0.9394,
+      "step": 12055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8823789678320957e-06,
+      "loss": 0.8982,
+      "step": 12056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8811123955606903e-06,
+      "loss": 0.9009,
+      "step": 12057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8798462053032308e-06,
+      "loss": 0.9149,
+      "step": 12058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8785803971192839e-06,
+      "loss": 0.9096,
+      "step": 12059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.877314971068417e-06,
+      "loss": 0.9321,
+      "step": 12060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8760499272101684e-06,
+      "loss": 0.9203,
+      "step": 12061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8747852656040622e-06,
+      "loss": 0.9056,
+      "step": 12062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8735209863096037e-06,
+      "loss": 0.9738,
+      "step": 12063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8722570893862813e-06,
+      "loss": 0.9032,
+      "step": 12064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8709935748935625e-06,
+      "loss": 0.9651,
+      "step": 12065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8697304428909003e-06,
+      "loss": 0.9312,
+      "step": 12066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8684676934377254e-06,
+      "loss": 0.9915,
+      "step": 12067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8672053265934587e-06,
+      "loss": 0.9308,
+      "step": 12068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8659433424174954e-06,
+      "loss": 0.8739,
+      "step": 12069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8646817409692132e-06,
+      "loss": 0.8541,
+      "step": 12070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.863420522307976e-06,
+      "loss": 0.9174,
+      "step": 12071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8621596864931225e-06,
+      "loss": 0.874,
+      "step": 12072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.860899233583987e-06,
+      "loss": 0.9287,
+      "step": 12073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8596391636398703e-06,
+      "loss": 0.936,
+      "step": 12074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.858379476720059e-06,
+      "loss": 0.9276,
+      "step": 12075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8571201728838328e-06,
+      "loss": 0.9463,
+      "step": 12076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8558612521904407e-06,
+      "loss": 0.9067,
+      "step": 12077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8546027146991186e-06,
+      "loss": 0.9527,
+      "step": 12078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8533445604690836e-06,
+      "loss": 0.8904,
+      "step": 12079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8520867895595318e-06,
+      "loss": 0.9171,
+      "step": 12080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8508294020296536e-06,
+      "loss": 0.9514,
+      "step": 12081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8495723979386016e-06,
+      "loss": 0.959,
+      "step": 12082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8483157773455273e-06,
+      "loss": 0.9455,
+      "step": 12083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8470595403095571e-06,
+      "loss": 0.9946,
+      "step": 12084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8458036868897988e-06,
+      "loss": 0.8538,
+      "step": 12085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8445482171453445e-06,
+      "loss": 0.8597,
+      "step": 12086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.843293131135263e-06,
+      "loss": 0.8919,
+      "step": 12087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8420384289186167e-06,
+      "loss": 0.9411,
+      "step": 12088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.840784110554441e-06,
+      "loss": 0.9445,
+      "step": 12089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.839530176101747e-06,
+      "loss": 0.9926,
+      "step": 12090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8382766256195428e-06,
+      "loss": 0.8335,
+      "step": 12091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.837023459166809e-06,
+      "loss": 0.8439,
+      "step": 12092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8357706768025108e-06,
+      "loss": 0.8991,
+      "step": 12093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.834518278585594e-06,
+      "loss": 0.951,
+      "step": 12094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8332662645749856e-06,
+      "loss": 0.9251,
+      "step": 12095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8320146348295997e-06,
+      "loss": 0.9948,
+      "step": 12096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8307633894083253e-06,
+      "loss": 0.9022,
+      "step": 12097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8295125283700398e-06,
+      "loss": 0.888,
+      "step": 12098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.828262051773596e-06,
+      "loss": 0.9872,
+      "step": 12099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8270119596778346e-06,
+      "loss": 0.9971,
+      "step": 12100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.825762252141573e-06,
+      "loss": 0.9254,
+      "step": 12101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8245129292236153e-06,
+      "loss": 0.9247,
+      "step": 12102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.823263990982742e-06,
+      "loss": 0.9025,
+      "step": 12103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8220154374777254e-06,
+      "loss": 0.829,
+      "step": 12104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.820767268767304e-06,
+      "loss": 0.8385,
+      "step": 12105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8195194849102149e-06,
+      "loss": 0.9681,
+      "step": 12106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.818272085965166e-06,
+      "loss": 0.9186,
+      "step": 12107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8170250719908523e-06,
+      "loss": 1.0317,
+      "step": 12108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8157784430459468e-06,
+      "loss": 0.9472,
+      "step": 12109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.814532199189104e-06,
+      "loss": 0.9463,
+      "step": 12110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8132863404789703e-06,
+      "loss": 0.929,
+      "step": 12111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.812040866974164e-06,
+      "loss": 0.9629,
+      "step": 12112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.810795778733282e-06,
+      "loss": 0.9079,
+      "step": 12113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8095510758149148e-06,
+      "loss": 0.9941,
+      "step": 12114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.808306758277627e-06,
+      "loss": 0.8826,
+      "step": 12115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8070628261799673e-06,
+      "loss": 0.9731,
+      "step": 12116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8058192795804652e-06,
+      "loss": 0.9457,
+      "step": 12117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8045761185376298e-06,
+      "loss": 0.8846,
+      "step": 12118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8033333431099598e-06,
+      "loss": 0.9135,
+      "step": 12119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8020909533559306e-06,
+      "loss": 0.8863,
+      "step": 12120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8008489493339965e-06,
+      "loss": 0.9448,
+      "step": 12121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7996073311025997e-06,
+      "loss": 0.9789,
+      "step": 12122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.798366098720159e-06,
+      "loss": 0.9209,
+      "step": 12123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.79712525224508e-06,
+      "loss": 0.9564,
+      "step": 12124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7958847917357436e-06,
+      "loss": 0.8783,
+      "step": 12125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.794644717250521e-06,
+      "loss": 0.9625,
+      "step": 12126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.79340502884776e-06,
+      "loss": 0.8846,
+      "step": 12127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7921657265857906e-06,
+      "loss": 0.8846,
+      "step": 12128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7909268105229238e-06,
+      "loss": 0.8282,
+      "step": 12129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.789688280717452e-06,
+      "loss": 0.9502,
+      "step": 12130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7884501372276585e-06,
+      "loss": 0.8978,
+      "step": 12131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7872123801117946e-06,
+      "loss": 1.0382,
+      "step": 12132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.785975009428098e-06,
+      "loss": 0.9289,
+      "step": 12133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7847380252347967e-06,
+      "loss": 0.9848,
+      "step": 12134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7835014275900907e-06,
+      "loss": 0.9115,
+      "step": 12135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7822652165521648e-06,
+      "loss": 0.8857,
+      "step": 12136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7810293921791855e-06,
+      "loss": 1.0208,
+      "step": 12137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.779793954529301e-06,
+      "loss": 0.8659,
+      "step": 12138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7785589036606466e-06,
+      "loss": 0.9373,
+      "step": 12139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7773242396313284e-06,
+      "loss": 0.9312,
+      "step": 12140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7760899624994399e-06,
+      "loss": 0.911,
+      "step": 12141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7748560723230624e-06,
+      "loss": 0.8805,
+      "step": 12142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7736225691602505e-06,
+      "loss": 0.987,
+      "step": 12143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7723894530690443e-06,
+      "loss": 0.9009,
+      "step": 12144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7711567241074646e-06,
+      "loss": 0.9299,
+      "step": 12145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.769924382333511e-06,
+      "loss": 0.9162,
+      "step": 12146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7686924278051775e-06,
+      "loss": 0.9475,
+      "step": 12147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.76746086058042e-06,
+      "loss": 0.9444,
+      "step": 12148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7662296807171942e-06,
+      "loss": 0.918,
+      "step": 12149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.764998888273427e-06,
+      "loss": 0.9731,
+      "step": 12150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.76376848330703e-06,
+      "loss": 0.9842,
+      "step": 12151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7625384658758993e-06,
+      "loss": 0.9486,
+      "step": 12152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.761308836037905e-06,
+      "loss": 0.8882,
+      "step": 12153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7600795938509141e-06,
+      "loss": 0.9051,
+      "step": 12154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.758850739372756e-06,
+      "loss": 0.8843,
+      "step": 12155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7576222726612524e-06,
+      "loss": 0.9019,
+      "step": 12156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7563941937742102e-06,
+      "loss": 0.9193,
+      "step": 12157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7551665027694108e-06,
+      "loss": 0.8912,
+      "step": 12158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7539391997046218e-06,
+      "loss": 0.933,
+      "step": 12159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.75271228463759e-06,
+      "loss": 0.9219,
+      "step": 12160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7514857576260414e-06,
+      "loss": 0.944,
+      "step": 12161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7502596187276954e-06,
+      "loss": 0.9825,
+      "step": 12162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.749033868000235e-06,
+      "loss": 1.0051,
+      "step": 12163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7478085055013428e-06,
+      "loss": 0.9877,
+      "step": 12164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7465835312886714e-06,
+      "loss": 0.9581,
+      "step": 12165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7453589454198606e-06,
+      "loss": 0.9013,
+      "step": 12166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7441347479525283e-06,
+      "loss": 0.9726,
+      "step": 12167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7429109389442744e-06,
+      "loss": 0.8711,
+      "step": 12168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.741687518452687e-06,
+      "loss": 0.969,
+      "step": 12169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7404644865353314e-06,
+      "loss": 0.9241,
+      "step": 12170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.739241843249747e-06,
+      "loss": 0.9686,
+      "step": 12171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.73801958865347e-06,
+      "loss": 0.888,
+      "step": 12172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7367977228040067e-06,
+      "loss": 0.9028,
+      "step": 12173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.73557624575885e-06,
+      "loss": 0.9153,
+      "step": 12174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7343551575754724e-06,
+      "loss": 0.914,
+      "step": 12175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7331344583113285e-06,
+      "loss": 0.8924,
+      "step": 12176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7319141480238588e-06,
+      "loss": 0.9308,
+      "step": 12177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7306942267704817e-06,
+      "loss": 0.9815,
+      "step": 12178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7294746946085917e-06,
+      "loss": 0.9106,
+      "step": 12179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7282555515955769e-06,
+      "loss": 0.963,
+      "step": 12180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.727036797788798e-06,
+      "loss": 0.9387,
+      "step": 12181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7258184332456018e-06,
+      "loss": 0.9662,
+      "step": 12182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7246004580233156e-06,
+      "loss": 0.9729,
+      "step": 12183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7233828721792445e-06,
+      "loss": 0.9051,
+      "step": 12184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.722165675770684e-06,
+      "loss": 0.9413,
+      "step": 12185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7209488688549048e-06,
+      "loss": 0.9428,
+      "step": 12186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.71973245148916e-06,
+      "loss": 0.894,
+      "step": 12187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.718516423730685e-06,
+      "loss": 0.8304,
+      "step": 12188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7173007856366973e-06,
+      "loss": 0.9272,
+      "step": 12189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7160855372643947e-06,
+      "loss": 0.8872,
+      "step": 12190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7148706786709568e-06,
+      "loss": 0.9511,
+      "step": 12191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.71365620991355e-06,
+      "loss": 0.9544,
+      "step": 12192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7124421310493144e-06,
+      "loss": 0.8772,
+      "step": 12193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7112284421353775e-06,
+      "loss": 0.8716,
+      "step": 12194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.710015143228846e-06,
+      "loss": 0.9132,
+      "step": 12195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7088022343868038e-06,
+      "loss": 0.8206,
+      "step": 12196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7075897156663323e-06,
+      "loss": 0.9617,
+      "step": 12197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7063775871244736e-06,
+      "loss": 0.9453,
+      "step": 12198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.705165848818262e-06,
+      "loss": 0.9619,
+      "step": 12199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7039545008047175e-06,
+      "loss": 0.8942,
+      "step": 12200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7027435431408356e-06,
+      "loss": 1.0155,
+      "step": 12201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.701532975883594e-06,
+      "loss": 0.9135,
+      "step": 12202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7003227990899528e-06,
+      "loss": 0.9119,
+      "step": 12203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.699113012816852e-06,
+      "loss": 0.8396,
+      "step": 12204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6979036171212216e-06,
+      "loss": 0.9253,
+      "step": 12205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6966946120599582e-06,
+      "loss": 0.9099,
+      "step": 12206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6954859976899552e-06,
+      "loss": 0.9042,
+      "step": 12207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6942777740680782e-06,
+      "loss": 0.8914,
+      "step": 12208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6930699412511765e-06,
+      "loss": 0.9215,
+      "step": 12209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.691862499296083e-06,
+      "loss": 0.937,
+      "step": 12210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6906554482596104e-06,
+      "loss": 0.857,
+      "step": 12211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6894487881985532e-06,
+      "loss": 1.0008,
+      "step": 12212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6882425191696882e-06,
+      "loss": 0.9581,
+      "step": 12213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6870366412297701e-06,
+      "loss": 0.8626,
+      "step": 12214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6858311544355443e-06,
+      "loss": 0.9551,
+      "step": 12215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6846260588437302e-06,
+      "loss": 0.9273,
+      "step": 12216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.683421354511028e-06,
+      "loss": 0.9846,
+      "step": 12217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6822170414941241e-06,
+      "loss": 0.9004,
+      "step": 12218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6810131198496815e-06,
+      "loss": 0.8816,
+      "step": 12219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798095896343548e-06,
+      "loss": 1.0,
+      "step": 12220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.678606450904766e-06,
+      "loss": 0.995,
+      "step": 12221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.677403703717526e-06,
+      "loss": 0.9277,
+      "step": 12222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6762013481292316e-06,
+      "loss": 0.9492,
+      "step": 12223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.674999384196454e-06,
+      "loss": 0.9183,
+      "step": 12224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.673797811975749e-06,
+      "loss": 0.8585,
+      "step": 12225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6725966315236542e-06,
+      "loss": 0.9603,
+      "step": 12226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6713958428966836e-06,
+      "loss": 0.9561,
+      "step": 12227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.670195446151347e-06,
+      "loss": 0.8867,
+      "step": 12228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6689954413441144e-06,
+      "loss": 0.9805,
+      "step": 12229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6677958285314588e-06,
+      "loss": 0.9547,
+      "step": 12230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6665966077698193e-06,
+      "loss": 0.9108,
+      "step": 12231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6653977791156251e-06,
+      "loss": 0.9558,
+      "step": 12232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6641993426252823e-06,
+      "loss": 0.9283,
+      "step": 12233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6630012983551792e-06,
+      "loss": 0.98,
+      "step": 12234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6618036463616905e-06,
+      "loss": 0.9642,
+      "step": 12235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6606063867011691e-06,
+      "loss": 0.9754,
+      "step": 12236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6594095194299409e-06,
+      "loss": 0.9446,
+      "step": 12237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.658213044604331e-06,
+      "loss": 0.9082,
+      "step": 12238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657016962280632e-06,
+      "loss": 0.9213,
+      "step": 12239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6558212725151235e-06,
+      "loss": 0.9716,
+      "step": 12240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6546259753640659e-06,
+      "loss": 0.9286,
+      "step": 12241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6534310708836976e-06,
+      "loss": 0.8985,
+      "step": 12242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652236559130248e-06,
+      "loss": 0.8913,
+      "step": 12243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6510424401599179e-06,
+      "loss": 0.9698,
+      "step": 12244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6498487140288954e-06,
+      "loss": 0.9477,
+      "step": 12245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6486553807933458e-06,
+      "loss": 0.9616,
+      "step": 12246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6474624405094209e-06,
+      "loss": 0.8804,
+      "step": 12247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.64626989323325e-06,
+      "loss": 0.9895,
+      "step": 12248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.645077739020946e-06,
+      "loss": 0.919,
+      "step": 12249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6438859779286008e-06,
+      "loss": 0.972,
+      "step": 12250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6426946100122965e-06,
+      "loss": 0.9087,
+      "step": 12251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.641503635328079e-06,
+      "loss": 0.8593,
+      "step": 12252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403130539319957e-06,
+      "loss": 0.9232,
+      "step": 12253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.639122865880063e-06,
+      "loss": 1.0058,
+      "step": 12254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6379330712282837e-06,
+      "loss": 0.9457,
+      "step": 12255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6367436700326378e-06,
+      "loss": 0.9578,
+      "step": 12256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6355546623490904e-06,
+      "loss": 0.9757,
+      "step": 12257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.63436604823359e-06,
+      "loss": 0.9058,
+      "step": 12258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.633177827742064e-06,
+      "loss": 0.9929,
+      "step": 12259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6319900009304146e-06,
+      "loss": 0.9272,
+      "step": 12260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6308025678545392e-06,
+      "loss": 0.8695,
+      "step": 12261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6296155285703052e-06,
+      "loss": 0.9198,
+      "step": 12262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6284288831335716e-06,
+      "loss": 0.9516,
+      "step": 12263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6272426316001654e-06,
+      "loss": 0.9863,
+      "step": 12264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.626056774025906e-06,
+      "loss": 0.8386,
+      "step": 12265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6248713104665915e-06,
+      "loss": 0.9215,
+      "step": 12266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6236862409780019e-06,
+      "loss": 1.0008,
+      "step": 12267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.622501565615896e-06,
+      "loss": 0.8767,
+      "step": 12268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6213172844360158e-06,
+      "loss": 0.9383,
+      "step": 12269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6201333974940848e-06,
+      "loss": 0.9837,
+      "step": 12270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.618949904845809e-06,
+      "loss": 0.9213,
+      "step": 12271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.61776680654687e-06,
+      "loss": 0.9212,
+      "step": 12272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6165841026529428e-06,
+      "loss": 0.9451,
+      "step": 12273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.615401793219672e-06,
+      "loss": 0.8739,
+      "step": 12274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6142198783026896e-06,
+      "loss": 0.9149,
+      "step": 12275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.613038357957607e-06,
+      "loss": 0.8829,
+      "step": 12276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6118572322400162e-06,
+      "loss": 0.8841,
+      "step": 12277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6106765012054981e-06,
+      "loss": 0.9029,
+      "step": 12278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6094961649096031e-06,
+      "loss": 0.9196,
+      "step": 12279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6083162234078675e-06,
+      "loss": 0.9653,
+      "step": 12280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.607136676755816e-06,
+      "loss": 0.933,
+      "step": 12281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6059575250089476e-06,
+      "loss": 0.8328,
+      "step": 12282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6047787682227433e-06,
+      "loss": 0.9832,
+      "step": 12283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6036004064526667e-06,
+      "loss": 0.9502,
+      "step": 12284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6024224397541599e-06,
+      "loss": 0.8942,
+      "step": 12285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6012448681826575e-06,
+      "loss": 0.9596,
+      "step": 12286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.60006769179356e-06,
+      "loss": 0.896,
+      "step": 12287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5988909106422556e-06,
+      "loss": 0.9463,
+      "step": 12288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5977145247841197e-06,
+      "loss": 0.8992,
+      "step": 12289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.596538534274501e-06,
+      "loss": 0.9144,
+      "step": 12290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5953629391687352e-06,
+      "loss": 1.0263,
+      "step": 12291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.594187739522135e-06,
+      "loss": 1.0195,
+      "step": 12292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5930129353899937e-06,
+      "loss": 0.877,
+      "step": 12293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5918385268275981e-06,
+      "loss": 0.8893,
+      "step": 12294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5906645138901954e-06,
+      "loss": 0.9424,
+      "step": 12295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5894908966330347e-06,
+      "loss": 0.957,
+      "step": 12296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5883176751113338e-06,
+      "loss": 0.9508,
+      "step": 12297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5871448493802966e-06,
+      "loss": 0.9843,
+      "step": 12298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5859724194951076e-06,
+      "loss": 0.9742,
+      "step": 12299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5848003855109295e-06,
+      "loss": 0.9222,
+      "step": 12300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5836287474829138e-06,
+      "loss": 0.9293,
+      "step": 12301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5824575054661906e-06,
+      "loss": 0.9434,
+      "step": 12302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5812866595158615e-06,
+      "loss": 0.9699,
+      "step": 12303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5801162096870248e-06,
+      "loss": 0.9491,
+      "step": 12304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5789461560347508e-06,
+      "loss": 0.932,
+      "step": 12305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577776498614093e-06,
+      "loss": 0.9144,
+      "step": 12306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5766072374800879e-06,
+      "loss": 0.8968,
+      "step": 12307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5754383726877497e-06,
+      "loss": 0.9488,
+      "step": 12308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5742699042920828e-06,
+      "loss": 0.9637,
+      "step": 12309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5731018323480563e-06,
+      "loss": 0.9307,
+      "step": 12310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5719341569106406e-06,
+      "loss": 0.9185,
+      "step": 12311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5707668780347728e-06,
+      "loss": 1.026,
+      "step": 12312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5695999957753772e-06,
+      "loss": 0.9536,
+      "step": 12313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5684335101873583e-06,
+      "loss": 0.8811,
+      "step": 12314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5672674213256e-06,
+      "loss": 0.959,
+      "step": 12315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5661017292449754e-06,
+      "loss": 0.9043,
+      "step": 12316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5649364340003325e-06,
+      "loss": 0.96,
+      "step": 12317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5637715356464934e-06,
+      "loss": 0.965,
+      "step": 12318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5626070342382771e-06,
+      "loss": 0.881,
+      "step": 12319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5614429298304724e-06,
+      "loss": 0.9292,
+      "step": 12320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5602792224778584e-06,
+      "loss": 0.9239,
+      "step": 12321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5591159122351863e-06,
+      "loss": 0.9352,
+      "step": 12322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5579529991571896e-06,
+      "loss": 0.8602,
+      "step": 12323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5567904832985937e-06,
+      "loss": 0.9074,
+      "step": 12324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5556283647140957e-06,
+      "loss": 0.9232,
+      "step": 12325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5544666434583712e-06,
+      "loss": 0.9396,
+      "step": 12326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.553305319586087e-06,
+      "loss": 0.9247,
+      "step": 12327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5521443931518865e-06,
+      "loss": 0.8916,
+      "step": 12328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5509838642103913e-06,
+      "loss": 0.9414,
+      "step": 12329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.549823732816209e-06,
+      "loss": 0.8989,
+      "step": 12330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5486639990239239e-06,
+      "loss": 0.8453,
+      "step": 12331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5475046628881085e-06,
+      "loss": 0.8778,
+      "step": 12332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5463457244633117e-06,
+      "loss": 0.8714,
+      "step": 12333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5451871838040632e-06,
+      "loss": 0.9593,
+      "step": 12334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.544029040964875e-06,
+      "loss": 0.9283,
+      "step": 12335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5428712960002411e-06,
+      "loss": 0.8887,
+      "step": 12336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.541713948964636e-06,
+      "loss": 0.9474,
+      "step": 12337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5405569999125148e-06,
+      "loss": 0.9575,
+      "step": 12338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5394004488983172e-06,
+      "loss": 0.8555,
+      "step": 12339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.538244295976461e-06,
+      "loss": 0.9403,
+      "step": 12340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5370885412013458e-06,
+      "loss": 0.8722,
+      "step": 12341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.535933184627354e-06,
+      "loss": 0.9503,
+      "step": 12342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.534778226308843e-06,
+      "loss": 0.9295,
+      "step": 12343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5336236663001647e-06,
+      "loss": 0.8997,
+      "step": 12344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.532469504655638e-06,
+      "loss": 0.8958,
+      "step": 12345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5313157414295676e-06,
+      "loss": 0.9154,
+      "step": 12346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5301623766762475e-06,
+      "loss": 0.9893,
+      "step": 12347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.529009410449942e-06,
+      "loss": 0.854,
+      "step": 12348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5278568428049013e-06,
+      "loss": 0.891,
+      "step": 12349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5267046737953583e-06,
+      "loss": 0.9308,
+      "step": 12350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5255529034755211e-06,
+      "loss": 0.9457,
+      "step": 12351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5244015318995919e-06,
+      "loss": 0.9298,
+      "step": 12352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.523250559121735e-06,
+      "loss": 0.8969,
+      "step": 12353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5220999851961149e-06,
+      "loss": 0.9498,
+      "step": 12354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.520949810176865e-06,
+      "loss": 0.9416,
+      "step": 12355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5198000341181051e-06,
+      "loss": 0.9165,
+      "step": 12356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5186506570739356e-06,
+      "loss": 0.957,
+      "step": 12357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5175016790984343e-06,
+      "loss": 0.9159,
+      "step": 12358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5163531002456676e-06,
+      "loss": 0.936,
+      "step": 12359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5152049205696794e-06,
+      "loss": 1.01,
+      "step": 12360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5140571401244886e-06,
+      "loss": 0.9626,
+      "step": 12361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5129097589641062e-06,
+      "loss": 0.9597,
+      "step": 12362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.511762777142518e-06,
+      "loss": 0.8826,
+      "step": 12363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5106161947136933e-06,
+      "loss": 0.9584,
+      "step": 12364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5094700117315797e-06,
+      "loss": 0.882,
+      "step": 12365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5083242282501077e-06,
+      "loss": 0.9443,
+      "step": 12366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.507178844323195e-06,
+      "loss": 0.9615,
+      "step": 12367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5060338600047276e-06,
+      "loss": 0.8967,
+      "step": 12368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.504889275348581e-06,
+      "loss": 0.8982,
+      "step": 12369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5037450904086138e-06,
+      "loss": 0.9501,
+      "step": 12370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5026013052386623e-06,
+      "loss": 0.9967,
+      "step": 12371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5014579198925428e-06,
+      "loss": 1.0049,
+      "step": 12372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5003149344240552e-06,
+      "loss": 0.9414,
+      "step": 12373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.499172348886978e-06,
+      "loss": 0.9993,
+      "step": 12374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4980301633350792e-06,
+      "loss": 0.9704,
+      "step": 12375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.496888377822091e-06,
+      "loss": 0.9395,
+      "step": 12376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4957469924017464e-06,
+      "loss": 0.9883,
+      "step": 12377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.494606007127747e-06,
+      "loss": 0.8051,
+      "step": 12378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.493465422053778e-06,
+      "loss": 0.8279,
+      "step": 12379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4923252372335084e-06,
+      "loss": 0.8667,
+      "step": 12380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4911854527205838e-06,
+      "loss": 0.8669,
+      "step": 12381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.490046068568639e-06,
+      "loss": 0.9346,
+      "step": 12382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4889070848312848e-06,
+      "loss": 1.0702,
+      "step": 12383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4877685015621046e-06,
+      "loss": 0.8925,
+      "step": 12384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4866303188146813e-06,
+      "loss": 0.9646,
+      "step": 12385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4854925366425643e-06,
+      "loss": 0.9371,
+      "step": 12386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.484355155099292e-06,
+      "loss": 0.8482,
+      "step": 12387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4832181742383778e-06,
+      "loss": 0.9232,
+      "step": 12388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4820815941133194e-06,
+      "loss": 0.8801,
+      "step": 12389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4809454147775992e-06,
+      "loss": 0.911,
+      "step": 12390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.479809636284676e-06,
+      "loss": 0.8702,
+      "step": 12391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.47867425868799e-06,
+      "loss": 0.9213,
+      "step": 12392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.477539282040964e-06,
+      "loss": 0.8995,
+      "step": 12393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.476404706397002e-06,
+      "loss": 0.8862,
+      "step": 12394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4752705318094874e-06,
+      "loss": 0.8968,
+      "step": 12395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4741367583317844e-06,
+      "loss": 0.9209,
+      "step": 12396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4730033860172455e-06,
+      "loss": 0.9268,
+      "step": 12397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4718704149191943e-06,
+      "loss": 0.9497,
+      "step": 12398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4707378450909416e-06,
+      "loss": 0.9227,
+      "step": 12399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.469605676585777e-06,
+      "loss": 0.8652,
+      "step": 12400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.468473909456971e-06,
+      "loss": 0.8675,
+      "step": 12401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.467342543757778e-06,
+      "loss": 0.9359,
+      "step": 12402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.46621157954143e-06,
+      "loss": 0.9371,
+      "step": 12403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4650810168611396e-06,
+      "loss": 0.8929,
+      "step": 12404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4639508557701078e-06,
+      "loss": 0.9477,
+      "step": 12405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4628210963215118e-06,
+      "loss": 0.9349,
+      "step": 12406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4616917385685004e-06,
+      "loss": 0.9631,
+      "step": 12407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4605627825642232e-06,
+      "loss": 0.9312,
+      "step": 12408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4594342283617924e-06,
+      "loss": 0.9062,
+      "step": 12409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4583060760143186e-06,
+      "loss": 0.8683,
+      "step": 12410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4571783255748762e-06,
+      "loss": 0.8795,
+      "step": 12411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4560509770965292e-06,
+      "loss": 0.9442,
+      "step": 12412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4549240306323253e-06,
+      "loss": 1.0039,
+      "step": 12413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.45379748623529e-06,
+      "loss": 0.9188,
+      "step": 12414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4526713439584284e-06,
+      "loss": 0.9075,
+      "step": 12415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4515456038547292e-06,
+      "loss": 0.934,
+      "step": 12416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4504202659771593e-06,
+      "loss": 1.016,
+      "step": 12417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449295330378674e-06,
+      "loss": 0.8704,
+      "step": 12418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448170797112196e-06,
+      "loss": 0.9059,
+      "step": 12419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447046666230645e-06,
+      "loss": 0.9227,
+      "step": 12420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.445922937786911e-06,
+      "loss": 0.9644,
+      "step": 12421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444799611833868e-06,
+      "loss": 0.9234,
+      "step": 12422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4436766884243724e-06,
+      "loss": 0.9672,
+      "step": 12423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4425541676112576e-06,
+      "loss": 0.9557,
+      "step": 12424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4414320494473477e-06,
+      "loss": 0.9255,
+      "step": 12425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.440310333985434e-06,
+      "loss": 0.939,
+      "step": 12426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4391890212782978e-06,
+      "loss": 0.9369,
+      "step": 12427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.438068111378702e-06,
+      "loss": 0.9518,
+      "step": 12428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.436947604339387e-06,
+      "loss": 0.9222,
+      "step": 12429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.435827500213075e-06,
+      "loss": 0.9642,
+      "step": 12430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4347077990524705e-06,
+      "loss": 0.8801,
+      "step": 12431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4335885009102556e-06,
+      "loss": 0.8547,
+      "step": 12432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.432469605839102e-06,
+      "loss": 0.9437,
+      "step": 12433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4313511138916491e-06,
+      "loss": 0.9047,
+      "step": 12434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.43023302512053e-06,
+      "loss": 0.9661,
+      "step": 12435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.429115339578353e-06,
+      "loss": 0.9226,
+      "step": 12436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4279980573177055e-06,
+      "loss": 0.9121,
+      "step": 12437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4268811783911606e-06,
+      "loss": 0.9075,
+      "step": 12438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4257647028512688e-06,
+      "loss": 0.9376,
+      "step": 12439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4246486307505626e-06,
+      "loss": 0.9829,
+      "step": 12440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4235329621415616e-06,
+      "loss": 0.9355,
+      "step": 12441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4224176970767512e-06,
+      "loss": 0.9606,
+      "step": 12442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4213028356086144e-06,
+      "loss": 0.9228,
+      "step": 12443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4201883777896076e-06,
+      "loss": 0.9169,
+      "step": 12444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.419074323672167e-06,
+      "loss": 0.9262,
+      "step": 12445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4179606733087126e-06,
+      "loss": 0.9381,
+      "step": 12446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4168474267516408e-06,
+      "loss": 0.9694,
+      "step": 12447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4157345840533387e-06,
+      "loss": 0.9096,
+      "step": 12448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.414622145266168e-06,
+      "loss": 0.9202,
+      "step": 12449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4135101104424643e-06,
+      "loss": 0.938,
+      "step": 12450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412398479634559e-06,
+      "loss": 1.0446,
+      "step": 12451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4112872528947553e-06,
+      "loss": 1.017,
+      "step": 12452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4101764302753373e-06,
+      "loss": 0.9295,
+      "step": 12453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4090660118285738e-06,
+      "loss": 0.9109,
+      "step": 12454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4079559976067103e-06,
+      "loss": 0.9799,
+      "step": 12455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4068463876619798e-06,
+      "loss": 0.9589,
+      "step": 12456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4057371820465893e-06,
+      "loss": 0.9535,
+      "step": 12457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4046283808127314e-06,
+      "loss": 0.9695,
+      "step": 12458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4035199840125757e-06,
+      "loss": 0.9723,
+      "step": 12459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4024119916982782e-06,
+      "loss": 1.0419,
+      "step": 12460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4013044039219691e-06,
+      "loss": 0.8634,
+      "step": 12461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.400197220735764e-06,
+      "loss": 0.9303,
+      "step": 12462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3990904421917606e-06,
+      "loss": 0.9758,
+      "step": 12463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3979840683420376e-06,
+      "loss": 0.9125,
+      "step": 12464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3968780992386454e-06,
+      "loss": 0.9366,
+      "step": 12465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3957725349336282e-06,
+      "loss": 0.9239,
+      "step": 12466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.394667375479003e-06,
+      "loss": 0.9775,
+      "step": 12467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.393562620926775e-06,
+      "loss": 1.0348,
+      "step": 12468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3924582713289204e-06,
+      "loss": 0.8853,
+      "step": 12469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3913543267374009e-06,
+      "loss": 0.9171,
+      "step": 12470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3902507872041647e-06,
+      "loss": 0.9008,
+      "step": 12471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3891476527811342e-06,
+      "loss": 0.9455,
+      "step": 12472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3880449235202132e-06,
+      "loss": 0.9597,
+      "step": 12473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3869425994732888e-06,
+      "loss": 0.9448,
+      "step": 12474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3858406806922287e-06,
+      "loss": 0.9077,
+      "step": 12475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.38473916722888e-06,
+      "loss": 0.9847,
+      "step": 12476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3836380591350728e-06,
+      "loss": 0.926,
+      "step": 12477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3825373564626133e-06,
+      "loss": 0.9621,
+      "step": 12478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3814370592632965e-06,
+      "loss": 0.9557,
+      "step": 12479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3803371675888943e-06,
+      "loss": 0.8621,
+      "step": 12480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3792376814911568e-06,
+      "loss": 0.9799,
+      "step": 12481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3781386010218189e-06,
+      "loss": 0.9578,
+      "step": 12482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3770399262325951e-06,
+      "loss": 0.8967,
+      "step": 12483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3759416571751804e-06,
+      "loss": 0.8796,
+      "step": 12484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3748437939012493e-06,
+      "loss": 0.8925,
+      "step": 12485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3737463364624627e-06,
+      "loss": 1.038,
+      "step": 12486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3726492849104578e-06,
+      "loss": 0.9328,
+      "step": 12487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3715526392968516e-06,
+      "loss": 0.8746,
+      "step": 12488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3704563996732457e-06,
+      "loss": 0.8751,
+      "step": 12489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.369360566091218e-06,
+      "loss": 0.8609,
+      "step": 12490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.368265138602337e-06,
+      "loss": 0.9592,
+      "step": 12491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3671701172581387e-06,
+      "loss": 0.8703,
+      "step": 12492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.366075502110147e-06,
+      "loss": 0.9598,
+      "step": 12493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3649812932098694e-06,
+      "loss": 0.9104,
+      "step": 12494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3638874906087907e-06,
+      "loss": 0.9229,
+      "step": 12495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.362794094358375e-06,
+      "loss": 0.9937,
+      "step": 12496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3617011045100714e-06,
+      "loss": 0.9291,
+      "step": 12497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3606085211153054e-06,
+      "loss": 0.9103,
+      "step": 12498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.35951634422549e-06,
+      "loss": 0.9924,
+      "step": 12499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3584245738920088e-06,
+      "loss": 0.8879,
+      "step": 12500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3573332101662373e-06,
+      "loss": 0.8664,
+      "step": 12501
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3562422530995256e-06,
+      "loss": 1.0131,
+      "step": 12502
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3551517027432048e-06,
+      "loss": 0.9381,
+      "step": 12503
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3540615591485896e-06,
+      "loss": 0.9688,
+      "step": 12504
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3529718223669707e-06,
+      "loss": 0.9981,
+      "step": 12505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3518824924496277e-06,
+      "loss": 0.8322,
+      "step": 12506
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.350793569447817e-06,
+      "loss": 0.9196,
+      "step": 12507
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3497050534127665e-06,
+      "loss": 0.9575,
+      "step": 12508
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.348616944395702e-06,
+      "loss": 0.9223,
+      "step": 12509
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3475292424478193e-06,
+      "loss": 0.9873,
+      "step": 12510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3464419476202962e-06,
+      "loss": 0.9098,
+      "step": 12511
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3453550599642939e-06,
+      "loss": 0.9262,
+      "step": 12512
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3442685795309518e-06,
+      "loss": 0.8865,
+      "step": 12513
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3431825063713932e-06,
+      "loss": 0.8714,
+      "step": 12514
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3420968405367229e-06,
+      "loss": 0.9351,
+      "step": 12515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.341011582078018e-06,
+      "loss": 0.8982,
+      "step": 12516
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339926731046347e-06,
+      "loss": 0.8893,
+      "step": 12517
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3388422874927531e-06,
+      "loss": 1.0076,
+      "step": 12518
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3377582514682642e-06,
+      "loss": 0.96,
+      "step": 12519
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3366746230238846e-06,
+      "loss": 0.9053,
+      "step": 12520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3355914022105998e-06,
+      "loss": 0.9274,
+      "step": 12521
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3345085890793863e-06,
+      "loss": 0.8999,
+      "step": 12522
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3334261836811824e-06,
+      "loss": 0.9868,
+      "step": 12523
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3323441860669261e-06,
+      "loss": 0.89,
+      "step": 12524
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3312625962875248e-06,
+      "loss": 0.8853,
+      "step": 12525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3301814143938706e-06,
+      "loss": 0.941,
+      "step": 12526
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3291006404368355e-06,
+      "loss": 0.9167,
+      "step": 12527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.328020274467271e-06,
+      "loss": 0.9145,
+      "step": 12528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.326940316536014e-06,
+      "loss": 0.9361,
+      "step": 12529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3258607666938806e-06,
+      "loss": 0.8962,
+      "step": 12530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.32478162499166e-06,
+      "loss": 0.9631,
+      "step": 12531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3237028914801331e-06,
+      "loss": 0.9533,
+      "step": 12532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3226245662100568e-06,
+      "loss": 0.9397,
+      "step": 12533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3215466492321693e-06,
+      "loss": 0.9443,
+      "step": 12534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.320469140597187e-06,
+      "loss": 0.8765,
+      "step": 12535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.319392040355809e-06,
+      "loss": 0.9348,
+      "step": 12536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3183153485587185e-06,
+      "loss": 0.9065,
+      "step": 12537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3172390652565769e-06,
+      "loss": 0.8944,
+      "step": 12538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3161631905000227e-06,
+      "loss": 0.9794,
+      "step": 12539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.315087724339682e-06,
+      "loss": 0.8885,
+      "step": 12540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3140126668261543e-06,
+      "loss": 0.9348,
+      "step": 12541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.312938018010027e-06,
+      "loss": 0.9186,
+      "step": 12542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.311863777941862e-06,
+      "loss": 0.964,
+      "step": 12543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3107899466722085e-06,
+      "loss": 0.8831,
+      "step": 12544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3097165242515908e-06,
+      "loss": 0.897,
+      "step": 12545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3086435107305173e-06,
+      "loss": 0.926,
+      "step": 12546
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3075709061594744e-06,
+      "loss": 0.9363,
+      "step": 12547
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3064987105889294e-06,
+      "loss": 0.9182,
+      "step": 12548
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3054269240693397e-06,
+      "loss": 0.9084,
+      "step": 12549
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304355546651126e-06,
+      "loss": 0.8697,
+      "step": 12550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3032845783847014e-06,
+      "loss": 0.9224,
+      "step": 12551
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3022140193204623e-06,
+      "loss": 0.9757,
+      "step": 12552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3011438695087775e-06,
+      "loss": 0.905,
+      "step": 12553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3000741290000008e-06,
+      "loss": 0.8825,
+      "step": 12554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2990047978444676e-06,
+      "loss": 0.895,
+      "step": 12555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2979358760924875e-06,
+      "loss": 0.9388,
+      "step": 12556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2968673637943652e-06,
+      "loss": 0.9309,
+      "step": 12557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295799261000369e-06,
+      "loss": 0.9093,
+      "step": 12558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2947315677607554e-06,
+      "loss": 0.9863,
+      "step": 12559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2936642841257675e-06,
+      "loss": 0.9843,
+      "step": 12560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2925974101456207e-06,
+      "loss": 0.9169,
+      "step": 12561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291530945870515e-06,
+      "loss": 0.8754,
+      "step": 12562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.290464891350629e-06,
+      "loss": 0.8757,
+      "step": 12563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2893992466361227e-06,
+      "loss": 0.8953,
+      "step": 12564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2883340117771415e-06,
+      "loss": 0.8995,
+      "step": 12565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2872691868238007e-06,
+      "loss": 0.9201,
+      "step": 12566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2862047718262094e-06,
+      "loss": 0.9727,
+      "step": 12567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2851407668344474e-06,
+      "loss": 0.9632,
+      "step": 12568
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.284077171898579e-06,
+      "loss": 0.9663,
+      "step": 12569
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2830139870686508e-06,
+      "loss": 1.0214,
+      "step": 12570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2819512123946841e-06,
+      "loss": 0.9434,
+      "step": 12571
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2808888479266934e-06,
+      "loss": 0.9941,
+      "step": 12572
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2798268937146574e-06,
+      "loss": 0.8964,
+      "step": 12573
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.278765349808544e-06,
+      "loss": 0.9117,
+      "step": 12574
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2777042162583075e-06,
+      "loss": 0.9731,
+      "step": 12575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2766434931138727e-06,
+      "loss": 0.945,
+      "step": 12576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2755831804251494e-06,
+      "loss": 0.9336,
+      "step": 12577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2745232782420303e-06,
+      "loss": 0.8946,
+      "step": 12578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2734637866143817e-06,
+      "loss": 0.925,
+      "step": 12579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.272404705592062e-06,
+      "loss": 0.9593,
+      "step": 12580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2713460352248964e-06,
+      "loss": 0.9025,
+      "step": 12581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.270287775562704e-06,
+      "loss": 0.9551,
+      "step": 12582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2692299266552755e-06,
+      "loss": 0.951,
+      "step": 12583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2681724885523861e-06,
+      "loss": 0.8902,
+      "step": 12584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2671154613037907e-06,
+      "loss": 0.8739,
+      "step": 12585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2660588449592259e-06,
+      "loss": 0.8826,
+      "step": 12586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2650026395684045e-06,
+      "loss": 0.9605,
+      "step": 12587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2639468451810311e-06,
+      "loss": 0.9543,
+      "step": 12588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2628914618467747e-06,
+      "loss": 0.9624,
+      "step": 12589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.261836489615299e-06,
+      "loss": 0.9253,
+      "step": 12590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2607819285362432e-06,
+      "loss": 0.886,
+      "step": 12591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2597277786592254e-06,
+      "loss": 0.9394,
+      "step": 12592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.258674040033846e-06,
+      "loss": 0.9756,
+      "step": 12593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.257620712709685e-06,
+      "loss": 0.9047,
+      "step": 12594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2565677967363065e-06,
+      "loss": 0.9401,
+      "step": 12595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.255515292163255e-06,
+      "loss": 0.8995,
+      "step": 12596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2544631990400458e-06,
+      "loss": 0.8502,
+      "step": 12597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2534115174161887e-06,
+      "loss": 0.9992,
+      "step": 12598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2523602473411666e-06,
+      "loss": 0.8537,
+      "step": 12599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2513093888644434e-06,
+      "loss": 0.9108,
+      "step": 12600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.250258942035466e-06,
+      "loss": 0.868,
+      "step": 12601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2492089069036573e-06,
+      "loss": 0.9355,
+      "step": 12602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2481592835184286e-06,
+      "loss": 0.8986,
+      "step": 12603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2471100719291652e-06,
+      "loss": 0.9599,
+      "step": 12604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.246061272185235e-06,
+      "loss": 0.8632,
+      "step": 12605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2450128843359876e-06,
+      "loss": 0.862,
+      "step": 12606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2439649084307515e-06,
+      "loss": 0.9368,
+      "step": 12607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2429173445188359e-06,
+      "loss": 0.9365,
+      "step": 12608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2418701926495303e-06,
+      "loss": 0.8589,
+      "step": 12609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2408234528721085e-06,
+      "loss": 0.8789,
+      "step": 12610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2397771252358226e-06,
+      "loss": 0.9308,
+      "step": 12611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2387312097899028e-06,
+      "loss": 0.8257,
+      "step": 12612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2376857065835634e-06,
+      "loss": 0.8398,
+      "step": 12613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.236640615665995e-06,
+      "loss": 1.0046,
+      "step": 12614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.235595937086378e-06,
+      "loss": 0.8694,
+      "step": 12615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.234551670893861e-06,
+      "loss": 1.009,
+      "step": 12616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2335078171375802e-06,
+      "loss": 1.0095,
+      "step": 12617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2324643758666544e-06,
+      "loss": 0.856,
+      "step": 12618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2314213471301784e-06,
+      "loss": 0.8907,
+      "step": 12619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2303787309772298e-06,
+      "loss": 0.9167,
+      "step": 12620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2293365274568657e-06,
+      "loss": 0.9238,
+      "step": 12621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.228294736618122e-06,
+      "loss": 0.9353,
+      "step": 12622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2272533585100255e-06,
+      "loss": 0.9013,
+      "step": 12623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2262123931815672e-06,
+      "loss": 0.9473,
+      "step": 12624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2251718406817292e-06,
+      "loss": 0.9722,
+      "step": 12625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2241317010594744e-06,
+      "loss": 0.9551,
+      "step": 12626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2230919743637438e-06,
+      "loss": 0.9202,
+      "step": 12627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2220526606434568e-06,
+      "loss": 0.8237,
+      "step": 12628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.221013759947518e-06,
+      "loss": 0.9448,
+      "step": 12629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2199752723248082e-06,
+      "loss": 0.9076,
+      "step": 12630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2189371978241925e-06,
+      "loss": 0.9814,
+      "step": 12631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2178995364945123e-06,
+      "loss": 0.9178,
+      "step": 12632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2168622883845959e-06,
+      "loss": 0.918,
+      "step": 12633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2158254535432467e-06,
+      "loss": 0.9317,
+      "step": 12634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2147890320192501e-06,
+      "loss": 0.8365,
+      "step": 12635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2137530238613737e-06,
+      "loss": 0.9428,
+      "step": 12636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2127174291183596e-06,
+      "loss": 0.8545,
+      "step": 12637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2116822478389434e-06,
+      "loss": 0.9427,
+      "step": 12638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.210647480071826e-06,
+      "loss": 0.8241,
+      "step": 12639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2096131258656962e-06,
+      "loss": 0.9341,
+      "step": 12640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2085791852692264e-06,
+      "loss": 0.9774,
+      "step": 12641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2075456583310653e-06,
+      "loss": 0.919,
+      "step": 12642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2065125450998416e-06,
+      "loss": 0.8931,
+      "step": 12643
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2054798456241656e-06,
+      "loss": 0.975,
+      "step": 12644
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2044475599526284e-06,
+      "loss": 1.0271,
+      "step": 12645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2034156881338055e-06,
+      "loss": 0.9168,
+      "step": 12646
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2023842302162415e-06,
+      "loss": 0.9123,
+      "step": 12647
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2013531862484761e-06,
+      "loss": 0.9389,
+      "step": 12648
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.20032255627902e-06,
+      "loss": 0.9589,
+      "step": 12649
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.199292340356366e-06,
+      "loss": 0.9097,
+      "step": 12650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1982625385289893e-06,
+      "loss": 0.8485,
+      "step": 12651
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1972331508453428e-06,
+      "loss": 0.9407,
+      "step": 12652
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1962041773538657e-06,
+      "loss": 0.9114,
+      "step": 12653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1951756181029728e-06,
+      "loss": 0.908,
+      "step": 12654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1941474731410552e-06,
+      "loss": 0.9062,
+      "step": 12655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1931197425164954e-06,
+      "loss": 0.8726,
+      "step": 12656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.192092426277649e-06,
+      "loss": 0.9118,
+      "step": 12657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1910655244728542e-06,
+      "loss": 0.8705,
+      "step": 12658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1900390371504278e-06,
+      "loss": 0.9695,
+      "step": 12659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1890129643586678e-06,
+      "loss": 0.9761,
+      "step": 12660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.187987306145858e-06,
+      "loss": 0.9741,
+      "step": 12661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1869620625602562e-06,
+      "loss": 0.8682,
+      "step": 12662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1859372336500984e-06,
+      "loss": 0.9021,
+      "step": 12663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1849128194636107e-06,
+      "loss": 0.8585,
+      "step": 12664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183888820048993e-06,
+      "loss": 0.9612,
+      "step": 12665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1828652354544257e-06,
+      "loss": 0.869,
+      "step": 12666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1818420657280728e-06,
+      "loss": 0.8992,
+      "step": 12667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1808193109180733e-06,
+      "loss": 0.9184,
+      "step": 12668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1797969710725565e-06,
+      "loss": 0.9595,
+      "step": 12669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1787750462396196e-06,
+      "loss": 0.9644,
+      "step": 12670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1777535364673508e-06,
+      "loss": 0.8904,
+      "step": 12671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1767324418038139e-06,
+      "loss": 1.0055,
+      "step": 12672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1757117622970548e-06,
+      "loss": 0.9333,
+      "step": 12673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1746914979950963e-06,
+      "loss": 0.9265,
+      "step": 12674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1736716489459454e-06,
+      "loss": 0.926,
+      "step": 12675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1726522151975895e-06,
+      "loss": 0.976,
+      "step": 12676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1716331967979987e-06,
+      "loss": 0.8958,
+      "step": 12677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.170614593795112e-06,
+      "loss": 0.8932,
+      "step": 12678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1695964062368637e-06,
+      "loss": 0.9695,
+      "step": 12679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1685786341711592e-06,
+      "loss": 0.9211,
+      "step": 12680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.167561277645891e-06,
+      "loss": 0.8121,
+      "step": 12681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1665443367089247e-06,
+      "loss": 0.9114,
+      "step": 12682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1655278114081092e-06,
+      "loss": 0.8907,
+      "step": 12683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1645117017912777e-06,
+      "loss": 0.9193,
+      "step": 12684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1634960079062397e-06,
+      "loss": 0.8827,
+      "step": 12685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1624807298007845e-06,
+      "loss": 0.9601,
+      "step": 12686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1614658675226854e-06,
+      "loss": 0.9846,
+      "step": 12687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1604514211196928e-06,
+      "loss": 0.9538,
+      "step": 12688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1594373906395405e-06,
+      "loss": 0.91,
+      "step": 12689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1584237761299377e-06,
+      "loss": 0.9421,
+      "step": 12690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1574105776385824e-06,
+      "loss": 0.9918,
+      "step": 12691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1563977952131444e-06,
+      "loss": 0.9113,
+      "step": 12692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.15538542890128e-06,
+      "loss": 0.8914,
+      "step": 12693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1543734787506233e-06,
+      "loss": 0.9323,
+      "step": 12694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1533619448087852e-06,
+      "loss": 0.9184,
+      "step": 12695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1523508271233684e-06,
+      "loss": 0.9027,
+      "step": 12696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.151340125741942e-06,
+      "loss": 0.9489,
+      "step": 12697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1503298407120621e-06,
+      "loss": 0.9119,
+      "step": 12698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1493199720812698e-06,
+      "loss": 0.8924,
+      "step": 12699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.148310519897079e-06,
+      "loss": 0.918,
+      "step": 12700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1473014842069864e-06,
+      "loss": 0.9449,
+      "step": 12701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1462928650584715e-06,
+      "loss": 1.0128,
+      "step": 12702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.145284662498989e-06,
+      "loss": 0.9374,
+      "step": 12703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1442768765759836e-06,
+      "loss": 0.9729,
+      "step": 12704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1432695073368694e-06,
+      "loss": 0.9239,
+      "step": 12705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1422625548290434e-06,
+      "loss": 0.8364,
+      "step": 12706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.14125601909989e-06,
+      "loss": 1.0104,
+      "step": 12707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1402499001967692e-06,
+      "loss": 0.9014,
+      "step": 12708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1392441981670189e-06,
+      "loss": 0.9916,
+      "step": 12709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1382389130579597e-06,
+      "loss": 0.8837,
+      "step": 12710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1372340449168928e-06,
+      "loss": 0.8825,
+      "step": 12711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.136229593791105e-06,
+      "loss": 0.9155,
+      "step": 12712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1352255597278495e-06,
+      "loss": 0.938,
+      "step": 12713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134221942774375e-06,
+      "loss": 0.9562,
+      "step": 12714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.133218742977903e-06,
+      "loss": 0.8965,
+      "step": 12715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1322159603856363e-06,
+      "loss": 0.9299,
+      "step": 12716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1312135950447567e-06,
+      "loss": 0.9754,
+      "step": 12717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1302116470024282e-06,
+      "loss": 0.8728,
+      "step": 12718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1292101163057967e-06,
+      "loss": 0.9204,
+      "step": 12719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1282090030019888e-06,
+      "loss": 0.9364,
+      "step": 12720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1272083071381023e-06,
+      "loss": 0.9228,
+      "step": 12721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1262080287612298e-06,
+      "loss": 0.8862,
+      "step": 12722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1252081679184324e-06,
+      "loss": 0.9642,
+      "step": 12723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1242087246567579e-06,
+      "loss": 1.0134,
+      "step": 12724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1232096990232322e-06,
+      "loss": 0.9205,
+      "step": 12725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1222110910648598e-06,
+      "loss": 0.9917,
+      "step": 12726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1212129008286343e-06,
+      "loss": 0.9397,
+      "step": 12727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1202151283615138e-06,
+      "loss": 0.9625,
+      "step": 12728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1192177737104525e-06,
+      "loss": 0.8953,
+      "step": 12729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.118220836922378e-06,
+      "loss": 0.9263,
+      "step": 12730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1172243180441954e-06,
+      "loss": 0.9092,
+      "step": 12731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1162282171227967e-06,
+      "loss": 0.9187,
+      "step": 12732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1152325342050474e-06,
+      "loss": 0.8885,
+      "step": 12733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1142372693378012e-06,
+      "loss": 0.872,
+      "step": 12734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1132424225678872e-06,
+      "loss": 0.8358,
+      "step": 12735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1122479939421093e-06,
+      "loss": 0.9345,
+      "step": 12736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1112539835072655e-06,
+      "loss": 0.8716,
+      "step": 12737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1102603913101229e-06,
+      "loss": 0.9572,
+      "step": 12738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1092672173974328e-06,
+      "loss": 0.8932,
+      "step": 12739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.108274461815927e-06,
+      "loss": 0.9456,
+      "step": 12740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1072821246123145e-06,
+      "loss": 0.9251,
+      "step": 12741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1062902058332924e-06,
+      "loss": 0.9447,
+      "step": 12742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1052987055255315e-06,
+      "loss": 0.8864,
+      "step": 12743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1043076237356786e-06,
+      "loss": 0.9604,
+      "step": 12744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1033169605103733e-06,
+      "loss": 0.9771,
+      "step": 12745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.102326715896227e-06,
+      "loss": 0.9543,
+      "step": 12746
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1013368899398314e-06,
+      "loss": 0.9708,
+      "step": 12747
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1003474826877625e-06,
+      "loss": 0.9235,
+      "step": 12748
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0993584941865709e-06,
+      "loss": 0.8864,
+      "step": 12749
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.098369924482795e-06,
+      "loss": 0.9461,
+      "step": 12750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0973817736229486e-06,
+      "loss": 0.9191,
+      "step": 12751
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0963940416535256e-06,
+      "loss": 0.8813,
+      "step": 12752
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0954067286210024e-06,
+      "loss": 0.9857,
+      "step": 12753
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0944198345718339e-06,
+      "loss": 0.8077,
+      "step": 12754
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093433359552455e-06,
+      "loss": 0.9133,
+      "step": 12755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0924473036092808e-06,
+      "loss": 0.9471,
+      "step": 12756
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0914616667887113e-06,
+      "loss": 0.8984,
+      "step": 12757
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0904764491371223e-06,
+      "loss": 0.8535,
+      "step": 12758
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.089491650700869e-06,
+      "loss": 0.9129,
+      "step": 12759
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.08850727152629e-06,
+      "loss": 1.0433,
+      "step": 12760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0875233116596995e-06,
+      "loss": 0.9253,
+      "step": 12761
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0865397711474035e-06,
+      "loss": 0.9383,
+      "step": 12762
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0855566500356706e-06,
+      "loss": 0.9226,
+      "step": 12763
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.084573948370763e-06,
+      "loss": 0.8909,
+      "step": 12764
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.083591666198921e-06,
+      "loss": 0.9132,
+      "step": 12765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0826098035663613e-06,
+      "loss": 0.9264,
+      "step": 12766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0816283605192834e-06,
+      "loss": 0.9424,
+      "step": 12767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.080647337103866e-06,
+      "loss": 0.9291,
+      "step": 12768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.079666733366269e-06,
+      "loss": 0.947,
+      "step": 12769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0786865493526354e-06,
+      "loss": 0.9192,
+      "step": 12770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0777067851090794e-06,
+      "loss": 0.9364,
+      "step": 12771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0767274406817063e-06,
+      "loss": 0.9562,
+      "step": 12772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.075748516116596e-06,
+      "loss": 0.8872,
+      "step": 12773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.074770011459807e-06,
+      "loss": 0.8336,
+      "step": 12774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0737919267573814e-06,
+      "loss": 0.9208,
+      "step": 12775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0728142620553417e-06,
+      "loss": 0.9013,
+      "step": 12776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0718370173996862e-06,
+      "loss": 0.8293,
+      "step": 12777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0708601928364037e-06,
+      "loss": 0.9117,
+      "step": 12778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0698837884114466e-06,
+      "loss": 0.9213,
+      "step": 12779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0689078041707646e-06,
+      "loss": 0.9264,
+      "step": 12780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0679322401602788e-06,
+      "loss": 0.895,
+      "step": 12781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0669570964258891e-06,
+      "loss": 1.0025,
+      "step": 12782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0659823730134811e-06,
+      "loss": 0.9464,
+      "step": 12783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0650080699689158e-06,
+      "loss": 0.8723,
+      "step": 12784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.064034187338041e-06,
+      "loss": 0.8666,
+      "step": 12785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0630607251666757e-06,
+      "loss": 0.9199,
+      "step": 12786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0620876835006222e-06,
+      "loss": 0.8575,
+      "step": 12787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0611150623856714e-06,
+      "loss": 0.9441,
+      "step": 12788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0601428618675824e-06,
+      "loss": 0.869,
+      "step": 12789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0591710819921009e-06,
+      "loss": 0.9247,
+      "step": 12790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0581997228049523e-06,
+      "loss": 0.8874,
+      "step": 12791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.057228784351839e-06,
+      "loss": 0.9372,
+      "step": 12792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0562582666784515e-06,
+      "loss": 0.9647,
+      "step": 12793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0552881698304472e-06,
+      "loss": 1.0234,
+      "step": 12794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.054318493853479e-06,
+      "loss": 0.9757,
+      "step": 12795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0533492387931687e-06,
+      "loss": 0.8892,
+      "step": 12796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0523804046951224e-06,
+      "loss": 0.9487,
+      "step": 12797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.051411991604928e-06,
+      "loss": 0.9636,
+      "step": 12798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0504439995681482e-06,
+      "loss": 0.8999,
+      "step": 12799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0494764286303328e-06,
+      "loss": 0.954,
+      "step": 12800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.04850927883701e-06,
+      "loss": 0.935,
+      "step": 12801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.04754255023368e-06,
+      "loss": 0.8688,
+      "step": 12802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0465762428658355e-06,
+      "loss": 0.962,
+      "step": 12803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0456103567789422e-06,
+      "loss": 0.9524,
+      "step": 12804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.044644892018447e-06,
+      "loss": 0.8999,
+      "step": 12805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0436798486297773e-06,
+      "loss": 0.8897,
+      "step": 12806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0427152266583396e-06,
+      "loss": 0.9344,
+      "step": 12807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0417510261495245e-06,
+      "loss": 1.0327,
+      "step": 12808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.040787247148699e-06,
+      "loss": 0.9363,
+      "step": 12809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0398238897012115e-06,
+      "loss": 0.8272,
+      "step": 12810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.038860953852391e-06,
+      "loss": 0.9314,
+      "step": 12811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0378984396475433e-06,
+      "loss": 0.8689,
+      "step": 12812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0369363471319593e-06,
+      "loss": 0.9525,
+      "step": 12813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0359746763509081e-06,
+      "loss": 0.9354,
+      "step": 12814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0350134273496348e-06,
+      "loss": 0.9786,
+      "step": 12815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0340526001733753e-06,
+      "loss": 0.8969,
+      "step": 12816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0330921948673344e-06,
+      "loss": 0.8492,
+      "step": 12817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.032132211476703e-06,
+      "loss": 0.86,
+      "step": 12818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.031172650046649e-06,
+      "loss": 0.9171,
+      "step": 12819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0302135106223244e-06,
+      "loss": 0.9414,
+      "step": 12820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0292547932488583e-06,
+      "loss": 0.9103,
+      "step": 12821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0282964979713584e-06,
+      "loss": 0.9657,
+      "step": 12822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0273386248349193e-06,
+      "loss": 0.8931,
+      "step": 12823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0263811738846107e-06,
+      "loss": 0.9669,
+      "step": 12824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0254241451654778e-06,
+      "loss": 0.8704,
+      "step": 12825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0244675387225567e-06,
+      "loss": 0.9092,
+      "step": 12826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0235113546008546e-06,
+      "loss": 0.8155,
+      "step": 12827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.022555592845368e-06,
+      "loss": 0.8715,
+      "step": 12828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0216002535010616e-06,
+      "loss": 0.9552,
+      "step": 12829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0206453366128865e-06,
+      "loss": 0.9212,
+      "step": 12830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.019690842225779e-06,
+      "loss": 0.9762,
+      "step": 12831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0187367703846462e-06,
+      "loss": 0.8679,
+      "step": 12832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0177831211343825e-06,
+      "loss": 0.8915,
+      "step": 12833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0168298945198573e-06,
+      "loss": 0.9264,
+      "step": 12834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0158770905859238e-06,
+      "loss": 0.8683,
+      "step": 12835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0149247093774118e-06,
+      "loss": 0.8483,
+      "step": 12836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.013972750939134e-06,
+      "loss": 0.9319,
+      "step": 12837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0130212153158836e-06,
+      "loss": 0.9124,
+      "step": 12838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0120701025524328e-06,
+      "loss": 0.8756,
+      "step": 12839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.011119412693532e-06,
+      "loss": 0.8858,
+      "step": 12840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0101691457839147e-06,
+      "loss": 0.8993,
+      "step": 12841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0092193018682916e-06,
+      "loss": 0.9374,
+      "step": 12842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0082698809913605e-06,
+      "loss": 0.9734,
+      "step": 12843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0073208831977887e-06,
+      "loss": 0.9052,
+      "step": 12844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0063723085322286e-06,
+      "loss": 0.8946,
+      "step": 12845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0054241570393163e-06,
+      "loss": 1.0058,
+      "step": 12846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0044764287636643e-06,
+      "loss": 0.8964,
+      "step": 12847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0035291237498635e-06,
+      "loss": 0.8869,
+      "step": 12848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0025822420424892e-06,
+      "loss": 0.9186,
+      "step": 12849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0016357836860914e-06,
+      "loss": 0.9072,
+      "step": 12850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0006897487252087e-06,
+      "loss": 0.8718,
+      "step": 12851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.997441372043493e-07,
+      "loss": 0.9801,
+      "step": 12852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.987989491680062e-07,
+      "loss": 0.9797,
+      "step": 12853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.978541846606572e-07,
+      "loss": 0.9506,
+      "step": 12854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.969098437267532e-07,
+      "loss": 0.9115,
+      "step": 12855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.959659264107292e-07,
+      "loss": 0.922,
+      "step": 12856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.950224327569968e-07,
+      "loss": 0.9132,
+      "step": 12857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.940793628099499e-07,
+      "loss": 0.924,
+      "step": 12858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.93136716613966e-07,
+      "loss": 0.8501,
+      "step": 12859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.921944942133932e-07,
+      "loss": 0.9131,
+      "step": 12860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.912526956525702e-07,
+      "loss": 0.963,
+      "step": 12861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.903113209758098e-07,
+      "loss": 0.9482,
+      "step": 12862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.893703702274039e-07,
+      "loss": 0.8851,
+      "step": 12863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.884298434516294e-07,
+      "loss": 0.9759,
+      "step": 12864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.874897406927365e-07,
+      "loss": 0.9593,
+      "step": 12865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.86550061994964e-07,
+      "loss": 0.9274,
+      "step": 12866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.856108074025273e-07,
+      "loss": 0.8733,
+      "step": 12867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.846719769596124e-07,
+      "loss": 0.8591,
+      "step": 12868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.837335707104013e-07,
+      "loss": 0.9309,
+      "step": 12869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.827955886990449e-07,
+      "loss": 0.8654,
+      "step": 12870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.81858030969679e-07,
+      "loss": 0.9797,
+      "step": 12871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.809208975664187e-07,
+      "loss": 0.9619,
+      "step": 12872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.79984188533354e-07,
+      "loss": 0.9219,
+      "step": 12873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.79047903914565e-07,
+      "loss": 0.9506,
+      "step": 12874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.781120437541037e-07,
+      "loss": 0.928,
+      "step": 12875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.771766080960055e-07,
+      "loss": 1.0102,
+      "step": 12876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.762415969842832e-07,
+      "loss": 0.9577,
+      "step": 12877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.753070104629336e-07,
+      "loss": 0.8649,
+      "step": 12878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.743728485759296e-07,
+      "loss": 0.885,
+      "step": 12879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.734391113672258e-07,
+      "loss": 0.9756,
+      "step": 12880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.725057988807585e-07,
+      "loss": 0.957,
+      "step": 12881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.715729111604433e-07,
+      "loss": 0.7973,
+      "step": 12882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.706404482501708e-07,
+      "loss": 0.8849,
+      "step": 12883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.69708410193818e-07,
+      "loss": 0.9613,
+      "step": 12884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.68776797035239e-07,
+      "loss": 0.9543,
+      "step": 12885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.678456088182731e-07,
+      "loss": 0.8838,
+      "step": 12886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.669148455867283e-07,
+      "loss": 0.8678,
+      "step": 12887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.659845073844009e-07,
+      "loss": 0.8737,
+      "step": 12888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.650545942550682e-07,
+      "loss": 0.8623,
+      "step": 12889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.64125106242486e-07,
+      "loss": 0.9336,
+      "step": 12890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.63196043390382e-07,
+      "loss": 0.9299,
+      "step": 12891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.622674057424785e-07,
+      "loss": 0.9644,
+      "step": 12892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.613391933424664e-07,
+      "loss": 0.976,
+      "step": 12893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.604114062340219e-07,
+      "loss": 0.8763,
+      "step": 12894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.594840444607989e-07,
+      "loss": 0.9123,
+      "step": 12895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.585571080664301e-07,
+      "loss": 0.9814,
+      "step": 12896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.576305970945355e-07,
+      "loss": 0.9433,
+      "step": 12897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.567045115887053e-07,
+      "loss": 0.9738,
+      "step": 12898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.55778851592516e-07,
+      "loss": 0.8938,
+      "step": 12899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.548536171495216e-07,
+      "loss": 0.8749,
+      "step": 12900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.53928808303256e-07,
+      "loss": 0.938,
+      "step": 12901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.530044250972359e-07,
+      "loss": 0.8827,
+      "step": 12902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.520804675749517e-07,
+      "loss": 0.9397,
+      "step": 12903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.511569357798822e-07,
+      "loss": 0.7834,
+      "step": 12904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.502338297554814e-07,
+      "loss": 0.9992,
+      "step": 12905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.493111495451823e-07,
+      "loss": 0.9292,
+      "step": 12906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.483888951923991e-07,
+      "loss": 0.9765,
+      "step": 12907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.474670667405262e-07,
+      "loss": 0.8676,
+      "step": 12908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.465456642329407e-07,
+      "loss": 0.97,
+      "step": 12909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.456246877129938e-07,
+      "loss": 0.9295,
+      "step": 12910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.447041372240184e-07,
+      "loss": 0.906,
+      "step": 12911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.437840128093322e-07,
+      "loss": 0.9084,
+      "step": 12912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.428643145122296e-07,
+      "loss": 0.8717,
+      "step": 12913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.419450423759824e-07,
+      "loss": 0.9386,
+      "step": 12914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.41026196443845e-07,
+      "loss": 0.948,
+      "step": 12915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.401077767590505e-07,
+      "loss": 0.9266,
+      "step": 12916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.391897833648178e-07,
+      "loss": 0.9369,
+      "step": 12917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.382722163043334e-07,
+      "loss": 0.8609,
+      "step": 12918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.373550756207772e-07,
+      "loss": 0.918,
+      "step": 12919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.364383613573003e-07,
+      "loss": 0.9353,
+      "step": 12920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.355220735570359e-07,
+      "loss": 0.8585,
+      "step": 12921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.346062122630994e-07,
+      "loss": 0.9548,
+      "step": 12922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.33690777518581e-07,
+      "loss": 0.9204,
+      "step": 12923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.327757693665595e-07,
+      "loss": 0.9427,
+      "step": 12924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.31861187850086e-07,
+      "loss": 0.885,
+      "step": 12925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.309470330121906e-07,
+      "loss": 0.8838,
+      "step": 12926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.30033304895891e-07,
+      "loss": 0.9339,
+      "step": 12927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.291200035441783e-07,
+      "loss": 0.8663,
+      "step": 12928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.28207129000026e-07,
+      "loss": 0.9172,
+      "step": 12929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.272946813063865e-07,
+      "loss": 0.9953,
+      "step": 12930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.263826605061921e-07,
+      "loss": 0.9245,
+      "step": 12931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.254710666423605e-07,
+      "loss": 0.9011,
+      "step": 12932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.245598997577787e-07,
+      "loss": 0.8644,
+      "step": 12933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.236491598953201e-07,
+      "loss": 0.9089,
+      "step": 12934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.227388470978404e-07,
+      "loss": 0.9757,
+      "step": 12935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.21828961408171e-07,
+      "loss": 0.9589,
+      "step": 12936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.209195028691231e-07,
+      "loss": 0.8608,
+      "step": 12937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.200104715234892e-07,
+      "loss": 0.893,
+      "step": 12938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.191018674140406e-07,
+      "loss": 0.9046,
+      "step": 12939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.181936905835343e-07,
+      "loss": 0.9029,
+      "step": 12940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.172859410746948e-07,
+      "loss": 0.8294,
+      "step": 12941
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.163786189302403e-07,
+      "loss": 0.9189,
+      "step": 12942
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15471724192859e-07,
+      "loss": 0.8794,
+      "step": 12943
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.145652569052244e-07,
+      "loss": 0.897,
+      "step": 12944
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.136592171099879e-07,
+      "loss": 0.8995,
+      "step": 12945
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.127536048497776e-07,
+      "loss": 0.9778,
+      "step": 12946
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.118484201672096e-07,
+      "loss": 0.8652,
+      "step": 12947
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.109436631048751e-07,
+      "loss": 0.8967,
+      "step": 12948
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.100393337053393e-07,
+      "loss": 0.9511,
+      "step": 12949
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.091354320111589e-07,
+      "loss": 0.8486,
+      "step": 12950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.082319580648635e-07,
+      "loss": 0.9128,
+      "step": 12951
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.073289119089623e-07,
+      "loss": 0.8751,
+      "step": 12952
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.064262935859469e-07,
+      "loss": 0.8844,
+      "step": 12953
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.055241031382856e-07,
+      "loss": 0.9611,
+      "step": 12954
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.046223406084331e-07,
+      "loss": 0.8656,
+      "step": 12955
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.037210060388169e-07,
+      "loss": 0.8606,
+      "step": 12956
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.028200994718472e-07,
+      "loss": 0.8486,
+      "step": 12957
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.019196209499149e-07,
+      "loss": 0.8735,
+      "step": 12958
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.010195705153879e-07,
+      "loss": 0.982,
+      "step": 12959
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.001199482106171e-07,
+      "loss": 0.9089,
+      "step": 12960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.992207540779297e-07,
+      "loss": 0.9695,
+      "step": 12961
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.983219881596384e-07,
+      "loss": 0.9552,
+      "step": 12962
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.974236504980316e-07,
+      "loss": 0.9126,
+      "step": 12963
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.965257411353778e-07,
+      "loss": 0.9755,
+      "step": 12964
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.956282601139255e-07,
+      "loss": 0.9585,
+      "step": 12965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.94731207475904e-07,
+      "loss": 1.0104,
+      "step": 12966
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.93834583263522e-07,
+      "loss": 0.9191,
+      "step": 12967
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.929383875189668e-07,
+      "loss": 0.9147,
+      "step": 12968
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.920426202844057e-07,
+      "loss": 0.9439,
+      "step": 12969
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.911472816019906e-07,
+      "loss": 0.9293,
+      "step": 12970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.902523715138467e-07,
+      "loss": 0.9166,
+      "step": 12971
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.893578900620825e-07,
+      "loss": 0.9158,
+      "step": 12972
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.884638372887855e-07,
+      "loss": 0.9163,
+      "step": 12973
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.87570213236022e-07,
+      "loss": 0.992,
+      "step": 12974
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.866770179458428e-07,
+      "loss": 0.8935,
+      "step": 12975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.85784251460271e-07,
+      "loss": 0.897,
+      "step": 12976
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.84891913821313e-07,
+      "loss": 0.9904,
+      "step": 12977
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.840000050709585e-07,
+      "loss": 0.9129,
+      "step": 12978
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.831085252511739e-07,
+      "loss": 0.9397,
+      "step": 12979
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.822174744039047e-07,
+      "loss": 0.9417,
+      "step": 12980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.813268525710772e-07,
+      "loss": 0.9345,
+      "step": 12981
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.804366597945946e-07,
+      "loss": 0.913,
+      "step": 12982
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.79546896116349e-07,
+      "loss": 0.9412,
+      "step": 12983
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.786575615781989e-07,
+      "loss": 0.9097,
+      "step": 12984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.777686562219956e-07,
+      "loss": 0.9164,
+      "step": 12985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.76880180089561e-07,
+      "loss": 0.9061,
+      "step": 12986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.759921332227006e-07,
+      "loss": 0.9947,
+      "step": 12987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.751045156631999e-07,
+      "loss": 0.926,
+      "step": 12988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.742173274528221e-07,
+      "loss": 0.9141,
+      "step": 12989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.733305686333149e-07,
+      "loss": 0.8871,
+      "step": 12990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.724442392463994e-07,
+      "loss": 0.9673,
+      "step": 12991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.71558339333779e-07,
+      "loss": 0.9531,
+      "step": 12992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.706728689371402e-07,
+      "loss": 0.9342,
+      "step": 12993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.697878280981453e-07,
+      "loss": 0.9254,
+      "step": 12994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.689032168584389e-07,
+      "loss": 0.9414,
+      "step": 12995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.680190352596419e-07,
+      "loss": 0.9824,
+      "step": 12996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.671352833433566e-07,
+      "loss": 0.9202,
+      "step": 12997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.662519611511722e-07,
+      "loss": 0.9507,
+      "step": 12998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.653690687246441e-07,
+      "loss": 0.8779,
+      "step": 12999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.644866061053147e-07,
+      "loss": 0.8569,
+      "step": 13000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.636045733347109e-07,
+      "loss": 0.9793,
+      "step": 13001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.627229704543317e-07,
+      "loss": 0.9311,
+      "step": 13002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.618417975056603e-07,
+      "loss": 0.9045,
+      "step": 13003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.60961054530155e-07,
+      "loss": 0.9022,
+      "step": 13004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.600807415692581e-07,
+      "loss": 0.96,
+      "step": 13005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.592008586643952e-07,
+      "loss": 0.9313,
+      "step": 13006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.583214058569588e-07,
+      "loss": 0.921,
+      "step": 13007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.57442383188336e-07,
+      "loss": 0.9673,
+      "step": 13008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.565637906998847e-07,
+      "loss": 0.8988,
+      "step": 13009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.55685628432944e-07,
+      "loss": 0.9777,
+      "step": 13010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.548078964288343e-07,
+      "loss": 0.9391,
+      "step": 13011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.539305947288545e-07,
+      "loss": 0.9282,
+      "step": 13012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.530537233742853e-07,
+      "loss": 0.9135,
+      "step": 13013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.521772824063878e-07,
+      "loss": 0.8469,
+      "step": 13014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.513012718663927e-07,
+      "loss": 0.9121,
+      "step": 13015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.504256917955267e-07,
+      "loss": 0.9535,
+      "step": 13016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.495505422349848e-07,
+      "loss": 0.9106,
+      "step": 13017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.486758232259451e-07,
+      "loss": 0.884,
+      "step": 13018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.478015348095648e-07,
+      "loss": 0.9492,
+      "step": 13019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.469276770269808e-07,
+      "loss": 0.929,
+      "step": 13020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.460542499193125e-07,
+      "loss": 0.8968,
+      "step": 13021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.451812535276571e-07,
+      "loss": 0.9607,
+      "step": 13022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.443086878930884e-07,
+      "loss": 0.9388,
+      "step": 13023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.434365530566646e-07,
+      "loss": 0.879,
+      "step": 13024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425648490594219e-07,
+      "loss": 0.9326,
+      "step": 13025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.416935759423762e-07,
+      "loss": 0.9481,
+      "step": 13026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.408227337465203e-07,
+      "loss": 0.917,
+      "step": 13027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.399523225128336e-07,
+      "loss": 0.9139,
+      "step": 13028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.390823422822713e-07,
+      "loss": 0.9392,
+      "step": 13029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.382127930957617e-07,
+      "loss": 0.9759,
+      "step": 13030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.373436749942266e-07,
+      "loss": 0.8303,
+      "step": 13031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.364749880185552e-07,
+      "loss": 0.898,
+      "step": 13032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.356067322096273e-07,
+      "loss": 0.941,
+      "step": 13033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.347389076082912e-07,
+      "loss": 0.8714,
+      "step": 13034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.338715142553788e-07,
+      "loss": 0.979,
+      "step": 13035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.330045521917096e-07,
+      "loss": 0.8785,
+      "step": 13036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321380214580732e-07,
+      "loss": 0.9243,
+      "step": 13037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312719220952382e-07,
+      "loss": 0.8933,
+      "step": 13038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.304062541439628e-07,
+      "loss": 0.9355,
+      "step": 13039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.295410176449736e-07,
+      "loss": 0.9326,
+      "step": 13040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.286762126389892e-07,
+      "loss": 0.9131,
+      "step": 13041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.278118391666934e-07,
+      "loss": 0.8978,
+      "step": 13042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.269478972687595e-07,
+      "loss": 0.8771,
+      "step": 13043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.260843869858392e-07,
+      "loss": 0.9175,
+      "step": 13044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.252213083585636e-07,
+      "loss": 0.8996,
+      "step": 13045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.243586614275412e-07,
+      "loss": 0.8932,
+      "step": 13046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.234964462333617e-07,
+      "loss": 0.943,
+      "step": 13047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.226346628165948e-07,
+      "loss": 0.8721,
+      "step": 13048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.217733112177895e-07,
+      "loss": 0.9135,
+      "step": 13049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.20912391477473e-07,
+      "loss": 0.9556,
+      "step": 13050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.200519036361565e-07,
+      "loss": 0.9174,
+      "step": 13051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.191918477343275e-07,
+      "loss": 0.9036,
+      "step": 13052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.183322238124536e-07,
+      "loss": 0.922,
+      "step": 13053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.174730319109802e-07,
+      "loss": 0.931,
+      "step": 13054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.166142720703351e-07,
+      "loss": 0.9172,
+      "step": 13055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.157559443309304e-07,
+      "loss": 0.8896,
+      "step": 13056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.148980487331459e-07,
+      "loss": 0.896,
+      "step": 13057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.140405853173471e-07,
+      "loss": 0.9432,
+      "step": 13058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.131835541238853e-07,
+      "loss": 0.8848,
+      "step": 13059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.123269551930835e-07,
+      "loss": 0.8829,
+      "step": 13060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.114707885652473e-07,
+      "loss": 0.9281,
+      "step": 13061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.106150542806601e-07,
+      "loss": 0.9697,
+      "step": 13062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.097597523795853e-07,
+      "loss": 0.8962,
+      "step": 13063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.089048829022728e-07,
+      "loss": 0.9065,
+      "step": 13064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.080504458889382e-07,
+      "loss": 0.9989,
+      "step": 13065
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.071964413797906e-07,
+      "loss": 0.9797,
+      "step": 13066
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.063428694150121e-07,
+      "loss": 0.8569,
+      "step": 13067
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.054897300347653e-07,
+      "loss": 0.8859,
+      "step": 13068
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.046370232791911e-07,
+      "loss": 0.8325,
+      "step": 13069
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.03784749188411e-07,
+      "loss": 0.8413,
+      "step": 13070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.029329078025294e-07,
+      "loss": 0.9011,
+      "step": 13071
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.020814991616288e-07,
+      "loss": 0.8808,
+      "step": 13072
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.012305233057627e-07,
+      "loss": 0.9466,
+      "step": 13073
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.00379980274979e-07,
+      "loss": 0.9579,
+      "step": 13074
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.995298701092957e-07,
+      "loss": 0.8968,
+      "step": 13075
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.986801928487109e-07,
+      "loss": 0.9439,
+      "step": 13076
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.978309485332059e-07,
+      "loss": 0.9546,
+      "step": 13077
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.969821372027387e-07,
+      "loss": 0.9047,
+      "step": 13078
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.961337588972484e-07,
+      "loss": 0.9634,
+      "step": 13079
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.952858136566566e-07,
+      "loss": 0.923,
+      "step": 13080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.944383015208535e-07,
+      "loss": 0.9043,
+      "step": 13081
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.935912225297238e-07,
+      "loss": 0.8989,
+      "step": 13082
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.927445767231212e-07,
+      "loss": 0.9455,
+      "step": 13083
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.918983641408839e-07,
+      "loss": 0.8989,
+      "step": 13084
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.910525848228279e-07,
+      "loss": 0.9191,
+      "step": 13085
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.902072388087479e-07,
+      "loss": 0.9372,
+      "step": 13086
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.893623261384242e-07,
+      "loss": 0.8832,
+      "step": 13087
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.885178468516042e-07,
+      "loss": 0.844,
+      "step": 13088
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.876738009880291e-07,
+      "loss": 0.9168,
+      "step": 13089
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.868301885874119e-07,
+      "loss": 0.9146,
+      "step": 13090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.859870096894462e-07,
+      "loss": 0.9259,
+      "step": 13091
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.851442643338059e-07,
+      "loss": 0.9418,
+      "step": 13092
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.843019525601425e-07,
+      "loss": 0.9525,
+      "step": 13093
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.834600744080922e-07,
+      "loss": 0.859,
+      "step": 13094
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.826186299172666e-07,
+      "loss": 0.9245,
+      "step": 13095
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.817776191272541e-07,
+      "loss": 0.9718,
+      "step": 13096
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.809370420776308e-07,
+      "loss": 0.9239,
+      "step": 13097
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.800968988079449e-07,
+      "loss": 0.9904,
+      "step": 13098
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.792571893577317e-07,
+      "loss": 0.9217,
+      "step": 13099
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.784179137664972e-07,
+      "loss": 0.871,
+      "step": 13100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.775790720737308e-07,
+      "loss": 0.9304,
+      "step": 13101
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.767406643189057e-07,
+      "loss": 0.9562,
+      "step": 13102
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.759026905414691e-07,
+      "loss": 0.8885,
+      "step": 13103
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.750651507808505e-07,
+      "loss": 0.844,
+      "step": 13104
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.742280450764583e-07,
+      "loss": 0.9622,
+      "step": 13105
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.73391373467679e-07,
+      "loss": 0.9468,
+      "step": 13106
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.725551359938821e-07,
+      "loss": 0.884,
+      "step": 13107
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.717193326944117e-07,
+      "loss": 0.9154,
+      "step": 13108
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.708839636085974e-07,
+      "loss": 0.9354,
+      "step": 13109
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.700490287757445e-07,
+      "loss": 0.9655,
+      "step": 13110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.692145282351393e-07,
+      "loss": 0.9479,
+      "step": 13111
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.68380462026046e-07,
+      "loss": 0.9439,
+      "step": 13112
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.675468301877098e-07,
+      "loss": 0.8387,
+      "step": 13113
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.66713632759356e-07,
+      "loss": 0.9001,
+      "step": 13114
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.658808697801867e-07,
+      "loss": 0.9755,
+      "step": 13115
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.65048541289386e-07,
+      "loss": 0.9636,
+      "step": 13116
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.642166473261192e-07,
+      "loss": 0.9003,
+      "step": 13117
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.633851879295285e-07,
+      "loss": 0.8635,
+      "step": 13118
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.625541631387334e-07,
+      "loss": 0.9141,
+      "step": 13119
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.617235729928385e-07,
+      "loss": 0.9936,
+      "step": 13120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.608934175309224e-07,
+      "loss": 0.9806,
+      "step": 13121
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.600636967920516e-07,
+      "loss": 0.8983,
+      "step": 13122
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.592344108152594e-07,
+      "loss": 0.8779,
+      "step": 13123
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.58405559639569e-07,
+      "loss": 0.8936,
+      "step": 13124
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.575771433039803e-07,
+      "loss": 0.9845,
+      "step": 13125
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.567491618474731e-07,
+      "loss": 0.8816,
+      "step": 13126
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.559216153090043e-07,
+      "loss": 0.8813,
+      "step": 13127
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.550945037275126e-07,
+      "loss": 0.8062,
+      "step": 13128
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.542678271419135e-07,
+      "loss": 0.9805,
+      "step": 13129
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.534415855911103e-07,
+      "loss": 0.8707,
+      "step": 13130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.526157791139732e-07,
+      "loss": 0.8981,
+      "step": 13131
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.517904077493621e-07,
+      "loss": 0.8707,
+      "step": 13132
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.509654715361114e-07,
+      "loss": 0.9542,
+      "step": 13133
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.501409705130369e-07,
+      "loss": 0.8698,
+      "step": 13134
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.493169047189331e-07,
+      "loss": 0.9244,
+      "step": 13135
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.484932741925732e-07,
+      "loss": 0.926,
+      "step": 13136
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.476700789727143e-07,
+      "loss": 0.9582,
+      "step": 13137
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.468473190980896e-07,
+      "loss": 0.9567,
+      "step": 13138
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.460249946074061e-07,
+      "loss": 0.9232,
+      "step": 13139
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.452031055393638e-07,
+      "loss": 0.9577,
+      "step": 13140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.443816519326297e-07,
+      "loss": 0.9367,
+      "step": 13141
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.435606338258572e-07,
+      "loss": 0.9585,
+      "step": 13142
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.427400512576777e-07,
+      "loss": 0.9356,
+      "step": 13143
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.41919904266698e-07,
+      "loss": 0.9689,
+      "step": 13144
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.411001928915151e-07,
+      "loss": 0.8837,
+      "step": 13145
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.402809171706904e-07,
+      "loss": 0.9675,
+      "step": 13146
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.394620771427785e-07,
+      "loss": 0.9414,
+      "step": 13147
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.386436728463054e-07,
+      "loss": 0.9468,
+      "step": 13148
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.378257043197812e-07,
+      "loss": 0.9698,
+      "step": 13149
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.370081716016908e-07,
+      "loss": 0.9134,
+      "step": 13150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.361910747305034e-07,
+      "loss": 0.8702,
+      "step": 13151
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.353744137446616e-07,
+      "loss": 0.9331,
+      "step": 13152
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.345581886825981e-07,
+      "loss": 0.9417,
+      "step": 13153
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.337423995827098e-07,
+      "loss": 0.9899,
+      "step": 13154
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329270464833893e-07,
+      "loss": 0.9733,
+      "step": 13155
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.321121294229971e-07,
+      "loss": 0.9294,
+      "step": 13156
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.312976484398781e-07,
+      "loss": 0.9036,
+      "step": 13157
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.30483603572355e-07,
+      "loss": 0.8965,
+      "step": 13158
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.296699948587294e-07,
+      "loss": 0.8812,
+      "step": 13159
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.288568223372871e-07,
+      "loss": 0.9144,
+      "step": 13160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.28044086046289e-07,
+      "loss": 0.9511,
+      "step": 13161
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.27231786023973e-07,
+      "loss": 0.9492,
+      "step": 13162
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.264199223085633e-07,
+      "loss": 0.9408,
+      "step": 13163
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.256084949382591e-07,
+      "loss": 0.9303,
+      "step": 13164
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.2479750395124e-07,
+      "loss": 0.903,
+      "step": 13165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.239869493856655e-07,
+      "loss": 0.9626,
+      "step": 13166
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.231768312796727e-07,
+      "loss": 1.009,
+      "step": 13167
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.223671496713813e-07,
+      "loss": 0.917,
+      "step": 13168
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.215579045988896e-07,
+      "loss": 0.9491,
+      "step": 13169
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.207490961002728e-07,
+      "loss": 0.8537,
+      "step": 13170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.19940724213588e-07,
+      "loss": 0.8638,
+      "step": 13171
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.191327889768717e-07,
+      "loss": 0.9167,
+      "step": 13172
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.183252904281379e-07,
+      "loss": 1.0272,
+      "step": 13173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.175182286053817e-07,
+      "loss": 0.9516,
+      "step": 13174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.167116035465782e-07,
+      "loss": 0.8654,
+      "step": 13175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.159054152896816e-07,
+      "loss": 0.9963,
+      "step": 13176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.150996638726237e-07,
+      "loss": 0.9145,
+      "step": 13177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.142943493333187e-07,
+      "loss": 0.9698,
+      "step": 13178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.134894717096552e-07,
+      "loss": 0.9371,
+      "step": 13179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.126850310395106e-07,
+      "loss": 0.8985,
+      "step": 13180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.118810273607313e-07,
+      "loss": 0.8934,
+      "step": 13181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.110774607111459e-07,
+      "loss": 0.8945,
+      "step": 13182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.102743311285698e-07,
+      "loss": 0.9177,
+      "step": 13183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.094716386507894e-07,
+      "loss": 0.978,
+      "step": 13184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.086693833155733e-07,
+      "loss": 0.9269,
+      "step": 13185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.078675651606692e-07,
+      "loss": 0.9208,
+      "step": 13186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.070661842238047e-07,
+      "loss": 0.8839,
+      "step": 13187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.062652405426906e-07,
+      "loss": 0.9267,
+      "step": 13188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.05464734155008e-07,
+      "loss": 0.912,
+      "step": 13189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.046646650984235e-07,
+      "loss": 0.9481,
+      "step": 13190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.038650334105846e-07,
+      "loss": 0.9103,
+      "step": 13191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.030658391291156e-07,
+      "loss": 0.9461,
+      "step": 13192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.0226708229162e-07,
+      "loss": 0.8897,
+      "step": 13193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.014687629356809e-07,
+      "loss": 0.9512,
+      "step": 13194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.006708810988605e-07,
+      "loss": 0.9044,
+      "step": 13195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.998734368187055e-07,
+      "loss": 0.8767,
+      "step": 13196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.990764301327313e-07,
+      "loss": 0.8136,
+      "step": 13197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.982798610784436e-07,
+      "loss": 0.9212,
+      "step": 13198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.974837296933212e-07,
+      "loss": 0.9176,
+      "step": 13199
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.966880360148243e-07,
+      "loss": 0.8716,
+      "step": 13200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.958927800803928e-07,
+      "loss": 0.9201,
+      "step": 13201
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.950979619274434e-07,
+      "loss": 0.9554,
+      "step": 13202
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.943035815933797e-07,
+      "loss": 0.8951,
+      "step": 13203
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.935096391155727e-07,
+      "loss": 0.9239,
+      "step": 13204
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.927161345313815e-07,
+      "loss": 0.9299,
+      "step": 13205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.919230678781441e-07,
+      "loss": 0.8495,
+      "step": 13206
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.911304391931762e-07,
+      "loss": 0.9181,
+      "step": 13207
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.90338248513771e-07,
+      "loss": 0.8961,
+      "step": 13208
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.895464958772047e-07,
+      "loss": 0.9761,
+      "step": 13209
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.887551813207294e-07,
+      "loss": 0.9315,
+      "step": 13210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.879643048815821e-07,
+      "loss": 0.8649,
+      "step": 13211
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.871738665969707e-07,
+      "loss": 1.0038,
+      "step": 13212
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.863838665040911e-07,
+      "loss": 0.8681,
+      "step": 13213
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.855943046401136e-07,
+      "loss": 0.955,
+      "step": 13214
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.848051810421885e-07,
+      "loss": 0.8605,
+      "step": 13215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.84016495747446e-07,
+      "loss": 0.8779,
+      "step": 13216
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.832282487929953e-07,
+      "loss": 0.8811,
+      "step": 13217
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.824404402159268e-07,
+      "loss": 0.9224,
+      "step": 13218
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.816530700533109e-07,
+      "loss": 0.8866,
+      "step": 13219
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.808661383421888e-07,
+      "loss": 0.9715,
+      "step": 13220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.800796451195935e-07,
+      "loss": 0.9343,
+      "step": 13221
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.792935904225285e-07,
+      "loss": 0.9301,
+      "step": 13222
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.785079742879818e-07,
+      "loss": 0.9103,
+      "step": 13223
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.777227967529166e-07,
+      "loss": 0.9251,
+      "step": 13224
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.769380578542772e-07,
+      "loss": 0.8813,
+      "step": 13225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.761537576289889e-07,
+      "loss": 0.8889,
+      "step": 13226
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.753698961139577e-07,
+      "loss": 0.9435,
+      "step": 13227
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.745864733460595e-07,
+      "loss": 0.9214,
+      "step": 13228
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.738034893621614e-07,
+      "loss": 0.9159,
+      "step": 13229
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.730209441991042e-07,
+      "loss": 0.9036,
+      "step": 13230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.722388378937073e-07,
+      "loss": 0.9095,
+      "step": 13231
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.71457170482771e-07,
+      "loss": 0.9171,
+      "step": 13232
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.706759420030729e-07,
+      "loss": 0.9189,
+      "step": 13233
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.698951524913755e-07,
+      "loss": 0.9622,
+      "step": 13234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.691148019844163e-07,
+      "loss": 0.9026,
+      "step": 13235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.683348905189102e-07,
+      "loss": 0.8912,
+      "step": 13236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.675554181315558e-07,
+      "loss": 0.9248,
+      "step": 13237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.667763848590292e-07,
+      "loss": 0.9204,
+      "step": 13238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.659977907379855e-07,
+      "loss": 0.918,
+      "step": 13239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.652196358050578e-07,
+      "loss": 0.7907,
+      "step": 13240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.644419200968633e-07,
+      "loss": 0.9231,
+      "step": 13241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.636646436499961e-07,
+      "loss": 0.9682,
+      "step": 13242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.628878065010225e-07,
+      "loss": 0.9302,
+      "step": 13243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.621114086865022e-07,
+      "loss": 0.8647,
+      "step": 13244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.613354502429614e-07,
+      "loss": 0.9858,
+      "step": 13245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.605599312069155e-07,
+      "loss": 0.9357,
+      "step": 13246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.597848516148519e-07,
+      "loss": 0.8654,
+      "step": 13247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.59010211503237e-07,
+      "loss": 0.8808,
+      "step": 13248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.582360109085261e-07,
+      "loss": 0.8925,
+      "step": 13249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.574622498671434e-07,
+      "loss": 0.8601,
+      "step": 13250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.566889284154965e-07,
+      "loss": 0.9222,
+      "step": 13251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.559160465899728e-07,
+      "loss": 0.8553,
+      "step": 13252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.551436044269377e-07,
+      "loss": 0.936,
+      "step": 13253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.543716019627378e-07,
+      "loss": 0.8931,
+      "step": 13254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.536000392336939e-07,
+      "loss": 0.9646,
+      "step": 13255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.528289162761147e-07,
+      "loss": 0.9579,
+      "step": 13256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.520582331262826e-07,
+      "loss": 0.876,
+      "step": 13257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.512879898204583e-07,
+      "loss": 0.8522,
+      "step": 13258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.50518186394885e-07,
+      "loss": 0.9836,
+      "step": 13259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.497488228857818e-07,
+      "loss": 0.8794,
+      "step": 13260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.489798993293539e-07,
+      "loss": 0.8871,
+      "step": 13261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.482114157617758e-07,
+      "loss": 0.9726,
+      "step": 13262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.474433722192075e-07,
+      "loss": 0.8869,
+      "step": 13263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.4667576873779e-07,
+      "loss": 0.8767,
+      "step": 13264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.459086053536401e-07,
+      "loss": 0.9252,
+      "step": 13265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.451418821028543e-07,
+      "loss": 0.9755,
+      "step": 13266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.443755990215084e-07,
+      "loss": 0.8683,
+      "step": 13267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.436097561456567e-07,
+      "loss": 0.9189,
+      "step": 13268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.428443535113382e-07,
+      "loss": 0.905,
+      "step": 13269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.420793911545642e-07,
+      "loss": 0.8493,
+      "step": 13270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.413148691113258e-07,
+      "loss": 0.9373,
+      "step": 13271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.405507874175998e-07,
+      "loss": 0.9769,
+      "step": 13272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.397871461093363e-07,
+      "loss": 0.982,
+      "step": 13273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.390239452224678e-07,
+      "loss": 0.9513,
+      "step": 13274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.382611847929032e-07,
+      "loss": 0.9275,
+      "step": 13275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.374988648565306e-07,
+      "loss": 0.87,
+      "step": 13276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.367369854492256e-07,
+      "loss": 0.9069,
+      "step": 13277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.359755466068284e-07,
+      "loss": 0.9097,
+      "step": 13278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.352145483651717e-07,
+      "loss": 0.8999,
+      "step": 13279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.344539907600622e-07,
+      "loss": 0.8936,
+      "step": 13280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.336938738272835e-07,
+      "loss": 0.8605,
+      "step": 13281
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.329341976026038e-07,
+      "loss": 0.9932,
+      "step": 13282
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.321749621217644e-07,
+      "loss": 0.8926,
+      "step": 13283
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.314161674204921e-07,
+      "loss": 0.929,
+      "step": 13284
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.306578135344921e-07,
+      "loss": 0.8958,
+      "step": 13285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.298999004994399e-07,
+      "loss": 0.9102,
+      "step": 13286
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.291424283510028e-07,
+      "loss": 0.9571,
+      "step": 13287
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.283853971248211e-07,
+      "loss": 0.8723,
+      "step": 13288
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.276288068565129e-07,
+      "loss": 0.9351,
+      "step": 13289
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.268726575816797e-07,
+      "loss": 0.9134,
+      "step": 13290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.261169493358976e-07,
+      "loss": 0.9529,
+      "step": 13291
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.253616821547281e-07,
+      "loss": 0.9526,
+      "step": 13292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.246068560737062e-07,
+      "loss": 0.9078,
+      "step": 13293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.2385247112835e-07,
+      "loss": 0.876,
+      "step": 13294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230985273541523e-07,
+      "loss": 0.9276,
+      "step": 13295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.223450247865914e-07,
+      "loss": 0.9737,
+      "step": 13296
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.21591963461119e-07,
+      "loss": 0.8808,
+      "step": 13297
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.208393434131677e-07,
+      "loss": 0.8519,
+      "step": 13298
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.200871646781537e-07,
+      "loss": 0.9087,
+      "step": 13299
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.193354272914687e-07,
+      "loss": 0.9677,
+      "step": 13300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.185841312884788e-07,
+      "loss": 0.8687,
+      "step": 13301
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.178332767045392e-07,
+      "loss": 0.9557,
+      "step": 13302
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.170828635749771e-07,
+      "loss": 0.9566,
+      "step": 13303
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.16332891935103e-07,
+      "loss": 0.8994,
+      "step": 13304
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.155833618202045e-07,
+      "loss": 0.9217,
+      "step": 13305
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.148342732655466e-07,
+      "loss": 0.8985,
+      "step": 13306
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.140856263063788e-07,
+      "loss": 0.9133,
+      "step": 13307
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.133374209779275e-07,
+      "loss": 0.887,
+      "step": 13308
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.125896573153933e-07,
+      "loss": 0.8909,
+      "step": 13309
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.118423353539638e-07,
+      "loss": 0.9616,
+      "step": 13310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.110954551288007e-07,
+      "loss": 1.0014,
+      "step": 13311
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.103490166750481e-07,
+      "loss": 0.9169,
+      "step": 13312
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.096030200278269e-07,
+      "loss": 0.862,
+      "step": 13313
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.088574652222368e-07,
+      "loss": 0.9707,
+      "step": 13314
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.081123522933607e-07,
+      "loss": 0.9132,
+      "step": 13315
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.073676812762574e-07,
+      "loss": 0.9802,
+      "step": 13316
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.066234522059644e-07,
+      "loss": 0.9141,
+      "step": 13317
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.058796651174992e-07,
+      "loss": 0.9361,
+      "step": 13318
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.051363200458605e-07,
+      "loss": 0.9078,
+      "step": 13319
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.043934170260235e-07,
+      "loss": 0.9248,
+      "step": 13320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.036509560929427e-07,
+      "loss": 0.9444,
+      "step": 13321
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.029089372815544e-07,
+      "loss": 0.9069,
+      "step": 13322
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.02167360626773e-07,
+      "loss": 0.8746,
+      "step": 13323
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.014262261634896e-07,
+      "loss": 0.8749,
+      "step": 13324
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.006855339265771e-07,
+      "loss": 0.9767,
+      "step": 13325
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.999452839508868e-07,
+      "loss": 0.8924,
+      "step": 13326
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.992054762712518e-07,
+      "loss": 0.947,
+      "step": 13327
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.984661109224776e-07,
+      "loss": 0.9332,
+      "step": 13328
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.977271879393542e-07,
+      "loss": 0.9351,
+      "step": 13329
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.969887073566527e-07,
+      "loss": 0.8787,
+      "step": 13330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.962506692091186e-07,
+      "loss": 1.0157,
+      "step": 13331
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.955130735314785e-07,
+      "loss": 0.9519,
+      "step": 13332
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.94775920358438e-07,
+      "loss": 0.9015,
+      "step": 13333
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.940392097246805e-07,
+      "loss": 0.882,
+      "step": 13334
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.93302941664875e-07,
+      "loss": 0.857,
+      "step": 13335
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.925671162136571e-07,
+      "loss": 0.8952,
+      "step": 13336
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.918317334056567e-07,
+      "loss": 0.9714,
+      "step": 13337
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.910967932754707e-07,
+      "loss": 0.8947,
+      "step": 13338
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.903622958576827e-07,
+      "loss": 0.8884,
+      "step": 13339
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.896282411868504e-07,
+      "loss": 0.8587,
+      "step": 13340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.88894629297514e-07,
+      "loss": 0.9913,
+      "step": 13341
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.881614602241903e-07,
+      "loss": 0.9956,
+      "step": 13342
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.874287340013807e-07,
+      "loss": 0.8741,
+      "step": 13343
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.866964506635564e-07,
+      "loss": 0.9699,
+      "step": 13344
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.859646102451777e-07,
+      "loss": 0.9468,
+      "step": 13345
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.85233212780677e-07,
+      "loss": 0.9202,
+      "step": 13346
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.845022583044701e-07,
+      "loss": 0.8692,
+      "step": 13347
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.837717468509485e-07,
+      "loss": 0.8873,
+      "step": 13348
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.830416784544834e-07,
+      "loss": 0.88,
+      "step": 13349
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.82312053149432e-07,
+      "loss": 0.8667,
+      "step": 13350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.815828709701199e-07,
+      "loss": 0.8902,
+      "step": 13351
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.808541319508565e-07,
+      "loss": 0.9287,
+      "step": 13352
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.801258361259332e-07,
+      "loss": 0.8777,
+      "step": 13353
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.793979835296182e-07,
+      "loss": 0.8983,
+      "step": 13354
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.786705741961585e-07,
+      "loss": 0.9441,
+      "step": 13355
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7794360815978e-07,
+      "loss": 0.9413,
+      "step": 13356
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.772170854546865e-07,
+      "loss": 0.9673,
+      "step": 13357
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.764910061150675e-07,
+      "loss": 0.9544,
+      "step": 13358
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.75765370175082e-07,
+      "loss": 0.9613,
+      "step": 13359
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.750401776688752e-07,
+      "loss": 0.8901,
+      "step": 13360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.743154286305697e-07,
+      "loss": 0.9406,
+      "step": 13361
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.735911230942659e-07,
+      "loss": 0.9177,
+      "step": 13362
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.728672610940433e-07,
+      "loss": 0.9077,
+      "step": 13363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.721438426639626e-07,
+      "loss": 0.9651,
+      "step": 13364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.714208678380628e-07,
+      "loss": 0.8255,
+      "step": 13365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.706983366503627e-07,
+      "loss": 0.9042,
+      "step": 13366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.699762491348548e-07,
+      "loss": 0.9567,
+      "step": 13367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.692546053255188e-07,
+      "loss": 0.9357,
+      "step": 13368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.685334052563096e-07,
+      "loss": 0.9471,
+      "step": 13369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.678126489611602e-07,
+      "loss": 1.0017,
+      "step": 13370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.670923364739855e-07,
+      "loss": 0.942,
+      "step": 13371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.663724678286753e-07,
+      "loss": 0.9334,
+      "step": 13372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.656530430591045e-07,
+      "loss": 0.994,
+      "step": 13373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.64934062199124e-07,
+      "loss": 0.8278,
+      "step": 13374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.642155252825588e-07,
+      "loss": 0.9007,
+      "step": 13375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.634974323432219e-07,
+      "loss": 0.9388,
+      "step": 13376
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.627797834149018e-07,
+      "loss": 0.8688,
+      "step": 13377
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.620625785313639e-07,
+      "loss": 0.9303,
+      "step": 13378
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.613458177263553e-07,
+      "loss": 0.9064,
+      "step": 13379
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.606295010335994e-07,
+      "loss": 0.8635,
+      "step": 13380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.599136284868034e-07,
+      "loss": 0.9741,
+      "step": 13381
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.591982001196505e-07,
+      "loss": 0.9761,
+      "step": 13382
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.584832159658027e-07,
+      "loss": 0.902,
+      "step": 13383
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.577686760589007e-07,
+      "loss": 0.8895,
+      "step": 13384
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.570545804325678e-07,
+      "loss": 0.9492,
+      "step": 13385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.563409291204025e-07,
+      "loss": 0.8868,
+      "step": 13386
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.556277221559825e-07,
+      "loss": 0.9403,
+      "step": 13387
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.549149595728698e-07,
+      "loss": 1.0251,
+      "step": 13388
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.542026414045998e-07,
+      "loss": 0.9519,
+      "step": 13389
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.534907676846857e-07,
+      "loss": 0.9382,
+      "step": 13390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.527793384466274e-07,
+      "loss": 0.9613,
+      "step": 13391
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.520683537238969e-07,
+      "loss": 0.8832,
+      "step": 13392
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.513578135499508e-07,
+      "loss": 0.8306,
+      "step": 13393
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.506477179582181e-07,
+      "loss": 0.9361,
+      "step": 13394
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.499380669821108e-07,
+      "loss": 0.8449,
+      "step": 13395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.492288606550223e-07,
+      "loss": 0.9169,
+      "step": 13396
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.485200990103223e-07,
+      "loss": 0.9021,
+      "step": 13397
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.478117820813578e-07,
+      "loss": 0.9561,
+      "step": 13398
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.471039099014585e-07,
+      "loss": 0.8647,
+      "step": 13399
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.463964825039291e-07,
+      "loss": 0.843,
+      "step": 13400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.456894999220597e-07,
+      "loss": 0.9852,
+      "step": 13401
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.449829621891123e-07,
+      "loss": 0.9036,
+      "step": 13402
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.442768693383327e-07,
+      "loss": 0.9312,
+      "step": 13403
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.435712214029442e-07,
+      "loss": 0.9834,
+      "step": 13404
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.428660184161505e-07,
+      "loss": 0.8695,
+      "step": 13405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.421612604111304e-07,
+      "loss": 0.8713,
+      "step": 13406
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.414569474210452e-07,
+      "loss": 0.94,
+      "step": 13407
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.407530794790383e-07,
+      "loss": 0.9657,
+      "step": 13408
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.400496566182245e-07,
+      "loss": 0.9488,
+      "step": 13409
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.393466788717005e-07,
+      "loss": 0.8901,
+      "step": 13410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.386441462725478e-07,
+      "loss": 0.8455,
+      "step": 13411
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.379420588538187e-07,
+      "loss": 0.8442,
+      "step": 13412
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.372404166485501e-07,
+      "loss": 0.9372,
+      "step": 13413
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.365392196897545e-07,
+      "loss": 0.9711,
+      "step": 13414
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.358384680104256e-07,
+      "loss": 0.9287,
+      "step": 13415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.351381616435369e-07,
+      "loss": 0.9633,
+      "step": 13416
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.344383006220377e-07,
+      "loss": 0.8534,
+      "step": 13417
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.337388849788572e-07,
+      "loss": 0.8926,
+      "step": 13418
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.330399147469079e-07,
+      "loss": 0.987,
+      "step": 13419
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.323413899590757e-07,
+      "loss": 0.9254,
+      "step": 13420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.316433106482288e-07,
+      "loss": 1.0071,
+      "step": 13421
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.309456768472143e-07,
+      "loss": 0.9153,
+      "step": 13422
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.302484885888548e-07,
+      "loss": 0.8809,
+      "step": 13423
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.295517459059584e-07,
+      "loss": 0.9381,
+      "step": 13424
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.288554488313058e-07,
+      "loss": 0.9115,
+      "step": 13425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.281595973976617e-07,
+      "loss": 0.8826,
+      "step": 13426
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.274641916377655e-07,
+      "loss": 0.884,
+      "step": 13427
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.267692315843398e-07,
+      "loss": 0.8788,
+      "step": 13428
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.26074717270083e-07,
+      "loss": 0.8977,
+      "step": 13429
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253806487276724e-07,
+      "loss": 0.9117,
+      "step": 13430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.246870259897685e-07,
+      "loss": 0.9512,
+      "step": 13431
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.239938490890085e-07,
+      "loss": 0.9469,
+      "step": 13432
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.233011180580028e-07,
+      "loss": 0.9299,
+      "step": 13433
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.226088329293522e-07,
+      "loss": 0.8915,
+      "step": 13434
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.219169937356272e-07,
+      "loss": 0.8688,
+      "step": 13435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.212256005093807e-07,
+      "loss": 0.9199,
+      "step": 13436
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.205346532831457e-07,
+      "loss": 0.95,
+      "step": 13437
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.198441520894315e-07,
+      "loss": 0.9045,
+      "step": 13438
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.19154096960729e-07,
+      "loss": 0.9854,
+      "step": 13439
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.184644879295075e-07,
+      "loss": 0.9205,
+      "step": 13440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.177753250282136e-07,
+      "loss": 0.9434,
+      "step": 13441
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.170866082892745e-07,
+      "loss": 0.9648,
+      "step": 13442
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.163983377450965e-07,
+      "loss": 0.8299,
+      "step": 13443
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.157105134280638e-07,
+      "loss": 0.9374,
+      "step": 13444
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.150231353705392e-07,
+      "loss": 0.9139,
+      "step": 13445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.143362036048683e-07,
+      "loss": 0.8967,
+      "step": 13446
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.136497181633726e-07,
+      "loss": 0.9183,
+      "step": 13447
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.129636790783499e-07,
+      "loss": 0.898,
+      "step": 13448
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.122780863820831e-07,
+      "loss": 0.9369,
+      "step": 13449
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.115929401068298e-07,
+      "loss": 0.8696,
+      "step": 13450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.109082402848287e-07,
+      "loss": 0.9397,
+      "step": 13451
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.10223986948295e-07,
+      "loss": 0.9179,
+      "step": 13452
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.095401801294253e-07,
+      "loss": 0.9122,
+      "step": 13453
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08856819860395e-07,
+      "loss": 0.9408,
+      "step": 13454
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.081739061733592e-07,
+      "loss": 0.9124,
+      "step": 13455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.074914391004471e-07,
+      "loss": 0.8798,
+      "step": 13456
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.068094186737726e-07,
+      "loss": 0.9394,
+      "step": 13457
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.061278449254259e-07,
+      "loss": 0.9236,
+      "step": 13458
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.054467178874789e-07,
+      "loss": 0.9143,
+      "step": 13459
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.047660375919772e-07,
+      "loss": 0.9919,
+      "step": 13460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.040858040709495e-07,
+      "loss": 0.9068,
+      "step": 13461
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.034060173564038e-07,
+      "loss": 0.9115,
+      "step": 13462
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.027266774803241e-07,
+      "loss": 0.8779,
+      "step": 13463
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.020477844746762e-07,
+      "loss": 0.8951,
+      "step": 13464
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.013693383714035e-07,
+      "loss": 0.8786,
+      "step": 13465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.006913392024282e-07,
+      "loss": 0.8718,
+      "step": 13466
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.000137869996524e-07,
+      "loss": 0.9273,
+      "step": 13467
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.993366817949541e-07,
+      "loss": 0.8528,
+      "step": 13468
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.986600236201966e-07,
+      "loss": 0.9079,
+      "step": 13469
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.979838125072167e-07,
+      "loss": 0.905,
+      "step": 13470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.973080484878323e-07,
+      "loss": 0.9307,
+      "step": 13471
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.966327315938391e-07,
+      "loss": 0.8597,
+      "step": 13472
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.959578618570105e-07,
+      "loss": 0.9073,
+      "step": 13473
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.952834393091066e-07,
+      "loss": 0.9356,
+      "step": 13474
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.946094639818555e-07,
+      "loss": 0.8721,
+      "step": 13475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.939359359069684e-07,
+      "loss": 0.9223,
+      "step": 13476
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.932628551161422e-07,
+      "loss": 0.9293,
+      "step": 13477
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.925902216410427e-07,
+      "loss": 0.9502,
+      "step": 13478
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919180355133202e-07,
+      "loss": 0.9516,
+      "step": 13479
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.912462967646025e-07,
+      "loss": 0.9787,
+      "step": 13480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.905750054264957e-07,
+      "loss": 0.9542,
+      "step": 13481
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.899041615305889e-07,
+      "loss": 0.9122,
+      "step": 13482
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.892337651084433e-07,
+      "loss": 0.9245,
+      "step": 13483
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.885638161916051e-07,
+      "loss": 0.8767,
+      "step": 13484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.878943148115956e-07,
+      "loss": 0.9508,
+      "step": 13485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.872252609999173e-07,
+      "loss": 0.9951,
+      "step": 13486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.865566547880518e-07,
+      "loss": 0.8833,
+      "step": 13487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.858884962074573e-07,
+      "loss": 0.8569,
+      "step": 13488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.852207852895707e-07,
+      "loss": 0.9535,
+      "step": 13489
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.845535220658138e-07,
+      "loss": 0.9865,
+      "step": 13490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.83886706567579e-07,
+      "loss": 0.9377,
+      "step": 13491
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.832203388262435e-07,
+      "loss": 0.9203,
+      "step": 13492
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.825544188731623e-07,
+      "loss": 0.9179,
+      "step": 13493
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.81888946739667e-07,
+      "loss": 0.9747,
+      "step": 13494
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.812239224570703e-07,
+      "loss": 0.9435,
+      "step": 13495
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.805593460566615e-07,
+      "loss": 0.8779,
+      "step": 13496
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.798952175697146e-07,
+      "loss": 0.949,
+      "step": 13497
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.79231537027477e-07,
+      "loss": 0.8144,
+      "step": 13498
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785683044611733e-07,
+      "loss": 0.9417,
+      "step": 13499
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.779055199020144e-07,
+      "loss": 0.9131,
+      "step": 13500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.772431833811841e-07,
+      "loss": 0.9717,
+      "step": 13501
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.765812949298476e-07,
+      "loss": 0.9715,
+      "step": 13502
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.759198545791477e-07,
+      "loss": 0.9148,
+      "step": 13503
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7525886236020503e-07,
+      "loss": 1.0062,
+      "step": 13504
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7459831830412693e-07,
+      "loss": 0.9094,
+      "step": 13505
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.739382224419864e-07,
+      "loss": 0.8977,
+      "step": 13506
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.732785748048474e-07,
+      "loss": 0.9086,
+      "step": 13507
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7261937542374625e-07,
+      "loss": 0.8864,
+      "step": 13508
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7196062432970036e-07,
+      "loss": 0.9189,
+      "step": 13509
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.71302321553706e-07,
+      "loss": 0.883,
+      "step": 13510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.706444671267352e-07,
+      "loss": 0.844,
+      "step": 13511
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.699870610797463e-07,
+      "loss": 0.929,
+      "step": 13512
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6933010344367034e-07,
+      "loss": 0.9361,
+      "step": 13513
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.686735942494147e-07,
+      "loss": 0.8819,
+      "step": 13514
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.680175335278747e-07,
+      "loss": 0.9503,
+      "step": 13515
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.67361921309919e-07,
+      "loss": 0.9836,
+      "step": 13516
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6670675762639285e-07,
+      "loss": 0.9991,
+      "step": 13517
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.660520425081272e-07,
+      "loss": 0.9112,
+      "step": 13518
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6539777598592295e-07,
+      "loss": 0.9104,
+      "step": 13519
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6474395809057104e-07,
+      "loss": 0.9197,
+      "step": 13520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.640905888528313e-07,
+      "loss": 0.9356,
+      "step": 13521
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6343766830344694e-07,
+      "loss": 0.9534,
+      "step": 13522
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6278519647314114e-07,
+      "loss": 0.9065,
+      "step": 13523
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.621331733926127e-07,
+      "loss": 0.8288,
+      "step": 13524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.614815990925414e-07,
+      "loss": 0.9693,
+      "step": 13525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6083047360358514e-07,
+      "loss": 0.9326,
+      "step": 13526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.601797969563804e-07,
+      "loss": 0.8804,
+      "step": 13527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.595295691815449e-07,
+      "loss": 0.9621,
+      "step": 13528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.588797903096731e-07,
+      "loss": 0.9905,
+      "step": 13529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.582304603713372e-07,
+      "loss": 0.885,
+      "step": 13530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.575815793970917e-07,
+      "loss": 0.9544,
+      "step": 13531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.569331474174665e-07,
+      "loss": 0.9274,
+      "step": 13532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.56285164462974e-07,
+      "loss": 0.9681,
+      "step": 13533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.556376305640997e-07,
+      "loss": 0.8895,
+      "step": 13534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5499054575131483e-07,
+      "loss": 0.8899,
+      "step": 13535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5434391005506617e-07,
+      "loss": 0.9868,
+      "step": 13536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5369772350577935e-07,
+      "loss": 0.8753,
+      "step": 13537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5305198613385784e-07,
+      "loss": 0.9187,
+      "step": 13538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5240669796968506e-07,
+      "loss": 0.9123,
+      "step": 13539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.517618590436268e-07,
+      "loss": 0.9499,
+      "step": 13540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.51117469386021e-07,
+      "loss": 0.8688,
+      "step": 13541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5047352902718666e-07,
+      "loss": 0.9899,
+      "step": 13542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.498300379974263e-07,
+      "loss": 0.9878,
+      "step": 13543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.491869963270168e-07,
+      "loss": 0.8862,
+      "step": 13544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.485444040462139e-07,
+      "loss": 0.9088,
+      "step": 13545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.479022611852546e-07,
+      "loss": 0.9619,
+      "step": 13546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.472605677743502e-07,
+      "loss": 1.0117,
+      "step": 13547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4661932384369886e-07,
+      "loss": 0.8793,
+      "step": 13548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4597852942346754e-07,
+      "loss": 0.9342,
+      "step": 13549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4533818454381095e-07,
+      "loss": 0.9914,
+      "step": 13550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4469828923485724e-07,
+      "loss": 0.8932,
+      "step": 13551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.440588435267157e-07,
+      "loss": 0.875,
+      "step": 13552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.434198474494744e-07,
+      "loss": 0.9637,
+      "step": 13553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.427813010331972e-07,
+      "loss": 0.9355,
+      "step": 13554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.421432043079321e-07,
+      "loss": 0.9256,
+      "step": 13555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4150555730370305e-07,
+      "loss": 0.9212,
+      "step": 13556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4086836005050925e-07,
+      "loss": 0.9027,
+      "step": 13557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.402316125783379e-07,
+      "loss": 0.9307,
+      "step": 13558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3959531491714503e-07,
+      "loss": 0.9641,
+      "step": 13559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3895946709687334e-07,
+      "loss": 0.9364,
+      "step": 13560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3832406914743886e-07,
+      "loss": 0.9,
+      "step": 13561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.376891210987377e-07,
+      "loss": 0.9548,
+      "step": 13562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.370546229806505e-07,
+      "loss": 0.891,
+      "step": 13563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3642057482302656e-07,
+      "loss": 0.9459,
+      "step": 13564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.35786976655701e-07,
+      "loss": 0.9838,
+      "step": 13565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3515382850848885e-07,
+      "loss": 0.9833,
+      "step": 13566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.345211304111785e-07,
+      "loss": 0.9563,
+      "step": 13567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3388888239354167e-07,
+      "loss": 0.9684,
+      "step": 13568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3325708448532565e-07,
+      "loss": 0.884,
+      "step": 13569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3262573671625786e-07,
+      "loss": 0.8981,
+      "step": 13570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.319948391160489e-07,
+      "loss": 0.909,
+      "step": 13571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.313643917143773e-07,
+      "loss": 0.9161,
+      "step": 13572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.307343945409126e-07,
+      "loss": 0.9633,
+      "step": 13573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.301048476252967e-07,
+      "loss": 0.9454,
+      "step": 13574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.294757509971492e-07,
+      "loss": 0.9815,
+      "step": 13575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2884710468607314e-07,
+      "loss": 0.9252,
+      "step": 13576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.282189087216448e-07,
+      "loss": 0.9574,
+      "step": 13577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.275911631334262e-07,
+      "loss": 0.9154,
+      "step": 13578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.269638679509547e-07,
+      "loss": 0.8872,
+      "step": 13579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2633702320374006e-07,
+      "loss": 0.938,
+      "step": 13580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2571062892128313e-07,
+      "loss": 0.9614,
+      "step": 13581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.250846851330537e-07,
+      "loss": 0.8688,
+      "step": 13582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2445919186850704e-07,
+      "loss": 0.8443,
+      "step": 13583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.238341491570708e-07,
+      "loss": 0.9391,
+      "step": 13584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.232095570281558e-07,
+      "loss": 0.8432,
+      "step": 13585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2258541551115306e-07,
+      "loss": 0.9547,
+      "step": 13586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.219617246354279e-07,
+      "loss": 0.9072,
+      "step": 13587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2133848443032587e-07,
+      "loss": 0.9038,
+      "step": 13588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.207156949251745e-07,
+      "loss": 1.0195,
+      "step": 13589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2009335614927484e-07,
+      "loss": 0.9884,
+      "step": 13590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.194714681319112e-07,
+      "loss": 0.9059,
+      "step": 13591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1885003090234354e-07,
+      "loss": 0.8419,
+      "step": 13592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1822904448981293e-07,
+      "loss": 0.834,
+      "step": 13593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.176085089235393e-07,
+      "loss": 0.8714,
+      "step": 13594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.169884242327205e-07,
+      "loss": 0.8859,
+      "step": 13595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.163687904465308e-07,
+      "loss": 0.9002,
+      "step": 13596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1574960759412474e-07,
+      "loss": 0.9335,
+      "step": 13597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.151308757046424e-07,
+      "loss": 1.0147,
+      "step": 13598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1451259480719043e-07,
+      "loss": 0.9853,
+      "step": 13599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1389476493086224e-07,
+      "loss": 0.9636,
+      "step": 13600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.132773861047301e-07,
+      "loss": 0.9534,
+      "step": 13601
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1266045835784194e-07,
+      "loss": 0.901,
+      "step": 13602
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1204398171922453e-07,
+      "loss": 0.9263,
+      "step": 13603
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.114279562178858e-07,
+      "loss": 0.9602,
+      "step": 13604
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1081238188281137e-07,
+      "loss": 0.916,
+      "step": 13605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.10197258742967e-07,
+      "loss": 0.939,
+      "step": 13606
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.095825868272929e-07,
+      "loss": 0.9485,
+      "step": 13607
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.089683661647115e-07,
+      "loss": 0.9677,
+      "step": 13608
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0835459678412627e-07,
+      "loss": 0.9157,
+      "step": 13609
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0774127871441305e-07,
+      "loss": 0.9428,
+      "step": 13610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.071284119844321e-07,
+      "loss": 0.9403,
+      "step": 13611
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.065159966230192e-07,
+      "loss": 0.9568,
+      "step": 13612
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.059040326589914e-07,
+      "loss": 0.9914,
+      "step": 13613
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0529252012114216e-07,
+      "loss": 0.914,
+      "step": 13614
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0468145903824194e-07,
+      "loss": 0.8851,
+      "step": 13615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0407084943904883e-07,
+      "loss": 0.9055,
+      "step": 13616
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0346069135228874e-07,
+      "loss": 0.8799,
+      "step": 13617
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0285098480667305e-07,
+      "loss": 0.9358,
+      "step": 13618
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0224172983089004e-07,
+      "loss": 0.838,
+      "step": 13619
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.016329264536034e-07,
+      "loss": 0.8807,
+      "step": 13620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.010245747034658e-07,
+      "loss": 0.9096,
+      "step": 13621
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.004166746090954e-07,
+      "loss": 0.8804,
+      "step": 13622
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.99809226199096e-07,
+      "loss": 0.8748,
+      "step": 13623
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.992022295020526e-07,
+      "loss": 0.8652,
+      "step": 13624
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.985956845465255e-07,
+      "loss": 0.9336,
+      "step": 13625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9798959136105206e-07,
+      "loss": 0.9188,
+      "step": 13626
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9738394997415164e-07,
+      "loss": 0.8593,
+      "step": 13627
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9677876041431916e-07,
+      "loss": 1.0014,
+      "step": 13628
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.961740227100352e-07,
+      "loss": 0.9551,
+      "step": 13629
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.955697368897482e-07,
+      "loss": 0.9178,
+      "step": 13630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9496590298189643e-07,
+      "loss": 0.908,
+      "step": 13631
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.943625210148894e-07,
+      "loss": 1.0258,
+      "step": 13632
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.937595910171177e-07,
+      "loss": 0.937,
+      "step": 13633
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.931571130169509e-07,
+      "loss": 0.911,
+      "step": 13634
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.925550870427353e-07,
+      "loss": 0.8762,
+      "step": 13635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.919535131228014e-07,
+      "loss": 0.8695,
+      "step": 13636
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9135239128545446e-07,
+      "loss": 0.9125,
+      "step": 13637
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.90751721558974e-07,
+      "loss": 0.9601,
+      "step": 13638
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.901515039716286e-07,
+      "loss": 0.8686,
+      "step": 13639
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.895517385516567e-07,
+      "loss": 0.998,
+      "step": 13640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.889524253272792e-07,
+      "loss": 0.9945,
+      "step": 13641
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.883535643266967e-07,
+      "loss": 0.982,
+      "step": 13642
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.877551555780856e-07,
+      "loss": 0.8987,
+      "step": 13643
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8715719910960345e-07,
+      "loss": 0.9209,
+      "step": 13644
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.865596949493866e-07,
+      "loss": 0.8978,
+      "step": 13645
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.859626431255448e-07,
+      "loss": 0.9903,
+      "step": 13646
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.853660436661755e-07,
+      "loss": 0.9272,
+      "step": 13647
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.847698965993485e-07,
+      "loss": 0.8903,
+      "step": 13648
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8417420195311474e-07,
+      "loss": 0.9039,
+      "step": 13649
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8357895975550175e-07,
+      "loss": 0.9507,
+      "step": 13650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8298417003451605e-07,
+      "loss": 0.9855,
+      "step": 13651
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.823898328181475e-07,
+      "loss": 0.9254,
+      "step": 13652
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.817959481343603e-07,
+      "loss": 0.9156,
+      "step": 13653
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.812025160110966e-07,
+      "loss": 0.9144,
+      "step": 13654
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.806095364762785e-07,
+      "loss": 0.9391,
+      "step": 13655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.800170095578093e-07,
+      "loss": 0.9035,
+      "step": 13656
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7942493528356773e-07,
+      "loss": 0.8701,
+      "step": 13657
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.788333136814115e-07,
+      "loss": 0.9071,
+      "step": 13658
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7824214477918066e-07,
+      "loss": 0.9471,
+      "step": 13659
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.776514286046895e-07,
+      "loss": 0.892,
+      "step": 13660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7706116518573035e-07,
+      "loss": 0.9634,
+      "step": 13661
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.764713545500798e-07,
+      "loss": 0.8695,
+      "step": 13662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7588199672548784e-07,
+      "loss": 0.849,
+      "step": 13663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.752930917396891e-07,
+      "loss": 1.0157,
+      "step": 13664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7470463962038793e-07,
+      "loss": 0.9283,
+      "step": 13665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.741166403952745e-07,
+      "loss": 0.8604,
+      "step": 13666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7352909409201666e-07,
+      "loss": 0.9308,
+      "step": 13667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7294200073826005e-07,
+      "loss": 0.9952,
+      "step": 13668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.72355360361627e-07,
+      "loss": 0.9161,
+      "step": 13669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7176917298972217e-07,
+      "loss": 0.9459,
+      "step": 13670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7118343865012563e-07,
+      "loss": 0.9182,
+      "step": 13671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.705981573703987e-07,
+      "loss": 0.9067,
+      "step": 13672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7001332917807943e-07,
+      "loss": 0.9131,
+      "step": 13673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.694289541006857e-07,
+      "loss": 0.9823,
+      "step": 13674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6884503216571554e-07,
+      "loss": 0.9328,
+      "step": 13675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6826156340064256e-07,
+      "loss": 0.9192,
+      "step": 13676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6767854783291925e-07,
+      "loss": 0.9301,
+      "step": 13677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.670959854899803e-07,
+      "loss": 0.9146,
+      "step": 13678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6651387639923484e-07,
+      "loss": 0.8826,
+      "step": 13679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.659322205880733e-07,
+      "loss": 0.9111,
+      "step": 13680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6535101808386263e-07,
+      "loss": 0.9025,
+      "step": 13681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.647702689139532e-07,
+      "loss": 0.9567,
+      "step": 13682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.641899731056686e-07,
+      "loss": 0.9286,
+      "step": 13683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.636101306863127e-07,
+      "loss": 0.9432,
+      "step": 13684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.630307416831691e-07,
+      "loss": 0.956,
+      "step": 13685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.624518061234983e-07,
+      "loss": 0.9368,
+      "step": 13686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.618733240345451e-07,
+      "loss": 0.9741,
+      "step": 13687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6129529544352336e-07,
+      "loss": 0.9763,
+      "step": 13688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6071772037763244e-07,
+      "loss": 0.8603,
+      "step": 13689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.601405988640494e-07,
+      "loss": 0.8575,
+      "step": 13690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5956393092992924e-07,
+      "loss": 0.9427,
+      "step": 13691
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5898771660240474e-07,
+      "loss": 0.9043,
+      "step": 13692
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.584119559085897e-07,
+      "loss": 0.9733,
+      "step": 13693
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5783664887557134e-07,
+      "loss": 0.9375,
+      "step": 13694
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5726179553042586e-07,
+      "loss": 0.9563,
+      "step": 13695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.566873959001949e-07,
+      "loss": 0.8755,
+      "step": 13696
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5611345001191013e-07,
+      "loss": 0.8877,
+      "step": 13697
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5553995789257444e-07,
+      "loss": 0.9505,
+      "step": 13698
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5496691956917295e-07,
+      "loss": 0.9425,
+      "step": 13699
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5439433506866847e-07,
+      "loss": 0.9473,
+      "step": 13700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5382220441800174e-07,
+      "loss": 0.9288,
+      "step": 13701
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.532505276440945e-07,
+      "loss": 0.8825,
+      "step": 13702
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5267930477384636e-07,
+      "loss": 0.9068,
+      "step": 13703
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.521085358341303e-07,
+      "loss": 0.9459,
+      "step": 13704
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.51538220851807e-07,
+      "loss": 0.899,
+      "step": 13705
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.509683598537095e-07,
+      "loss": 0.9397,
+      "step": 13706
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5039895286665184e-07,
+      "loss": 0.9664,
+      "step": 13707
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.498299999174248e-07,
+      "loss": 0.884,
+      "step": 13708
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.492615010327982e-07,
+      "loss": 0.8777,
+      "step": 13709
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.48693456239525e-07,
+      "loss": 0.9006,
+      "step": 13710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4812586556432936e-07,
+      "loss": 0.9547,
+      "step": 13711
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.475587290339211e-07,
+      "loss": 0.9541,
+      "step": 13712
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.469920466749821e-07,
+      "loss": 0.906,
+      "step": 13713
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.464258185141789e-07,
+      "loss": 0.8126,
+      "step": 13714
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4586004457815235e-07,
+      "loss": 0.9518,
+      "step": 13715
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4529472489352455e-07,
+      "loss": 0.9012,
+      "step": 13716
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.447298594868931e-07,
+      "loss": 0.9735,
+      "step": 13717
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.441654483848411e-07,
+      "loss": 0.8932,
+      "step": 13718
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.436014916139196e-07,
+      "loss": 0.9036,
+      "step": 13719
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4303798920066744e-07,
+      "loss": 0.9035,
+      "step": 13720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4247494117159887e-07,
+      "loss": 0.9568,
+      "step": 13721
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.41912347553206e-07,
+      "loss": 0.8345,
+      "step": 13722
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4135020837196e-07,
+      "loss": 0.9624,
+      "step": 13723
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.407885236543107e-07,
+      "loss": 0.9272,
+      "step": 13724
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4022729342668813e-07,
+      "loss": 0.9039,
+      "step": 13725
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.396665177155001e-07,
+      "loss": 0.9553,
+      "step": 13726
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3910619654712875e-07,
+      "loss": 0.9157,
+      "step": 13727
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.385463299479419e-07,
+      "loss": 0.9323,
+      "step": 13728
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.37986917944283e-07,
+      "loss": 0.8857,
+      "step": 13729
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3742796056247195e-07,
+      "loss": 0.9579,
+      "step": 13730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.36869457828809e-07,
+      "loss": 0.8634,
+      "step": 13731
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.363114097695741e-07,
+      "loss": 0.8928,
+      "step": 13732
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.357538164110252e-07,
+      "loss": 0.935,
+      "step": 13733
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.351966777793969e-07,
+      "loss": 0.8978,
+      "step": 13734
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3463999390090595e-07,
+      "loss": 0.9167,
+      "step": 13735
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.340837648017448e-07,
+      "loss": 0.9189,
+      "step": 13736
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.335279905080846e-07,
+      "loss": 0.9714,
+      "step": 13737
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.329726710460768e-07,
+      "loss": 0.8782,
+      "step": 13738
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.324178064418493e-07,
+      "loss": 0.8996,
+      "step": 13739
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3186339672151123e-07,
+      "loss": 0.904,
+      "step": 13740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3130944191114953e-07,
+      "loss": 0.9152,
+      "step": 13741
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.307559420368278e-07,
+      "loss": 0.8503,
+      "step": 13742
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3020289712459077e-07,
+      "loss": 0.8025,
+      "step": 13743
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2965030720045865e-07,
+      "loss": 0.8774,
+      "step": 13744
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2909817229043515e-07,
+      "loss": 0.959,
+      "step": 13745
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.285464924204962e-07,
+      "loss": 0.9362,
+      "step": 13746
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.279952676166009e-07,
+      "loss": 0.9379,
+      "step": 13747
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2744449790468755e-07,
+      "loss": 0.8691,
+      "step": 13748
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.268941833106698e-07,
+      "loss": 0.8721,
+      "step": 13749
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2634432386044133e-07,
+      "loss": 0.8951,
+      "step": 13750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2579491957987485e-07,
+      "loss": 0.8945,
+      "step": 13751
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.252459704948197e-07,
+      "loss": 0.8574,
+      "step": 13752
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.246974766311084e-07,
+      "loss": 0.9168,
+      "step": 13753
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.241494380145449e-07,
+      "loss": 0.8847,
+      "step": 13754
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2360185467091856e-07,
+      "loss": 0.855,
+      "step": 13755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2305472662599426e-07,
+      "loss": 0.9043,
+      "step": 13756
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.225080539055148e-07,
+      "loss": 0.9504,
+      "step": 13757
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2196183653520396e-07,
+      "loss": 0.9415,
+      "step": 13758
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2141607454076127e-07,
+      "loss": 0.892,
+      "step": 13759
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.208707679478662e-07,
+      "loss": 0.9443,
+      "step": 13760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.203259167821782e-07,
+      "loss": 1.0266,
+      "step": 13761
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.197815210693322e-07,
+      "loss": 0.8989,
+      "step": 13762
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.192375808349446e-07,
+      "loss": 0.9449,
+      "step": 13763
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1869409610460813e-07,
+      "loss": 0.9319,
+      "step": 13764
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.181510669038956e-07,
+      "loss": 0.9104,
+      "step": 13765
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1760849325835895e-07,
+      "loss": 0.8665,
+      "step": 13766
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1706637519352437e-07,
+      "loss": 0.9752,
+      "step": 13767
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.165247127349047e-07,
+      "loss": 0.8705,
+      "step": 13768
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.15983505907983e-07,
+      "loss": 0.9561,
+      "step": 13769
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.154427547382244e-07,
+      "loss": 0.9314,
+      "step": 13770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.149024592510741e-07,
+      "loss": 0.9465,
+      "step": 13771
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.14362619471954e-07,
+      "loss": 0.8927,
+      "step": 13772
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.138232354262649e-07,
+      "loss": 0.9754,
+      "step": 13773
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.132843071393854e-07,
+      "loss": 0.9255,
+      "step": 13774
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.127458346366741e-07,
+      "loss": 0.8859,
+      "step": 13775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.122078179434684e-07,
+      "loss": 0.9271,
+      "step": 13776
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1167025708507934e-07,
+      "loss": 0.934,
+      "step": 13777
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.111331520868055e-07,
+      "loss": 0.9269,
+      "step": 13778
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1059650297391775e-07,
+      "loss": 0.9116,
+      "step": 13779
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1006030977166477e-07,
+      "loss": 0.9296,
+      "step": 13780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0952457250527754e-07,
+      "loss": 0.8378,
+      "step": 13781
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0898929119996144e-07,
+      "loss": 0.9111,
+      "step": 13782
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.084544658809063e-07,
+      "loss": 0.925,
+      "step": 13783
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.079200965732765e-07,
+      "loss": 0.8627,
+      "step": 13784
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.073861833022107e-07,
+      "loss": 0.858,
+      "step": 13785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.068527260928367e-07,
+      "loss": 0.9823,
+      "step": 13786
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0631972497025317e-07,
+      "loss": 0.9036,
+      "step": 13787
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.05787179959538e-07,
+      "loss": 0.928,
+      "step": 13788
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.052550910857499e-07,
+      "loss": 0.8971,
+      "step": 13789
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0472345837392226e-07,
+      "loss": 0.8906,
+      "step": 13790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.041922818490739e-07,
+      "loss": 0.9158,
+      "step": 13791
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0366156153619817e-07,
+      "loss": 0.8478,
+      "step": 13792
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.031312974602607e-07,
+      "loss": 0.9591,
+      "step": 13793
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0260148964621814e-07,
+      "loss": 0.9049,
+      "step": 13794
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0207213811899614e-07,
+      "loss": 0.9664,
+      "step": 13795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.015432429035037e-07,
+      "loss": 1.0024,
+      "step": 13796
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.010148040246241e-07,
+      "loss": 0.9377,
+      "step": 13797
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.004868215072232e-07,
+      "loss": 0.961,
+      "step": 13798
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9995929537614545e-07,
+      "loss": 0.9082,
+      "step": 13799
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9943222565620987e-07,
+      "loss": 0.9706,
+      "step": 13800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.989056123722189e-07,
+      "loss": 0.903,
+      "step": 13801
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.983794555489483e-07,
+      "loss": 0.9038,
+      "step": 13802
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.97853755211156e-07,
+      "loss": 0.8875,
+      "step": 13803
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.973285113835789e-07,
+      "loss": 0.8916,
+      "step": 13804
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9680372409092827e-07,
+      "loss": 0.8909,
+      "step": 13805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.962793933578989e-07,
+      "loss": 0.9149,
+      "step": 13806
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9575551920916213e-07,
+      "loss": 0.8696,
+      "step": 13807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.952321016693649e-07,
+      "loss": 0.9755,
+      "step": 13808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.947091407631386e-07,
+      "loss": 0.916,
+      "step": 13809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.941866365150869e-07,
+      "loss": 0.9546,
+      "step": 13810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.936645889497991e-07,
+      "loss": 0.9398,
+      "step": 13811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.931429980918343e-07,
+      "loss": 0.9066,
+      "step": 13812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.926218639657352e-07,
+      "loss": 0.8403,
+      "step": 13813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9210118659602547e-07,
+      "loss": 0.9167,
+      "step": 13814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.915809660072022e-07,
+      "loss": 0.9245,
+      "step": 13815
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.910612022237436e-07,
+      "loss": 0.9802,
+      "step": 13816
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9054189527010557e-07,
+      "loss": 0.9643,
+      "step": 13817
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.900230451707209e-07,
+      "loss": 0.966,
+      "step": 13818
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.895046519500078e-07,
+      "loss": 0.8935,
+      "step": 13819
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.889867156323534e-07,
+      "loss": 0.9392,
+      "step": 13820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.884692362421293e-07,
+      "loss": 0.9155,
+      "step": 13821
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.87952213803685e-07,
+      "loss": 0.9299,
+      "step": 13822
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.874356483413465e-07,
+      "loss": 0.8312,
+      "step": 13823
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8691953987942e-07,
+      "loss": 0.9134,
+      "step": 13824
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8640388844219047e-07,
+      "loss": 0.9156,
+      "step": 13825
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8588869405391963e-07,
+      "loss": 0.8392,
+      "step": 13826
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8537395673884807e-07,
+      "loss": 0.9465,
+      "step": 13827
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8485967652119527e-07,
+      "loss": 0.952,
+      "step": 13828
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8434585342516196e-07,
+      "loss": 0.9409,
+      "step": 13829
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.838324874749232e-07,
+      "loss": 0.9181,
+      "step": 13830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.833195786946341e-07,
+      "loss": 0.8479,
+      "step": 13831
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.828071271084287e-07,
+      "loss": 0.9538,
+      "step": 13832
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8229513274041776e-07,
+      "loss": 0.9201,
+      "step": 13833
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8178359561469413e-07,
+      "loss": 0.9364,
+      "step": 13834
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.812725157553253e-07,
+      "loss": 0.8653,
+      "step": 13835
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.807618931863587e-07,
+      "loss": 0.8778,
+      "step": 13836
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8025172793182063e-07,
+      "loss": 0.8719,
+      "step": 13837
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.797420200157175e-07,
+      "loss": 0.8815,
+      "step": 13838
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7923276946203003e-07,
+      "loss": 0.8728,
+      "step": 13839
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7872397629472026e-07,
+      "loss": 0.9277,
+      "step": 13840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.782156405377279e-07,
+      "loss": 0.9226,
+      "step": 13841
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7770776221497373e-07,
+      "loss": 0.9386,
+      "step": 13842
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.77200341350351e-07,
+      "loss": 0.9187,
+      "step": 13843
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.766933779677383e-07,
+      "loss": 0.9167,
+      "step": 13844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.761868720909877e-07,
+      "loss": 0.9962,
+      "step": 13845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.756808237439323e-07,
+      "loss": 0.8453,
+      "step": 13846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7517523295038316e-07,
+      "loss": 0.9096,
+      "step": 13847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7467009973412785e-07,
+      "loss": 0.8567,
+      "step": 13848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7416542411893623e-07,
+      "loss": 0.9932,
+      "step": 13849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7366120612855484e-07,
+      "loss": 0.9558,
+      "step": 13850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7315744578670587e-07,
+      "loss": 0.8747,
+      "step": 13851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7265414311709483e-07,
+      "loss": 0.9429,
+      "step": 13852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.721512981434027e-07,
+      "loss": 0.9072,
+      "step": 13853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7164891088928945e-07,
+      "loss": 0.8807,
+      "step": 13854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.711469813783929e-07,
+      "loss": 0.9183,
+      "step": 13855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7064550963433076e-07,
+      "loss": 0.947,
+      "step": 13856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7014449568069866e-07,
+      "loss": 0.9375,
+      "step": 13857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6964393954107106e-07,
+      "loss": 0.9427,
+      "step": 13858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.691438412390002e-07,
+      "loss": 0.8788,
+      "step": 13859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.686442007980161e-07,
+      "loss": 0.89,
+      "step": 13860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6814501824162897e-07,
+      "loss": 0.9705,
+      "step": 13861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.676462935933255e-07,
+      "loss": 0.8862,
+      "step": 13862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6714802687657247e-07,
+      "loss": 0.9078,
+      "step": 13863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.666502181148145e-07,
+      "loss": 0.9246,
+      "step": 13864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6615286733147613e-07,
+      "loss": 0.973,
+      "step": 13865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.656559745499565e-07,
+      "loss": 0.9606,
+      "step": 13866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6515953979363794e-07,
+      "loss": 0.8878,
+      "step": 13867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.646635630858774e-07,
+      "loss": 0.8645,
+      "step": 13868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6416804445001274e-07,
+      "loss": 0.9522,
+      "step": 13869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.636729839093588e-07,
+      "loss": 0.9417,
+      "step": 13870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6317838148720907e-07,
+      "loss": 0.9774,
+      "step": 13871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.626842372068361e-07,
+      "loss": 0.9653,
+      "step": 13872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.621905510914935e-07,
+      "loss": 0.9499,
+      "step": 13873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6169732316440486e-07,
+      "loss": 0.9521,
+      "step": 13874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.612045534487817e-07,
+      "loss": 0.8717,
+      "step": 13875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6071224196780766e-07,
+      "loss": 0.8994,
+      "step": 13876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.602203887446508e-07,
+      "loss": 1.0003,
+      "step": 13877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5972899380245165e-07,
+      "loss": 0.9737,
+      "step": 13878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5923805716433046e-07,
+      "loss": 0.9497,
+      "step": 13879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5874757885339e-07,
+      "loss": 0.9552,
+      "step": 13880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.582575588927061e-07,
+      "loss": 0.9095,
+      "step": 13881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5776799730533595e-07,
+      "loss": 0.9287,
+      "step": 13882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.572788941143156e-07,
+      "loss": 0.9297,
+      "step": 13883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.567902493426577e-07,
+      "loss": 0.8687,
+      "step": 13884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.563020630133539e-07,
+      "loss": 0.8634,
+      "step": 13885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.558143351493747e-07,
+      "loss": 0.8554,
+      "step": 13886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.553270657736695e-07,
+      "loss": 0.9307,
+      "step": 13887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.548402549091655e-07,
+      "loss": 0.9677,
+      "step": 13888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.543539025787689e-07,
+      "loss": 0.8764,
+      "step": 13889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5386800880536243e-07,
+      "loss": 0.8831,
+      "step": 13890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5338257361180896e-07,
+      "loss": 0.9124,
+      "step": 13891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5289759702095016e-07,
+      "loss": 0.8957,
+      "step": 13892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.524130790556056e-07,
+      "loss": 0.9003,
+      "step": 13893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.519290197385704e-07,
+      "loss": 0.9618,
+      "step": 13894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.514454190926241e-07,
+      "loss": 0.9362,
+      "step": 13895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5096227714052067e-07,
+      "loss": 0.9027,
+      "step": 13896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.50479593904992e-07,
+      "loss": 0.914,
+      "step": 13897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4999736940874986e-07,
+      "loss": 0.9158,
+      "step": 13898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.495156036744828e-07,
+      "loss": 0.9071,
+      "step": 13899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4903429672486265e-07,
+      "loss": 0.8979,
+      "step": 13900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.485534485825336e-07,
+      "loss": 0.8884,
+      "step": 13901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.480730592701208e-07,
+      "loss": 0.8645,
+      "step": 13902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4759312881022844e-07,
+      "loss": 1.033,
+      "step": 13903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4711365722543737e-07,
+      "loss": 0.9155,
+      "step": 13904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4663464453830944e-07,
+      "loss": 0.8757,
+      "step": 13905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4615609077138224e-07,
+      "loss": 0.9127,
+      "step": 13906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4567799594717336e-07,
+      "loss": 1.0226,
+      "step": 13907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.452003600881792e-07,
+      "loss": 0.9708,
+      "step": 13908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.447231832168717e-07,
+      "loss": 0.8953,
+      "step": 13909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4424646535570526e-07,
+      "loss": 0.9162,
+      "step": 13910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.437702065271097e-07,
+      "loss": 0.9187,
+      "step": 13911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.432944067534937e-07,
+      "loss": 0.8248,
+      "step": 13912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4281906605724603e-07,
+      "loss": 0.9231,
+      "step": 13913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4234418446072993e-07,
+      "loss": 0.8788,
+      "step": 13914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.418697619862942e-07,
+      "loss": 0.9147,
+      "step": 13915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4139579865625983e-07,
+      "loss": 0.9833,
+      "step": 13916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.409222944929246e-07,
+      "loss": 0.8937,
+      "step": 13917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4044924951857283e-07,
+      "loss": 0.9414,
+      "step": 13918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3997666375546014e-07,
+      "loss": 0.867,
+      "step": 13919
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.395045372258231e-07,
+      "loss": 0.9704,
+      "step": 13920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.390328699518762e-07,
+      "loss": 0.8075,
+      "step": 13921
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3856166195581287e-07,
+      "loss": 0.91,
+      "step": 13922
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3809091325980638e-07,
+      "loss": 0.9074,
+      "step": 13923
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3762062388600238e-07,
+      "loss": 0.9623,
+      "step": 13924
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3715079385653316e-07,
+      "loss": 0.8264,
+      "step": 13925
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.366814231935044e-07,
+      "loss": 0.8887,
+      "step": 13926
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3621251191900062e-07,
+      "loss": 0.8397,
+      "step": 13927
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3574406005508422e-07,
+      "loss": 0.9029,
+      "step": 13928
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3527606762379863e-07,
+      "loss": 0.9427,
+      "step": 13929
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3480853464716403e-07,
+      "loss": 0.902,
+      "step": 13930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3434146114717838e-07,
+      "loss": 0.9401,
+      "step": 13931
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3387484714581854e-07,
+      "loss": 0.9811,
+      "step": 13932
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3340869266504029e-07,
+      "loss": 0.9305,
+      "step": 13933
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3294299772677832e-07,
+      "loss": 0.9326,
+      "step": 13934
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3247776235294284e-07,
+      "loss": 0.8829,
+      "step": 13935
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3201298656542525e-07,
+      "loss": 0.9238,
+      "step": 13936
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3154867038609252e-07,
+      "loss": 0.8367,
+      "step": 13937
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3108481383679604e-07,
+      "loss": 0.9279,
+      "step": 13938
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3062141693935947e-07,
+      "loss": 0.8774,
+      "step": 13939
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3015847971558424e-07,
+      "loss": 0.8776,
+      "step": 13940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.296960021872563e-07,
+      "loss": 0.8918,
+      "step": 13941
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2923398437613487e-07,
+      "loss": 0.9297,
+      "step": 13942
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2877242630395924e-07,
+      "loss": 0.922,
+      "step": 13943
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.283113279924465e-07,
+      "loss": 0.9011,
+      "step": 13944
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2785068946329148e-07,
+      "loss": 0.8683,
+      "step": 13945
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2739051073817132e-07,
+      "loss": 0.8937,
+      "step": 13946
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2693079183873535e-07,
+      "loss": 0.8479,
+      "step": 13947
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2647153278661738e-07,
+      "loss": 0.8659,
+      "step": 13948
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2601273360342567e-07,
+      "loss": 0.9405,
+      "step": 13949
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2555439431074633e-07,
+      "loss": 0.9302,
+      "step": 13950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.250965149301476e-07,
+      "loss": 0.919,
+      "step": 13951
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.246390954831712e-07,
+      "loss": 0.906,
+      "step": 13952
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2418213599134208e-07,
+      "loss": 0.9216,
+      "step": 13953
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2372563647616196e-07,
+      "loss": 0.8695,
+      "step": 13954
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2326959695910812e-07,
+      "loss": 0.9253,
+      "step": 13955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.228140174616389e-07,
+      "loss": 0.9653,
+      "step": 13956
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.223588980051905e-07,
+      "loss": 1.0705,
+      "step": 13957
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2190423861118027e-07,
+      "loss": 0.9711,
+      "step": 13958
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2145003930099662e-07,
+      "loss": 0.9483,
+      "step": 13959
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2099630009601138e-07,
+      "loss": 0.9758,
+      "step": 13960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.205430210175774e-07,
+      "loss": 0.8612,
+      "step": 13961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2009020208701993e-07,
+      "loss": 0.9237,
+      "step": 13962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.196378433256463e-07,
+      "loss": 0.8392,
+      "step": 13963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1918594475474174e-07,
+      "loss": 0.8676,
+      "step": 13964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1873450639556594e-07,
+      "loss": 0.9617,
+      "step": 13965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1828352826936405e-07,
+      "loss": 0.8617,
+      "step": 13966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.178330103973536e-07,
+      "loss": 0.9328,
+      "step": 13967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.173829528007332e-07,
+      "loss": 0.9408,
+      "step": 13968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1693335550067918e-07,
+      "loss": 0.962,
+      "step": 13969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1648421851834577e-07,
+      "loss": 0.9694,
+      "step": 13970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1603554187486718e-07,
+      "loss": 0.9734,
+      "step": 13971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.155873255913532e-07,
+      "loss": 0.9098,
+      "step": 13972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.151395696888936e-07,
+      "loss": 0.9455,
+      "step": 13973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1469227418855932e-07,
+      "loss": 0.9635,
+      "step": 13974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.142454391113924e-07,
+      "loss": 0.9543,
+      "step": 13975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1379906447842048e-07,
+      "loss": 0.8862,
+      "step": 13976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1335315031064563e-07,
+      "loss": 0.8761,
+      "step": 13977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1290769662904997e-07,
+      "loss": 0.9205,
+      "step": 13978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.124627034545923e-07,
+      "loss": 0.8588,
+      "step": 13979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.120181708082103e-07,
+      "loss": 0.908,
+      "step": 13980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1157409871082168e-07,
+      "loss": 0.9438,
+      "step": 13981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1113048718332086e-07,
+      "loss": 0.9493,
+      "step": 13982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.106873362465789e-07,
+      "loss": 0.9304,
+      "step": 13983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1024464592145021e-07,
+      "loss": 0.8874,
+      "step": 13984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0980241622876262e-07,
+      "loss": 0.9268,
+      "step": 13985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0936064718932392e-07,
+      "loss": 0.957,
+      "step": 13986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.089193388239219e-07,
+      "loss": 0.9969,
+      "step": 13987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0847849115331887e-07,
+      "loss": 0.9577,
+      "step": 13988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0803810419826154e-07,
+      "loss": 0.9708,
+      "step": 13989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.075981779794667e-07,
+      "loss": 0.8775,
+      "step": 13990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0715871251763774e-07,
+      "loss": 0.9216,
+      "step": 13991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0671970783345041e-07,
+      "loss": 0.9262,
+      "step": 13992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0628116394756258e-07,
+      "loss": 0.8659,
+      "step": 13993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0584308088060777e-07,
+      "loss": 0.8615,
+      "step": 13994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0540545865319726e-07,
+      "loss": 0.879,
+      "step": 13995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0496829728592572e-07,
+      "loss": 0.9475,
+      "step": 13996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0453159679936218e-07,
+      "loss": 0.9414,
+      "step": 13997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.040953572140514e-07,
+      "loss": 0.9185,
+      "step": 13998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0365957855052244e-07,
+      "loss": 0.8594,
+      "step": 13999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0322426082927782e-07,
+      "loss": 0.9348,
+      "step": 14000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0278940407080228e-07,
+      "loss": 0.9309,
+      "step": 14001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.023550082955561e-07,
+      "loss": 0.8788,
+      "step": 14002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0192107352397739e-07,
+      "loss": 0.9077,
+      "step": 14003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0148759977648536e-07,
+      "loss": 0.8874,
+      "step": 14004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0105458707347702e-07,
+      "loss": 0.893,
+      "step": 14005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0062203543532388e-07,
+      "loss": 0.8704,
+      "step": 14006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0018994488238075e-07,
+      "loss": 0.9341,
+      "step": 14007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9975831543497803e-07,
+      "loss": 0.9087,
+      "step": 14008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9932714711342504e-07,
+      "loss": 0.9492,
+      "step": 14009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9889643993800777e-07,
+      "loss": 0.9475,
+      "step": 14010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9846619392899446e-07,
+      "loss": 0.885,
+      "step": 14011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.980364091066278e-07,
+      "loss": 0.9092,
+      "step": 14012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9760708549113161e-07,
+      "loss": 0.9797,
+      "step": 14013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.971782231027053e-07,
+      "loss": 0.9297,
+      "step": 14014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9674982196152716e-07,
+      "loss": 0.9962,
+      "step": 14015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.963218820877566e-07,
+      "loss": 0.9394,
+      "step": 14016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9589440350152866e-07,
+      "loss": 0.8949,
+      "step": 14017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9546738622295613e-07,
+      "loss": 0.8647,
+      "step": 14018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9504083027213183e-07,
+      "loss": 0.9229,
+      "step": 14019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.946147356691286e-07,
+      "loss": 0.8945,
+      "step": 14020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9418910243399037e-07,
+      "loss": 0.9608,
+      "step": 14021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.937639305867489e-07,
+      "loss": 0.9812,
+      "step": 14022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9333922014740603e-07,
+      "loss": 0.92,
+      "step": 14023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9291497113595014e-07,
+      "loss": 0.925,
+      "step": 14024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9249118357233754e-07,
+      "loss": 0.9551,
+      "step": 14025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9206785747651112e-07,
+      "loss": 0.9888,
+      "step": 14026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9164499286839056e-07,
+      "loss": 0.8889,
+      "step": 14027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9122258976787212e-07,
+      "loss": 0.9491,
+      "step": 14028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.908006481948299e-07,
+      "loss": 0.8661,
+      "step": 14029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9037916816911694e-07,
+      "loss": 0.8889,
+      "step": 14030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8995814971056736e-07,
+      "loss": 0.9427,
+      "step": 14031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8953759283898864e-07,
+      "loss": 0.8687,
+      "step": 14032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.891174975741683e-07,
+      "loss": 0.8988,
+      "step": 14033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.886978639358772e-07,
+      "loss": 0.895,
+      "step": 14034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8827869194385618e-07,
+      "loss": 0.9437,
+      "step": 14035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.878599816178306e-07,
+      "loss": 0.9511,
+      "step": 14036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8744173297750024e-07,
+      "loss": 0.9243,
+      "step": 14037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.870239460425438e-07,
+      "loss": 0.9469,
+      "step": 14038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8660662083262338e-07,
+      "loss": 0.8873,
+      "step": 14039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.861897573673721e-07,
+      "loss": 0.9251,
+      "step": 14040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.857733556664032e-07,
+      "loss": 0.9319,
+      "step": 14041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8535741574931209e-07,
+      "loss": 0.8856,
+      "step": 14042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.849419376356687e-07,
+      "loss": 0.9776,
+      "step": 14043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8452692134502294e-07,
+      "loss": 0.9102,
+      "step": 14044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8411236689690138e-07,
+      "loss": 0.8282,
+      "step": 14045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8369827431080955e-07,
+      "loss": 0.871,
+      "step": 14046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.832846436062341e-07,
+      "loss": 0.9584,
+      "step": 14047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8287147480263388e-07,
+      "loss": 0.9424,
+      "step": 14048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8245876791945226e-07,
+      "loss": 0.9561,
+      "step": 14049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8204652297610703e-07,
+      "loss": 0.92,
+      "step": 14050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8163473999199488e-07,
+      "loss": 0.8872,
+      "step": 14051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8122341898649254e-07,
+      "loss": 1.0397,
+      "step": 14052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.808125599789523e-07,
+      "loss": 0.8778,
+      "step": 14053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8040216298870538e-07,
+      "loss": 0.917,
+      "step": 14054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7999222803506634e-07,
+      "loss": 0.9446,
+      "step": 14055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.795827551373175e-07,
+      "loss": 0.8852,
+      "step": 14056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.791737443147301e-07,
+      "loss": 0.9164,
+      "step": 14057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.787651955865477e-07,
+      "loss": 0.9231,
+      "step": 14058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7835710897199377e-07,
+      "loss": 0.911,
+      "step": 14059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7794948449026961e-07,
+      "loss": 0.9573,
+      "step": 14060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7754232216055435e-07,
+      "loss": 0.8611,
+      "step": 14061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7713562200200706e-07,
+      "loss": 0.9733,
+      "step": 14062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7672938403376362e-07,
+      "loss": 0.9419,
+      "step": 14063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7632360827493754e-07,
+      "loss": 0.8678,
+      "step": 14064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7591829474462363e-07,
+      "loss": 0.8337,
+      "step": 14065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7551344346189102e-07,
+      "loss": 0.8788,
+      "step": 14066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7510905444579007e-07,
+      "loss": 0.8497,
+      "step": 14067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7470512771534776e-07,
+      "loss": 0.8121,
+      "step": 14068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7430166328956887e-07,
+      "loss": 0.9187,
+      "step": 14069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7389866118744048e-07,
+      "loss": 0.9321,
+      "step": 14070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7349612142792184e-07,
+      "loss": 0.8075,
+      "step": 14071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7309404402995556e-07,
+      "loss": 0.8701,
+      "step": 14072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7269242901245876e-07,
+      "loss": 0.8626,
+      "step": 14073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7229127639432852e-07,
+      "loss": 0.9386,
+      "step": 14074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7189058619444088e-07,
+      "loss": 0.8561,
+      "step": 14075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7149035843164853e-07,
+      "loss": 0.9349,
+      "step": 14076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7109059312478417e-07,
+      "loss": 0.8757,
+      "step": 14077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.706912902926583e-07,
+      "loss": 0.9546,
+      "step": 14078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7029244995405703e-07,
+      "loss": 0.9395,
+      "step": 14079
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6989407212774755e-07,
+      "loss": 0.9737,
+      "step": 14080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.694961568324749e-07,
+      "loss": 0.9204,
+      "step": 14081
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6909870408696295e-07,
+      "loss": 0.9261,
+      "step": 14082
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6870171390991118e-07,
+      "loss": 0.9328,
+      "step": 14083
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6830518631999916e-07,
+      "loss": 0.9089,
+      "step": 14084
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.679091213358852e-07,
+      "loss": 1.0162,
+      "step": 14085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.675135189762056e-07,
+      "loss": 0.827,
+      "step": 14086
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.671183792595732e-07,
+      "loss": 0.8715,
+      "step": 14087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6672370220458089e-07,
+      "loss": 0.9344,
+      "step": 14088
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.663294878297994e-07,
+      "loss": 0.9263,
+      "step": 14089
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6593573615377722e-07,
+      "loss": 0.9173,
+      "step": 14090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6554244719504063e-07,
+      "loss": 0.8552,
+      "step": 14091
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6514962097209598e-07,
+      "loss": 0.9294,
+      "step": 14092
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.647572575034262e-07,
+      "loss": 0.9305,
+      "step": 14093
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.643653568074932e-07,
+      "loss": 0.9068,
+      "step": 14094
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6397391890273783e-07,
+      "loss": 0.9183,
+      "step": 14095
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6358294380757645e-07,
+      "loss": 0.9302,
+      "step": 14096
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6319243154040543e-07,
+      "loss": 0.9235,
+      "step": 14097
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.628023821196012e-07,
+      "loss": 0.9621,
+      "step": 14098
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6241279556351465e-07,
+      "loss": 0.9372,
+      "step": 14099
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6202367189047774e-07,
+      "loss": 0.8595,
+      "step": 14100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.616350111188003e-07,
+      "loss": 0.9285,
+      "step": 14101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6124681326676884e-07,
+      "loss": 0.9457,
+      "step": 14102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6085907835264868e-07,
+      "loss": 0.8634,
+      "step": 14103
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6047180639468418e-07,
+      "loss": 0.9302,
+      "step": 14104
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6008499741109962e-07,
+      "loss": 0.8296,
+      "step": 14105
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5969865142009156e-07,
+      "loss": 0.8557,
+      "step": 14106
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5931276843983988e-07,
+      "loss": 0.8469,
+      "step": 14107
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5892734848850232e-07,
+      "loss": 0.9445,
+      "step": 14108
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5854239158421324e-07,
+      "loss": 0.8874,
+      "step": 14109
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5815789774508594e-07,
+      "loss": 0.8674,
+      "step": 14110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.577738669892115e-07,
+      "loss": 0.91,
+      "step": 14111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5739029933465878e-07,
+      "loss": 0.8894,
+      "step": 14112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.570071947994789e-07,
+      "loss": 0.8635,
+      "step": 14113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5662455340169302e-07,
+      "loss": 0.876,
+      "step": 14114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5624237515931008e-07,
+      "loss": 0.9296,
+      "step": 14115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.55860660090309e-07,
+      "loss": 0.9524,
+      "step": 14116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5547940821265205e-07,
+      "loss": 0.9532,
+      "step": 14117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.550986195442783e-07,
+      "loss": 0.9781,
+      "step": 14118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5471829410310334e-07,
+      "loss": 0.9527,
+      "step": 14119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5433843190702402e-07,
+      "loss": 0.9362,
+      "step": 14120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.539590329739138e-07,
+      "loss": 0.9853,
+      "step": 14121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5358009732162281e-07,
+      "loss": 0.8931,
+      "step": 14122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.532016249679813e-07,
+      "loss": 0.9009,
+      "step": 14123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5282361593079942e-07,
+      "loss": 0.9893,
+      "step": 14124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5244607022786073e-07,
+      "loss": 0.8592,
+      "step": 14125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5206898787693214e-07,
+      "loss": 1.0003,
+      "step": 14126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.516923688957528e-07,
+      "loss": 0.9607,
+      "step": 14127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.513162133020485e-07,
+      "loss": 0.9825,
+      "step": 14128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.50940521113514e-07,
+      "loss": 0.8864,
+      "step": 14129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5056529234782847e-07,
+      "loss": 0.9906,
+      "step": 14130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.501905270226467e-07,
+      "loss": 0.9212,
+      "step": 14131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4981622515560346e-07,
+      "loss": 0.9133,
+      "step": 14132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.494423867643091e-07,
+      "loss": 0.8458,
+      "step": 14133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4906901186635513e-07,
+      "loss": 0.9534,
+      "step": 14134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.486961004793075e-07,
+      "loss": 0.9004,
+      "step": 14135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4832365262071658e-07,
+      "loss": 0.963,
+      "step": 14136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4795166830810282e-07,
+      "loss": 0.9618,
+      "step": 14137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.475801475589711e-07,
+      "loss": 0.8471,
+      "step": 14138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4720909039080188e-07,
+      "loss": 0.8483,
+      "step": 14139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4683849682105456e-07,
+      "loss": 0.8979,
+      "step": 14140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4646836686716625e-07,
+      "loss": 0.8538,
+      "step": 14141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4609870054655305e-07,
+      "loss": 0.948,
+      "step": 14142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4572949787660774e-07,
+      "loss": 0.9169,
+      "step": 14143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4536075887470413e-07,
+      "loss": 0.9521,
+      "step": 14144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.449924835581895e-07,
+      "loss": 0.8479,
+      "step": 14145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4462467194439333e-07,
+      "loss": 0.9587,
+      "step": 14146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4425732405062287e-07,
+      "loss": 0.8778,
+      "step": 14147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.438904398941632e-07,
+      "loss": 0.9059,
+      "step": 14148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.435240194922749e-07,
+      "loss": 0.882,
+      "step": 14149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.431580628621998e-07,
+      "loss": 0.9681,
+      "step": 14150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4279257002115854e-07,
+      "loss": 0.9807,
+      "step": 14151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4242754098634738e-07,
+      "loss": 0.96,
+      "step": 14152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4206297577494143e-07,
+      "loss": 0.8535,
+      "step": 14153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.416988744040948e-07,
+      "loss": 0.9123,
+      "step": 14154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4133523689093932e-07,
+      "loss": 0.8865,
+      "step": 14155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4097206325258572e-07,
+      "loss": 0.8715,
+      "step": 14156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4060935350612037e-07,
+      "loss": 0.8943,
+      "step": 14157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4024710766861183e-07,
+      "loss": 0.8802,
+      "step": 14158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3988532575710423e-07,
+      "loss": 0.9149,
+      "step": 14159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3952400778861952e-07,
+      "loss": 0.9465,
+      "step": 14160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3916315378015854e-07,
+      "loss": 0.9269,
+      "step": 14161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3880276374870215e-07,
+      "loss": 0.9056,
+      "step": 14162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3844283771120458e-07,
+      "loss": 0.8416,
+      "step": 14163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3808337568460338e-07,
+      "loss": 0.9125,
+      "step": 14164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.377243776858117e-07,
+      "loss": 0.9006,
+      "step": 14165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3736584373172156e-07,
+      "loss": 0.9246,
+      "step": 14166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3700777383920282e-07,
+      "loss": 0.8996,
+      "step": 14167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36650168025102e-07,
+      "loss": 0.8676,
+      "step": 14168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3629302630624786e-07,
+      "loss": 0.8611,
+      "step": 14169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.359363486994425e-07,
+      "loss": 0.9355,
+      "step": 14170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.355801352214714e-07,
+      "loss": 0.9112,
+      "step": 14171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3522438588909225e-07,
+      "loss": 0.9172,
+      "step": 14172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3486910071904503e-07,
+      "loss": 0.9405,
+      "step": 14173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.345142797280474e-07,
+      "loss": 0.8933,
+      "step": 14174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.341599229327939e-07,
+      "loss": 0.9215,
+      "step": 14175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3380603034995775e-07,
+      "loss": 0.9521,
+      "step": 14176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3345260199619125e-07,
+      "loss": 0.8787,
+      "step": 14177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.330996378881233e-07,
+      "loss": 0.8864,
+      "step": 14178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3274713804236394e-07,
+      "loss": 0.9228,
+      "step": 14179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3239510247549549e-07,
+      "loss": 0.8754,
+      "step": 14180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3204353120408463e-07,
+      "loss": 0.9015,
+      "step": 14181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3169242424467266e-07,
+      "loss": 0.8837,
+      "step": 14182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3134178161378076e-07,
+      "loss": 0.9921,
+      "step": 14183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3099160332790684e-07,
+      "loss": 1.0004,
+      "step": 14184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3064188940352772e-07,
+      "loss": 1.0815,
+      "step": 14185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3029263985710028e-07,
+      "loss": 0.9103,
+      "step": 14186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2994385470505465e-07,
+      "loss": 0.9957,
+      "step": 14187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.295955339638033e-07,
+      "loss": 0.8525,
+      "step": 14188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2924767764973535e-07,
+      "loss": 0.9789,
+      "step": 14189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2890028577921875e-07,
+      "loss": 0.8642,
+      "step": 14190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2855335836859939e-07,
+      "loss": 0.9305,
+      "step": 14191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2820689543420084e-07,
+      "loss": 0.8163,
+      "step": 14192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.278608969923234e-07,
+      "loss": 0.8699,
+      "step": 14193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2751536305925073e-07,
+      "loss": 0.9302,
+      "step": 14194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2717029365123646e-07,
+      "loss": 0.885,
+      "step": 14195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2682568878452094e-07,
+      "loss": 0.954,
+      "step": 14196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.264815484753168e-07,
+      "loss": 0.9112,
+      "step": 14197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2613787273981771e-07,
+      "loss": 0.9554,
+      "step": 14198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2579466159419296e-07,
+      "loss": 0.9597,
+      "step": 14199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2545191505459298e-07,
+      "loss": 0.8727,
+      "step": 14200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2510963313714375e-07,
+      "loss": 0.8929,
+      "step": 14201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2476781585795127e-07,
+      "loss": 0.9276,
+      "step": 14202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2442646323309827e-07,
+      "loss": 0.9426,
+      "step": 14203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.240855752786474e-07,
+      "loss": 0.8054,
+      "step": 14204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2374515201063698e-07,
+      "loss": 0.8963,
+      "step": 14205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.234051934450853e-07,
+      "loss": 0.9059,
+      "step": 14206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2306569959798842e-07,
+      "loss": 0.9027,
+      "step": 14207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2272667048531918e-07,
+      "loss": 0.8835,
+      "step": 14208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2238810612303141e-07,
+      "loss": 0.9151,
+      "step": 14209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2205000652705578e-07,
+      "loss": 0.921,
+      "step": 14210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.217123717132984e-07,
+      "loss": 0.9015,
+      "step": 14211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.213752016976477e-07,
+      "loss": 0.9164,
+      "step": 14212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2103849649596765e-07,
+      "loss": 0.957,
+      "step": 14213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2070225612410115e-07,
+      "loss": 0.9081,
+      "step": 14214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2036648059786992e-07,
+      "loss": 0.9678,
+      "step": 14215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.200311699330714e-07,
+      "loss": 0.8631,
+      "step": 14216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1969632414548405e-07,
+      "loss": 0.9124,
+      "step": 14217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1936194325086303e-07,
+      "loss": 0.8489,
+      "step": 14218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1902802726494134e-07,
+      "loss": 0.919,
+      "step": 14219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1869457620343083e-07,
+      "loss": 0.8645,
+      "step": 14220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.183615900820212e-07,
+      "loss": 0.8768,
+      "step": 14221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1802906891638099e-07,
+      "loss": 0.9423,
+      "step": 14222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1769701272215439e-07,
+      "loss": 0.8508,
+      "step": 14223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1736542151496777e-07,
+      "loss": 0.954,
+      "step": 14224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1703429531042198e-07,
+      "loss": 0.935,
+      "step": 14225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.167036341240968e-07,
+      "loss": 0.8873,
+      "step": 14226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1637343797155087e-07,
+      "loss": 0.9512,
+      "step": 14227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1604370686832178e-07,
+      "loss": 0.905,
+      "step": 14228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1571444082992489e-07,
+      "loss": 0.9913,
+      "step": 14229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1538563987185003e-07,
+      "loss": 1.0144,
+      "step": 14230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1505730400957039e-07,
+      "loss": 0.935,
+      "step": 14231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.147294332585347e-07,
+      "loss": 0.8982,
+      "step": 14232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1440202763416952e-07,
+      "loss": 0.8965,
+      "step": 14233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1407508715188032e-07,
+      "loss": 0.9865,
+      "step": 14234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1374861182705033e-07,
+      "loss": 0.8796,
+      "step": 14235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1342260167504171e-07,
+      "loss": 0.961,
+      "step": 14236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1309705671119442e-07,
+      "loss": 0.9488,
+      "step": 14237
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1277197695082398e-07,
+      "loss": 0.945,
+      "step": 14238
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1244736240922704e-07,
+      "loss": 0.8502,
+      "step": 14239
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1212321310167917e-07,
+      "loss": 0.9355,
+      "step": 14240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.117995290434315e-07,
+      "loss": 0.9339,
+      "step": 14241
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1147631024971295e-07,
+      "loss": 0.859,
+      "step": 14242
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1115355673573358e-07,
+      "loss": 0.9489,
+      "step": 14243
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1083126851667902e-07,
+      "loss": 0.9114,
+      "step": 14244
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1050944560771271e-07,
+      "loss": 0.8857,
+      "step": 14245
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1018808802397807e-07,
+      "loss": 0.9174,
+      "step": 14246
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0986719578059635e-07,
+      "loss": 0.9269,
+      "step": 14247
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0954676889266658e-07,
+      "loss": 0.8481,
+      "step": 14248
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0922680737526448e-07,
+      "loss": 0.8804,
+      "step": 14249
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0890731124344578e-07,
+      "loss": 0.9898,
+      "step": 14250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.085882805122418e-07,
+      "loss": 0.9853,
+      "step": 14251
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0826971519666718e-07,
+      "loss": 0.8605,
+      "step": 14252
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0795161531170772e-07,
+      "loss": 0.9829,
+      "step": 14253
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0763398087233257e-07,
+      "loss": 0.9054,
+      "step": 14254
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0731681189348753e-07,
+      "loss": 0.9737,
+      "step": 14255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0700010839009511e-07,
+      "loss": 0.9443,
+      "step": 14256
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0668387037705785e-07,
+      "loss": 0.9071,
+      "step": 14257
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0636809786925606e-07,
+      "loss": 0.989,
+      "step": 14258
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0605279088154452e-07,
+      "loss": 0.8999,
+      "step": 14259
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0573794942876358e-07,
+      "loss": 0.9839,
+      "step": 14260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0542357352572474e-07,
+      "loss": 0.8962,
+      "step": 14261
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.051096631872195e-07,
+      "loss": 0.9092,
+      "step": 14262
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0479621842802046e-07,
+      "loss": 0.9627,
+      "step": 14263
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0448323926287473e-07,
+      "loss": 0.8371,
+      "step": 14264
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0417072570650832e-07,
+      "loss": 0.8741,
+      "step": 14265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0385867777362613e-07,
+      "loss": 0.881,
+      "step": 14266
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0354709547891195e-07,
+      "loss": 0.9315,
+      "step": 14267
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0323597883702519e-07,
+      "loss": 0.8704,
+      "step": 14268
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0292532786260412e-07,
+      "loss": 0.8895,
+      "step": 14269
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0261514257026706e-07,
+      "loss": 0.9032,
+      "step": 14270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.02305422974609e-07,
+      "loss": 0.9218,
+      "step": 14271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.019961690902016e-07,
+      "loss": 0.8681,
+      "step": 14272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0168738093159769e-07,
+      "loss": 0.919,
+      "step": 14273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0137905851332453e-07,
+      "loss": 0.9635,
+      "step": 14274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0107120184989161e-07,
+      "loss": 0.9507,
+      "step": 14275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0076381095578291e-07,
+      "loss": 0.9366,
+      "step": 14276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0045688584546242e-07,
+      "loss": 0.9407,
+      "step": 14277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0015042653337193e-07,
+      "loss": 0.9131,
+      "step": 14278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.984443303393099e-08,
+      "loss": 0.9514,
+      "step": 14279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.95389053615381e-08,
+      "loss": 0.9221,
+      "step": 14280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.92338435305673e-08,
+      "loss": 0.9576,
+      "step": 14281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.89292475553727e-08,
+      "loss": 0.9864,
+      "step": 14282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.862511745028835e-08,
+      "loss": 0.9597,
+      "step": 14283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.832145322962284e-08,
+      "loss": 0.8794,
+      "step": 14284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.80182549076647e-08,
+      "loss": 0.8703,
+      "step": 14285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.771552249867921e-08,
+      "loss": 0.8684,
+      "step": 14286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.741325601691276e-08,
+      "loss": 0.8413,
+      "step": 14287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.711145547658618e-08,
+      "loss": 0.9172,
+      "step": 14288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.681012089189923e-08,
+      "loss": 0.9536,
+      "step": 14289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.650925227703168e-08,
+      "loss": 0.9272,
+      "step": 14290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.620884964614108e-08,
+      "loss": 0.9285,
+      "step": 14291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590891301336059e-08,
+      "loss": 0.9621,
+      "step": 14292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.560944239280334e-08,
+      "loss": 0.8925,
+      "step": 14293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.53104377985603e-08,
+      "loss": 0.8268,
+      "step": 14294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.501189924470022e-08,
+      "loss": 0.8589,
+      "step": 14295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.471382674526963e-08,
+      "loss": 0.906,
+      "step": 14296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.441622031429398e-08,
+      "loss": 0.9047,
+      "step": 14297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.411907996577763e-08,
+      "loss": 0.9201,
+      "step": 14298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.382240571369938e-08,
+      "loss": 0.8821,
+      "step": 14299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.352619757202141e-08,
+      "loss": 0.8913,
+      "step": 14300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.323045555467814e-08,
+      "loss": 0.879,
+      "step": 14301
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.293517967558729e-08,
+      "loss": 0.9219,
+      "step": 14302
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.264036994864e-08,
+      "loss": 0.9631,
+      "step": 14303
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.234602638770963e-08,
+      "loss": 0.9066,
+      "step": 14304
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.205214900664505e-08,
+      "loss": 0.8646,
+      "step": 14305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.175873781927525e-08,
+      "loss": 0.9068,
+      "step": 14306
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.14657928394036e-08,
+      "loss": 0.8595,
+      "step": 14307
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.117331408081686e-08,
+      "loss": 0.8957,
+      "step": 14308
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.088130155727293e-08,
+      "loss": 0.8382,
+      "step": 14309
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.058975528251523e-08,
+      "loss": 0.9383,
+      "step": 14310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.02986752702606e-08,
+      "loss": 0.9372,
+      "step": 14311
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.000806153420472e-08,
+      "loss": 0.9589,
+      "step": 14312
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.971791408802111e-08,
+      "loss": 0.8704,
+      "step": 14313
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.94282329453644e-08,
+      "loss": 0.9396,
+      "step": 14314
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.913901811986148e-08,
+      "loss": 0.8602,
+      "step": 14315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.885026962512367e-08,
+      "loss": 0.9318,
+      "step": 14316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.856198747473455e-08,
+      "loss": 0.956,
+      "step": 14317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.827417168226215e-08,
+      "loss": 0.9125,
+      "step": 14318
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.798682226124456e-08,
+      "loss": 0.9118,
+      "step": 14319
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.769993922520536e-08,
+      "loss": 0.8995,
+      "step": 14320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.741352258764158e-08,
+      "loss": 0.9417,
+      "step": 14321
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.712757236203129e-08,
+      "loss": 0.946,
+      "step": 14322
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.684208856182707e-08,
+      "loss": 0.8769,
+      "step": 14323
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.655707120046375e-08,
+      "loss": 0.9545,
+      "step": 14324
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.627252029135057e-08,
+      "loss": 0.8774,
+      "step": 14325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.598843584787797e-08,
+      "loss": 0.9241,
+      "step": 14326
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.570481788341079e-08,
+      "loss": 0.8551,
+      "step": 14327
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.542166641129612e-08,
+      "loss": 0.8804,
+      "step": 14328
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.513898144485555e-08,
+      "loss": 0.8922,
+      "step": 14329
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.485676299739065e-08,
+      "loss": 0.9947,
+      "step": 14330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.45750110821797e-08,
+      "loss": 1.0196,
+      "step": 14331
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.429372571248095e-08,
+      "loss": 0.9442,
+      "step": 14332
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.40129069015283e-08,
+      "loss": 0.9537,
+      "step": 14333
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.373255466253671e-08,
+      "loss": 0.8942,
+      "step": 14334
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.345266900869564e-08,
+      "loss": 0.9155,
+      "step": 14335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.317324995317565e-08,
+      "loss": 0.9916,
+      "step": 14336
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.289429750912404e-08,
+      "loss": 0.9072,
+      "step": 14337
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.261581168966581e-08,
+      "loss": 0.924,
+      "step": 14338
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.233779250790386e-08,
+      "loss": 0.9811,
+      "step": 14339
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.206023997692103e-08,
+      "loss": 0.9553,
+      "step": 14340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.178315410977578e-08,
+      "loss": 0.9082,
+      "step": 14341
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.150653491950655e-08,
+      "loss": 0.8384,
+      "step": 14342
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.123038241912739e-08,
+      "loss": 0.9342,
+      "step": 14343
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.095469662163346e-08,
+      "loss": 0.8942,
+      "step": 14344
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.067947753999661e-08,
+      "loss": 0.8374,
+      "step": 14345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.040472518716647e-08,
+      "loss": 0.9594,
+      "step": 14346
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.013043957606825e-08,
+      "loss": 0.8681,
+      "step": 14347
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.985662071961165e-08,
+      "loss": 0.8696,
+      "step": 14348
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.958326863067856e-08,
+      "loss": 0.9862,
+      "step": 14349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.931038332213092e-08,
+      "loss": 0.9481,
+      "step": 14350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.903796480680848e-08,
+      "loss": 0.906,
+      "step": 14351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.876601309752985e-08,
+      "loss": 0.9515,
+      "step": 14352
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.849452820709147e-08,
+      "loss": 0.873,
+      "step": 14353
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.822351014826646e-08,
+      "loss": 0.8638,
+      "step": 14354
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.795295893380572e-08,
+      "loss": 0.8797,
+      "step": 14355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.768287457644241e-08,
+      "loss": 0.895,
+      "step": 14356
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.741325708888303e-08,
+      "loss": 0.9566,
+      "step": 14357
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.714410648381187e-08,
+      "loss": 0.9338,
+      "step": 14358
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.687542277389659e-08,
+      "loss": 0.8923,
+      "step": 14359
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.660720597177706e-08,
+      "loss": 0.9248,
+      "step": 14360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.633945609007542e-08,
+      "loss": 0.9682,
+      "step": 14361
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.607217314138826e-08,
+      "loss": 0.9019,
+      "step": 14362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.580535713829218e-08,
+      "loss": 0.8646,
+      "step": 14363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.553900809334269e-08,
+      "loss": 0.9661,
+      "step": 14364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.527312601907089e-08,
+      "loss": 0.9037,
+      "step": 14365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.500771092798787e-08,
+      "loss": 0.8922,
+      "step": 14366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.474276283258142e-08,
+      "loss": 0.9694,
+      "step": 14367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.447828174531935e-08,
+      "loss": 0.949,
+      "step": 14368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.421426767864504e-08,
+      "loss": 0.9713,
+      "step": 14369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.395072064497966e-08,
+      "loss": 0.8823,
+      "step": 14370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.368764065672773e-08,
+      "loss": 0.8134,
+      "step": 14371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.342502772626381e-08,
+      "loss": 0.9372,
+      "step": 14372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.316288186594689e-08,
+      "loss": 0.9213,
+      "step": 14373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.290120308811155e-08,
+      "loss": 0.9192,
+      "step": 14374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.263999140506794e-08,
+      "loss": 0.9071,
+      "step": 14375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.237924682910958e-08,
+      "loss": 0.9181,
+      "step": 14376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.21189693725044e-08,
+      "loss": 0.9193,
+      "step": 14377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.185915904749818e-08,
+      "loss": 0.8777,
+      "step": 14378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.159981586631671e-08,
+      "loss": 0.8905,
+      "step": 14379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.134093984116131e-08,
+      "loss": 0.8267,
+      "step": 14380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.108253098421448e-08,
+      "loss": 0.914,
+      "step": 14381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.082458930763425e-08,
+      "loss": 0.9755,
+      "step": 14382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.056711482355649e-08,
+      "loss": 1.0031,
+      "step": 14383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.031010754409817e-08,
+      "loss": 0.9331,
+      "step": 14384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.005356748135072e-08,
+      "loss": 0.9673,
+      "step": 14385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.979749464738451e-08,
+      "loss": 0.9177,
+      "step": 14386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.954188905424986e-08,
+      "loss": 0.897,
+      "step": 14387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.928675071397161e-08,
+      "loss": 0.9682,
+      "step": 14388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.903207963855685e-08,
+      "loss": 0.9436,
+      "step": 14389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.877787583998707e-08,
+      "loss": 0.8534,
+      "step": 14390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.852413933022384e-08,
+      "loss": 0.9307,
+      "step": 14391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.827087012120648e-08,
+      "loss": 0.8872,
+      "step": 14392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.801806822484991e-08,
+      "loss": 0.9615,
+      "step": 14393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.776573365305128e-08,
+      "loss": 0.8857,
+      "step": 14394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.751386641768331e-08,
+      "loss": 0.9299,
+      "step": 14395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.726246653059653e-08,
+      "loss": 0.8878,
+      "step": 14396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.701153400361926e-08,
+      "loss": 0.8632,
+      "step": 14397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.676106884855871e-08,
+      "loss": 0.9719,
+      "step": 14398
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.651107107720212e-08,
+      "loss": 0.8853,
+      "step": 14399
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.626154070130896e-08,
+      "loss": 0.868,
+      "step": 14400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.60124777326221e-08,
+      "loss": 0.8757,
+      "step": 14401
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.576388218286101e-08,
+      "loss": 0.9373,
+      "step": 14402
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.55157540637219e-08,
+      "loss": 0.8766,
+      "step": 14403
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.526809338688101e-08,
+      "loss": 0.9766,
+      "step": 14404
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.50209001639901e-08,
+      "loss": 0.9517,
+      "step": 14405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.477417440667988e-08,
+      "loss": 0.9851,
+      "step": 14406
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.452791612656106e-08,
+      "loss": 0.9049,
+      "step": 14407
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.428212533521883e-08,
+      "loss": 0.8828,
+      "step": 14408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.403680204421948e-08,
+      "loss": 1.0421,
+      "step": 14409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.37919462651071e-08,
+      "loss": 0.9952,
+      "step": 14410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.354755800940027e-08,
+      "loss": 0.9788,
+      "step": 14411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.330363728859978e-08,
+      "loss": 0.8677,
+      "step": 14412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.306018411418202e-08,
+      "loss": 0.8816,
+      "step": 14413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.281719849760227e-08,
+      "loss": 0.9504,
+      "step": 14414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.257468045029357e-08,
+      "loss": 0.933,
+      "step": 14415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.233262998366685e-08,
+      "loss": 0.9582,
+      "step": 14416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.209104710911184e-08,
+      "loss": 0.9369,
+      "step": 14417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.184993183799614e-08,
+      "loss": 0.9432,
+      "step": 14418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.16092841816629e-08,
+      "loss": 0.8783,
+      "step": 14419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.136910415143527e-08,
+      "loss": 0.8503,
+      "step": 14420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.112939175861643e-08,
+      "loss": 0.9514,
+      "step": 14421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.089014701448404e-08,
+      "loss": 0.9094,
+      "step": 14422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.065136993029464e-08,
+      "loss": 0.8873,
+      "step": 14423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.04130605172848e-08,
+      "loss": 0.9232,
+      "step": 14424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.017521878666666e-08,
+      "loss": 0.98,
+      "step": 14425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.993784474963127e-08,
+      "loss": 0.9977,
+      "step": 14426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.970093841734748e-08,
+      "loss": 0.8439,
+      "step": 14427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.946449980096303e-08,
+      "loss": 0.876,
+      "step": 14428
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9228528911601245e-08,
+      "loss": 0.9367,
+      "step": 14429
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8993025760367695e-08,
+      "loss": 0.9081,
+      "step": 14430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8757990358342396e-08,
+      "loss": 1.0142,
+      "step": 14431
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.852342271658318e-08,
+      "loss": 0.9238,
+      "step": 14432
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.828932284612898e-08,
+      "loss": 0.9424,
+      "step": 14433
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.805569075799322e-08,
+      "loss": 0.9054,
+      "step": 14434
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7822526463169324e-08,
+      "loss": 0.8961,
+      "step": 14435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.758982997262741e-08,
+      "loss": 0.8946,
+      "step": 14436
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.735760129731871e-08,
+      "loss": 0.9559,
+      "step": 14437
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.712584044816894e-08,
+      "loss": 0.8562,
+      "step": 14438
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6894547436081584e-08,
+      "loss": 0.8098,
+      "step": 14439
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.666372227194239e-08,
+      "loss": 0.948,
+      "step": 14440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.643336496660934e-08,
+      "loss": 0.9202,
+      "step": 14441
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.620347553092376e-08,
+      "loss": 0.9521,
+      "step": 14442
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5974053975701437e-08,
+      "loss": 0.8604,
+      "step": 14443
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5745100311735964e-08,
+      "loss": 0.8943,
+      "step": 14444
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.551661454980206e-08,
+      "loss": 0.8986,
+      "step": 14445
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.528859670065112e-08,
+      "loss": 0.8927,
+      "step": 14446
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.506104677501012e-08,
+      "loss": 0.9826,
+      "step": 14447
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4833964783586045e-08,
+      "loss": 1.006,
+      "step": 14448
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4607350737064803e-08,
+      "loss": 0.9208,
+      "step": 14449
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.438120464610785e-08,
+      "loss": 0.9093,
+      "step": 14450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.415552652135669e-08,
+      "loss": 0.8439,
+      "step": 14451
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3930316373430603e-08,
+      "loss": 0.9735,
+      "step": 14452
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.370557421292555e-08,
+      "loss": 0.9132,
+      "step": 14453
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.34813000504153e-08,
+      "loss": 0.91,
+      "step": 14454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.325749389645474e-08,
+      "loss": 0.9681,
+      "step": 14455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.303415576157212e-08,
+      "loss": 0.8861,
+      "step": 14456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.281128565627791e-08,
+      "loss": 0.8441,
+      "step": 14457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.258888359105818e-08,
+      "loss": 0.872,
+      "step": 14458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.236694957637678e-08,
+      "loss": 0.8659,
+      "step": 14459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.214548362267757e-08,
+      "loss": 0.8542,
+      "step": 14460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1924485740380006e-08,
+      "loss": 0.9453,
+      "step": 14461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.170395593988242e-08,
+      "loss": 0.9721,
+      "step": 14462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1483894231562084e-08,
+      "loss": 0.9525,
+      "step": 14463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1264300625772924e-08,
+      "loss": 0.8907,
+      "step": 14464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.104517513284779e-08,
+      "loss": 0.8353,
+      "step": 14465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0826517763097326e-08,
+      "loss": 0.9536,
+      "step": 14466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0608328526807745e-08,
+      "loss": 0.9364,
+      "step": 14467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0390607434248614e-08,
+      "loss": 1.0192,
+      "step": 14468
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.017335449566285e-08,
+      "loss": 0.9789,
+      "step": 14469
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.995656972127227e-08,
+      "loss": 0.8757,
+      "step": 14470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9740253121277614e-08,
+      "loss": 0.9182,
+      "step": 14471
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.952440470585629e-08,
+      "loss": 0.8416,
+      "step": 14472
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9309024485165725e-08,
+      "loss": 0.9477,
+      "step": 14473
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.909411246934004e-08,
+      "loss": 0.8834,
+      "step": 14474
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8879668668490034e-08,
+      "loss": 0.9017,
+      "step": 14475
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8665693092707636e-08,
+      "loss": 0.8328,
+      "step": 14476
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.845218575205923e-08,
+      "loss": 0.8168,
+      "step": 14477
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.823914665659235e-08,
+      "loss": 0.9221,
+      "step": 14478
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.802657581633008e-08,
+      "loss": 0.9025,
+      "step": 14479
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.781447324127442e-08,
+      "loss": 0.8954,
+      "step": 14480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.760283894140627e-08,
+      "loss": 0.921,
+      "step": 14481
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.739167292668212e-08,
+      "loss": 0.8794,
+      "step": 14482
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.718097520703846e-08,
+      "loss": 0.9848,
+      "step": 14483
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.697074579238958e-08,
+      "loss": 0.8505,
+      "step": 14484
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6760984692627575e-08,
+      "loss": 0.9229,
+      "step": 14485
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.655169191762121e-08,
+      "loss": 0.9387,
+      "step": 14486
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6342867477218165e-08,
+      "loss": 0.8801,
+      "step": 14487
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6134511381245025e-08,
+      "loss": 0.9181,
+      "step": 14488
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5926623639505065e-08,
+      "loss": 0.9026,
+      "step": 14489
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.571920426177934e-08,
+      "loss": 0.8423,
+      "step": 14490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.551225325782893e-08,
+      "loss": 0.8928,
+      "step": 14491
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.53057706373905e-08,
+      "loss": 0.9578,
+      "step": 14492
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5099756410179604e-08,
+      "loss": 0.9697,
+      "step": 14493
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.48942105858885e-08,
+      "loss": 0.9112,
+      "step": 14494
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4689133174191654e-08,
+      "loss": 0.9504,
+      "step": 14495
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.448452418473581e-08,
+      "loss": 0.8988,
+      "step": 14496
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4280383627148815e-08,
+      "loss": 0.9157,
+      "step": 14497
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.407671151103743e-08,
+      "loss": 0.9446,
+      "step": 14498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3873507845982876e-08,
+      "loss": 0.8575,
+      "step": 14499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.367077264154751e-08,
+      "loss": 0.895,
+      "step": 14500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.346850590727148e-08,
+      "loss": 0.8967,
+      "step": 14501
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.326670765266938e-08,
+      "loss": 0.9666,
+      "step": 14502
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.306537788723919e-08,
+      "loss": 0.971,
+      "step": 14503
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2864516620452214e-08,
+      "loss": 0.8675,
+      "step": 14504
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.266412386175867e-08,
+      "loss": 0.8436,
+      "step": 14505
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2464199620589895e-08,
+      "loss": 0.909,
+      "step": 14506
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.226474390635171e-08,
+      "loss": 0.9344,
+      "step": 14507
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.206575672842883e-08,
+      "loss": 0.8772,
+      "step": 14508
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.186723809618376e-08,
+      "loss": 0.8984,
+      "step": 14509
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1669188018957915e-08,
+      "loss": 0.9707,
+      "step": 14510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1471606506069406e-08,
+      "loss": 0.8832,
+      "step": 14511
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.127449356681634e-08,
+      "loss": 1.0018,
+      "step": 14512
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.107784921047242e-08,
+      "loss": 0.9596,
+      "step": 14513
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.088167344629024e-08,
+      "loss": 0.9176,
+      "step": 14514
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.068596628350019e-08,
+      "loss": 1.026,
+      "step": 14515
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.049072773131046e-08,
+      "loss": 0.9692,
+      "step": 14516
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.029595779890816e-08,
+      "loss": 0.9325,
+      "step": 14517
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0101656495458165e-08,
+      "loss": 0.8956,
+      "step": 14518
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.990782383010316e-08,
+      "loss": 0.9114,
+      "step": 14519
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.971445981196143e-08,
+      "loss": 0.9018,
+      "step": 14520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9521564450132335e-08,
+      "loss": 0.8704,
+      "step": 14521
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9329137753691956e-08,
+      "loss": 0.8949,
+      "step": 14522
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9137179731695286e-08,
+      "loss": 0.8403,
+      "step": 14523
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8945690393172866e-08,
+      "loss": 0.935,
+      "step": 14524
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.875466974713638e-08,
+      "loss": 0.8913,
+      "step": 14525
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8564117802573076e-08,
+      "loss": 0.8847,
+      "step": 14526
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8374034568448014e-08,
+      "loss": 0.9231,
+      "step": 14527
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.818442005370737e-08,
+      "loss": 0.9642,
+      "step": 14528
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.799527426727068e-08,
+      "loss": 0.8262,
+      "step": 14529
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7806597218038585e-08,
+      "loss": 0.9948,
+      "step": 14530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.761838891488956e-08,
+      "loss": 0.9053,
+      "step": 14531
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.743064936667762e-08,
+      "loss": 0.9284,
+      "step": 14532
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.724337858223792e-08,
+      "loss": 0.9543,
+      "step": 14533
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7056576570381196e-08,
+      "loss": 0.8987,
+      "step": 14534
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6870243339898195e-08,
+      "loss": 0.9375,
+      "step": 14535
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.668437889955412e-08,
+      "loss": 1.0173,
+      "step": 14536
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.649898325809531e-08,
+      "loss": 0.975,
+      "step": 14537
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.63140564242459e-08,
+      "loss": 0.8542,
+      "step": 14538
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6129598406707824e-08,
+      "loss": 0.8351,
+      "step": 14539
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5945609214157465e-08,
+      "loss": 0.959,
+      "step": 14540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.576208885525456e-08,
+      "loss": 0.8173,
+      "step": 14541
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.557903733863222e-08,
+      "loss": 0.9995,
+      "step": 14542
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.539645467290576e-08,
+      "loss": 0.8933,
+      "step": 14543
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5214340866663865e-08,
+      "loss": 0.8996,
+      "step": 14544
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5032695928476355e-08,
+      "loss": 0.9135,
+      "step": 14545
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.485151986689195e-08,
+      "loss": 0.9011,
+      "step": 14546
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.467081269043271e-08,
+      "loss": 0.9204,
+      "step": 14547
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.449057440760073e-08,
+      "loss": 0.8687,
+      "step": 14548
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4310805026880335e-08,
+      "loss": 0.8654,
+      "step": 14549
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.413150455672698e-08,
+      "loss": 0.9014,
+      "step": 14550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.395267300557947e-08,
+      "loss": 0.9032,
+      "step": 14551
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.377431038184997e-08,
+      "loss": 0.9266,
+      "step": 14552
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.359641669393177e-08,
+      "loss": 0.8832,
+      "step": 14553
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3418991950197044e-08,
+      "loss": 0.9339,
+      "step": 14554
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.324203615899024e-08,
+      "loss": 0.9177,
+      "step": 14555
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3065549328641364e-08,
+      "loss": 0.9352,
+      "step": 14556
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.288953146745266e-08,
+      "loss": 0.8762,
+      "step": 14557
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.27139825837064e-08,
+      "loss": 0.9801,
+      "step": 14558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.253890268566262e-08,
+      "loss": 0.8864,
+      "step": 14559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2364291781560306e-08,
+      "loss": 0.9355,
+      "step": 14560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.219014987961289e-08,
+      "loss": 0.8611,
+      "step": 14561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2016476988017134e-08,
+      "loss": 0.9024,
+      "step": 14562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.184327311494317e-08,
+      "loss": 0.9642,
+      "step": 14563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1670538268540054e-08,
+      "loss": 0.8849,
+      "step": 14564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1498272456936816e-08,
+      "loss": 0.9433,
+      "step": 14565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.13264756882381e-08,
+      "loss": 0.8867,
+      "step": 14566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1155147970528545e-08,
+      "loss": 0.8375,
+      "step": 14567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.098428931186725e-08,
+      "loss": 0.8251,
+      "step": 14568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.081389972029558e-08,
+      "loss": 0.9489,
+      "step": 14569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.064397920383044e-08,
+      "loss": 0.8071,
+      "step": 14570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0474527770466553e-08,
+      "loss": 0.9505,
+      "step": 14571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.030554542817643e-08,
+      "loss": 0.9344,
+      "step": 14572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.013703218491371e-08,
+      "loss": 0.8406,
+      "step": 14573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9968988048604264e-08,
+      "loss": 0.8969,
+      "step": 14574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9801413027156223e-08,
+      "loss": 0.9114,
+      "step": 14575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9634307128454387e-08,
+      "loss": 1.0305,
+      "step": 14576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9467670360362465e-08,
+      "loss": 0.9223,
+      "step": 14577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9301502730719743e-08,
+      "loss": 0.9705,
+      "step": 14578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.913580424734441e-08,
+      "loss": 0.9142,
+      "step": 14579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8970574918035787e-08,
+      "loss": 0.8651,
+      "step": 14580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.880581475056543e-08,
+      "loss": 0.9763,
+      "step": 14581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.864152375268603e-08,
+      "loss": 0.8981,
+      "step": 14582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8477701932129175e-08,
+      "loss": 0.9708,
+      "step": 14583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8314349296603148e-08,
+      "loss": 0.9277,
+      "step": 14584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8151465853792913e-08,
+      "loss": 0.8654,
+      "step": 14585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.798905161136345e-08,
+      "loss": 0.908,
+      "step": 14586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.782710657695531e-08,
+      "loss": 0.916,
+      "step": 14587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7665630758190175e-08,
+      "loss": 1.024,
+      "step": 14588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7504624162666417e-08,
+      "loss": 0.8621,
+      "step": 14589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7344086797956858e-08,
+      "loss": 0.9314,
+      "step": 14590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7184018671617684e-08,
+      "loss": 0.949,
+      "step": 14591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7024419791179534e-08,
+      "loss": 0.9083,
+      "step": 14592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6865290164151957e-08,
+      "loss": 0.9369,
+      "step": 14593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.670662979802341e-08,
+      "loss": 0.8433,
+      "step": 14594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6548438700257918e-08,
+      "loss": 0.9214,
+      "step": 14595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6390716878298416e-08,
+      "loss": 0.9314,
+      "step": 14596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.623346433956786e-08,
+      "loss": 0.9161,
+      "step": 14597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6076681091464772e-08,
+      "loss": 0.8839,
+      "step": 14598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5920367141365475e-08,
+      "loss": 0.9688,
+      "step": 14599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.57645224966252e-08,
+      "loss": 0.914,
+      "step": 14600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5609147164576965e-08,
+      "loss": 0.8967,
+      "step": 14601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.545424115253159e-08,
+      "loss": 0.9255,
+      "step": 14602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.529980446777769e-08,
+      "loss": 0.9752,
+      "step": 14603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5145837117582784e-08,
+      "loss": 0.9325,
+      "step": 14604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4992339109189967e-08,
+      "loss": 1.018,
+      "step": 14605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.483931044982235e-08,
+      "loss": 0.8676,
+      "step": 14606
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4686751146680842e-08,
+      "loss": 0.844,
+      "step": 14607
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4534661206943032e-08,
+      "loss": 0.9262,
+      "step": 14608
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.438304063776542e-08,
+      "loss": 0.9549,
+      "step": 14609
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4231889446282296e-08,
+      "loss": 0.8875,
+      "step": 14610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4081207639604642e-08,
+      "loss": 0.8683,
+      "step": 14611
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.393099522482345e-08,
+      "loss": 0.8977,
+      "step": 14612
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.37812522090064e-08,
+      "loss": 0.9211,
+      "step": 14613
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3631978599198967e-08,
+      "loss": 0.9135,
+      "step": 14614
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3483174402425534e-08,
+      "loss": 0.8956,
+      "step": 14615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3334839625687166e-08,
+      "loss": 0.9701,
+      "step": 14616
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3186974275963837e-08,
+      "loss": 0.9909,
+      "step": 14617
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.30395783602122e-08,
+      "loss": 0.8789,
+      "step": 14618
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2892651885367822e-08,
+      "loss": 0.9585,
+      "step": 14619
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.274619485834517e-08,
+      "loss": 0.8984,
+      "step": 14620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2600207286033183e-08,
+      "loss": 0.8993,
+      "step": 14621
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2454689175303022e-08,
+      "loss": 0.922,
+      "step": 14622
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.230964053300144e-08,
+      "loss": 0.8444,
+      "step": 14623
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2165061365951866e-08,
+      "loss": 0.915,
+      "step": 14624
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.202095168095775e-08,
+      "loss": 0.9839,
+      "step": 14625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1877311484800324e-08,
+      "loss": 0.8813,
+      "step": 14626
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1734140784238632e-08,
+      "loss": 0.9794,
+      "step": 14627
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.159143958600729e-08,
+      "loss": 0.8802,
+      "step": 14628
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1449207896823144e-08,
+      "loss": 0.9345,
+      "step": 14629
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1307445723376397e-08,
+      "loss": 0.9637,
+      "step": 14630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.116615307233949e-08,
+      "loss": 0.9061,
+      "step": 14631
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.102532995035822e-08,
+      "loss": 0.9486,
+      "step": 14632
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0884976364060618e-08,
+      "loss": 0.9532,
+      "step": 14633
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0745092320049177e-08,
+      "loss": 0.9523,
+      "step": 14634
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0605677824907523e-08,
+      "loss": 0.8877,
+      "step": 14635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.046673288519374e-08,
+      "loss": 0.9035,
+      "step": 14636
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0328257507445935e-08,
+      "loss": 0.9362,
+      "step": 14637
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.019025169818112e-08,
+      "loss": 0.9008,
+      "step": 14638
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.005271546389076e-08,
+      "loss": 0.9811,
+      "step": 14639
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.991564881104857e-08,
+      "loss": 0.9573,
+      "step": 14640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.977905174610051e-08,
+      "loss": 0.8924,
+      "step": 14641
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9642924275476983e-08,
+      "loss": 0.8883,
+      "step": 14642
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.950726640558287e-08,
+      "loss": 0.877,
+      "step": 14643
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9372078142798622e-08,
+      "loss": 0.9407,
+      "step": 14644
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.923735949348804e-08,
+      "loss": 0.9871,
+      "step": 14645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.910311046398827e-08,
+      "loss": 0.9238,
+      "step": 14646
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.89693310606176e-08,
+      "loss": 1.0277,
+      "step": 14647
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.883602128966877e-08,
+      "loss": 0.912,
+      "step": 14648
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.870318115741565e-08,
+      "loss": 0.8598,
+      "step": 14649
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8570810670108795e-08,
+      "loss": 0.9569,
+      "step": 14650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.843890983397545e-08,
+      "loss": 0.9389,
+      "step": 14651
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8307478655222866e-08,
+      "loss": 0.8731,
+      "step": 14652
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.817651714003388e-08,
+      "loss": 0.9039,
+      "step": 14653
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8046025294572446e-08,
+      "loss": 0.9102,
+      "step": 14654
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7916003124978097e-08,
+      "loss": 0.9623,
+      "step": 14655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7786450637368168e-08,
+      "loss": 1.0044,
+      "step": 14656
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.765736783783778e-08,
+      "loss": 0.8652,
+      "step": 14657
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7528754732460963e-08,
+      "loss": 0.8803,
+      "step": 14658
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7400611327290652e-08,
+      "loss": 0.8717,
+      "step": 14659
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7272937628355356e-08,
+      "loss": 0.9552,
+      "step": 14660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.714573364166139e-08,
+      "loss": 0.9393,
+      "step": 14661
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7018999373196177e-08,
+      "loss": 0.8782,
+      "step": 14662
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.689273482892051e-08,
+      "loss": 0.9907,
+      "step": 14663
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.676694001477741e-08,
+      "loss": 0.9343,
+      "step": 14664
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.664161493668548e-08,
+      "loss": 0.9109,
+      "step": 14665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.651675960054111e-08,
+      "loss": 0.9737,
+      "step": 14666
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6392374012218492e-08,
+      "loss": 0.9769,
+      "step": 14667
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6268458177570724e-08,
+      "loss": 0.9166,
+      "step": 14668
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6145012102429802e-08,
+      "loss": 0.8632,
+      "step": 14669
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6022035792602197e-08,
+      "loss": 0.9271,
+      "step": 14670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.58995292538755e-08,
+      "loss": 0.8252,
+      "step": 14671
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.577749249201177e-08,
+      "loss": 0.8756,
+      "step": 14672
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5655925512756408e-08,
+      "loss": 0.9178,
+      "step": 14673
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5534828321827066e-08,
+      "loss": 0.8899,
+      "step": 14674
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5414200924923628e-08,
+      "loss": 0.8685,
+      "step": 14675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5294043327719332e-08,
+      "loss": 0.8718,
+      "step": 14676
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5174355535869655e-08,
+      "loss": 0.9058,
+      "step": 14677
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.505513755500676e-08,
+      "loss": 0.9584,
+      "step": 14678
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.493638939073838e-08,
+      "loss": 0.9088,
+      "step": 14679
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4818111048653384e-08,
+      "loss": 0.8683,
+      "step": 14680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4700302534315091e-08,
+      "loss": 0.9437,
+      "step": 14681
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4582963853269072e-08,
+      "loss": 0.9509,
+      "step": 14682
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4466095011035353e-08,
+      "loss": 0.8339,
+      "step": 14683
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4349696013112868e-08,
+      "loss": 0.9073,
+      "step": 14684
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.423376686497835e-08,
+      "loss": 0.8987,
+      "step": 14685
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.411830757208743e-08,
+      "loss": 0.9087,
+      "step": 14686
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4003318139871325e-08,
+      "loss": 0.8904,
+      "step": 14687
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3888798573741257e-08,
+      "loss": 0.9575,
+      "step": 14688
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3774748879086252e-08,
+      "loss": 0.9224,
+      "step": 14689
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3661169061272017e-08,
+      "loss": 0.8768,
+      "step": 14690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3548059125643165e-08,
+      "loss": 0.9676,
+      "step": 14691
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3435419077520995e-08,
+      "loss": 0.9555,
+      "step": 14692
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3323248922205711e-08,
+      "loss": 0.949,
+      "step": 14693
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3211548664976426e-08,
+      "loss": 0.9256,
+      "step": 14694
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3100318311086712e-08,
+      "loss": 0.9353,
+      "step": 14695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2989557865772383e-08,
+      "loss": 0.9303,
+      "step": 14696
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2879267334243717e-08,
+      "loss": 0.9109,
+      "step": 14697
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2769446721691004e-08,
+      "loss": 0.859,
+      "step": 14698
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2660096033281221e-08,
+      "loss": 0.8675,
+      "step": 14699
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2551215274158036e-08,
+      "loss": 1.01,
+      "step": 14700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2442804449447343e-08,
+      "loss": 0.8579,
+      "step": 14701
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.23348635642484e-08,
+      "loss": 0.9414,
+      "step": 14702
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2227392623641588e-08,
+      "loss": 0.8473,
+      "step": 14703
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.212039163268175e-08,
+      "loss": 0.8968,
+      "step": 14704
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2013860596403747e-08,
+      "loss": 0.9041,
+      "step": 14705
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.190779951982135e-08,
+      "loss": 0.95,
+      "step": 14706
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1802208407925008e-08,
+      "loss": 0.9245,
+      "step": 14707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1697087265681861e-08,
+      "loss": 0.9504,
+      "step": 14708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1592436098039061e-08,
+      "loss": 0.9227,
+      "step": 14709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1488254909920448e-08,
+      "loss": 0.9884,
+      "step": 14710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1384543706227658e-08,
+      "loss": 0.8433,
+      "step": 14711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1281302491840119e-08,
+      "loss": 0.937,
+      "step": 14712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1178531271617277e-08,
+      "loss": 0.937,
+      "step": 14713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1076230050394154e-08,
+      "loss": 0.8667,
+      "step": 14714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0974398832982458e-08,
+      "loss": 0.9114,
+      "step": 14715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.087303762417613e-08,
+      "loss": 0.9186,
+      "step": 14716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0772146428743579e-08,
+      "loss": 0.8852,
+      "step": 14717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.067172525143212e-08,
+      "loss": 0.8813,
+      "step": 14718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0571774096965747e-08,
+      "loss": 0.9356,
+      "step": 14719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0472292970048482e-08,
+      "loss": 0.9262,
+      "step": 14720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.037328187536102e-08,
+      "loss": 0.9017,
+      "step": 14721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0274740817561856e-08,
+      "loss": 0.899,
+      "step": 14722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0176669801287286e-08,
+      "loss": 0.9542,
+      "step": 14723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0079068831152505e-08,
+      "loss": 0.9052,
+      "step": 14724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.981937911749396e-09,
+      "loss": 1.0024,
+      "step": 14725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.885277047648744e-09,
+      "loss": 0.921,
+      "step": 14726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.789086243398028e-09,
+      "loss": 0.8291,
+      "step": 14727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.693365503523623e-09,
+      "loss": 0.884,
+      "step": 14728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.598114832529704e-09,
+      "loss": 0.868,
+      "step": 14729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.503334234897132e-09,
+      "loss": 1.0035,
+      "step": 14730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.409023715085674e-09,
+      "loss": 0.9321,
+      "step": 14731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.315183277534001e-09,
+      "loss": 0.9435,
+      "step": 14732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.22181292665747e-09,
+      "loss": 0.8552,
+      "step": 14733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.128912666848122e-09,
+      "loss": 0.9123,
+      "step": 14734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.036482502478017e-09,
+      "loss": 0.9762,
+      "step": 14735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.944522437895897e-09,
+      "loss": 1.0085,
+      "step": 14736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.853032477428303e-09,
+      "loss": 0.978,
+      "step": 14737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.762012625381789e-09,
+      "loss": 0.8907,
+      "step": 14738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.671462886037374e-09,
+      "loss": 0.9468,
+      "step": 14739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.581383263656095e-09,
+      "loss": 0.8696,
+      "step": 14740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.491773762476784e-09,
+      "loss": 0.9227,
+      "step": 14741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.402634386714958e-09,
+      "loss": 0.9869,
+      "step": 14742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.31396514056504e-09,
+      "loss": 0.9428,
+      "step": 14743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.225766028200354e-09,
+      "loss": 0.947,
+      "step": 14744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.138037053769809e-09,
+      "loss": 0.8857,
+      "step": 14745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.050778221402321e-09,
+      "loss": 0.966,
+      "step": 14746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.963989535201277e-09,
+      "loss": 0.8814,
+      "step": 14747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.877670999253405e-09,
+      "loss": 0.9572,
+      "step": 14748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.791822617617684e-09,
+      "loss": 0.7814,
+      "step": 14749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.706444394334212e-09,
+      "loss": 0.9235,
+      "step": 14750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.621536333420887e-09,
+      "loss": 0.8701,
+      "step": 14751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.537098438872292e-09,
+      "loss": 0.9189,
+      "step": 14752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.453130714661916e-09,
+      "loss": 0.8822,
+      "step": 14753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.369633164739931e-09,
+      "loss": 0.9213,
+      "step": 14754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.2866057930354174e-09,
+      "loss": 0.9148,
+      "step": 14755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.204048603455249e-09,
+      "loss": 0.9363,
+      "step": 14756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.121961599884097e-09,
+      "loss": 0.9578,
+      "step": 14757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.040344786183317e-09,
+      "loss": 0.9697,
+      "step": 14758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9591981661953915e-09,
+      "loss": 0.9057,
+      "step": 14759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8785217437350446e-09,
+      "loss": 0.9537,
+      "step": 14760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.7983155226014616e-09,
+      "loss": 0.9662,
+      "step": 14761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.71857950656718e-09,
+      "loss": 0.9232,
+      "step": 14762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.639313699384753e-09,
+      "loss": 0.9574,
+      "step": 14763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.560518104782309e-09,
+      "loss": 0.9075,
+      "step": 14764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.482192726467995e-09,
+      "loss": 0.9186,
+      "step": 14765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.40433756812886e-09,
+      "loss": 0.9645,
+      "step": 14766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.32695263342642e-09,
+      "loss": 0.9678,
+      "step": 14767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.250037926001096e-09,
+      "loss": 0.9219,
+      "step": 14768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.173593449474436e-09,
+      "loss": 0.9851,
+      "step": 14769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.097619207441341e-09,
+      "loss": 1.0223,
+      "step": 14770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.022115203477841e-09,
+      "loss": 0.9372,
+      "step": 14771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9470814411355386e-09,
+      "loss": 0.8944,
+      "step": 14772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.872517923944942e-09,
+      "loss": 0.8843,
+      "step": 14773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7984246554154685e-09,
+      "loss": 0.891,
+      "step": 14774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7248016390321075e-09,
+      "loss": 0.9372,
+      "step": 14775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6516488782609735e-09,
+      "loss": 0.9162,
+      "step": 14776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.578966376541539e-09,
+      "loss": 0.869,
+      "step": 14777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.506754137295511e-09,
+      "loss": 0.9372,
+      "step": 14778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.435012163921283e-09,
+      "loss": 1.0329,
+      "step": 14779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.363740459791711e-09,
+      "loss": 0.9072,
+      "step": 14780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.292939028263e-09,
+      "loss": 0.8924,
+      "step": 14781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.222607872664709e-09,
+      "loss": 0.906,
+      "step": 14782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.152746996307523e-09,
+      "loss": 0.88,
+      "step": 14783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0833564024777014e-09,
+      "loss": 0.9754,
+      "step": 14784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.014436094440411e-09,
+      "loss": 0.9931,
+      "step": 14785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9459860754386135e-09,
+      "loss": 0.9226,
+      "step": 14786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.878006348693065e-09,
+      "loss": 0.901,
+      "step": 14787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8104969174012084e-09,
+      "loss": 0.9989,
+      "step": 14788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7434577847416116e-09,
+      "loss": 0.9376,
+      "step": 14789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.676888953868419e-09,
+      "loss": 0.8402,
+      "step": 14790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.61079042791246e-09,
+      "loss": 0.8833,
+      "step": 14791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.545162209984577e-09,
+      "loss": 0.9449,
+      "step": 14792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.480004303172303e-09,
+      "loss": 0.9813,
+      "step": 14793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.415316710542072e-09,
+      "loss": 0.9061,
+      "step": 14794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.351099435137008e-09,
+      "loss": 0.9024,
+      "step": 14795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2873524799802444e-09,
+      "loss": 0.9127,
+      "step": 14796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2240758480682765e-09,
+      "loss": 0.9375,
+      "step": 14797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.161269542382051e-09,
+      "loss": 0.8854,
+      "step": 14798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0989335658736525e-09,
+      "loss": 0.8488,
+      "step": 14799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0370679214774e-09,
+      "loss": 0.8954,
+      "step": 14800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.975672612104298e-09,
+      "loss": 0.8641,
+      "step": 14801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.914747640643146e-09,
+      "loss": 0.9726,
+      "step": 14802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.854293009961652e-09,
+      "loss": 0.8831,
+      "step": 14803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.794308722901985e-09,
+      "loss": 0.9262,
+      "step": 14804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.734794782287443e-09,
+      "loss": 0.883,
+      "step": 14805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6757511909202293e-09,
+      "loss": 0.9732,
+      "step": 14806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.617177951575901e-09,
+      "loss": 0.8305,
+      "step": 14807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.559075067011142e-09,
+      "loss": 0.9861,
+      "step": 14808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5014425399604314e-09,
+      "loss": 0.9433,
+      "step": 14809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.444280373136044e-09,
+      "loss": 0.9033,
+      "step": 14810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3875885692269405e-09,
+      "loss": 0.945,
+      "step": 14811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3313671308998762e-09,
+      "loss": 0.9935,
+      "step": 14812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.275616060800513e-09,
+      "loss": 0.8751,
+      "step": 14813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.220335361553417e-09,
+      "loss": 0.9422,
+      "step": 14814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1655250357576216e-09,
+      "loss": 0.9284,
+      "step": 14815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1111850859932845e-09,
+      "loss": 0.9258,
+      "step": 14816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0573155148183597e-09,
+      "loss": 0.987,
+      "step": 14817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0039163247652658e-09,
+      "loss": 0.9048,
+      "step": 14818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.950987518347548e-09,
+      "loss": 0.9126,
+      "step": 14819
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8985290980565462e-09,
+      "loss": 0.9907,
+      "step": 14820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8465410663591765e-09,
+      "loss": 0.9176,
+      "step": 14821
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7950234257023702e-09,
+      "loss": 0.9607,
+      "step": 14822
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.743976178509744e-09,
+      "loss": 0.9246,
+      "step": 14823
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.693399327182711e-09,
+      "loss": 0.9535,
+      "step": 14824
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.643292874102699e-09,
+      "loss": 0.8908,
+      "step": 14825
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5936568216256008e-09,
+      "loss": 0.8813,
+      "step": 14826
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.544491172088437e-09,
+      "loss": 0.9176,
+      "step": 14827
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.495795927802691e-09,
+      "loss": 1.0429,
+      "step": 14828
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4475710910620843e-09,
+      "loss": 0.8064,
+      "step": 14829
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3998166641336917e-09,
+      "loss": 0.8671,
+      "step": 14830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.352532649264605e-09,
+      "loss": 0.9149,
+      "step": 14831
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3057190486808213e-09,
+      "loss": 0.9999,
+      "step": 14832
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2593758645839126e-09,
+      "loss": 1.0124,
+      "step": 14833
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2135030991554672e-09,
+      "loss": 0.9166,
+      "step": 14834
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.168100754553759e-09,
+      "loss": 0.9602,
+      "step": 14835
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1231688329137468e-09,
+      "loss": 0.9116,
+      "step": 14836
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.078707336350405e-09,
+      "loss": 0.9382,
+      "step": 14837
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0347162669565046e-09,
+      "loss": 0.9486,
+      "step": 14838
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.991195626801501e-09,
+      "loss": 0.908,
+      "step": 14839
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9481454179326453e-09,
+      "loss": 0.9866,
+      "step": 14840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.905565642376095e-09,
+      "loss": 0.8298,
+      "step": 14841
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.863456302134692e-09,
+      "loss": 0.9528,
+      "step": 14842
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8218173991901844e-09,
+      "loss": 0.8472,
+      "step": 14843
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7806489355021162e-09,
+      "loss": 0.893,
+      "step": 14844
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7399509130078263e-09,
+      "loss": 0.8955,
+      "step": 14845
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6997233336213392e-09,
+      "loss": 0.9628,
+      "step": 14846
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6599661992355853e-09,
+      "loss": 0.943,
+      "step": 14847
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6206795117212904e-09,
+      "loss": 0.874,
+      "step": 14848
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5818632729269755e-09,
+      "loss": 0.9393,
+      "step": 14849
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5435174846800682e-09,
+      "loss": 0.9787,
+      "step": 14850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5056421487835704e-09,
+      "loss": 0.9321,
+      "step": 14851
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.46823726701939e-09,
+      "loss": 0.9575,
+      "step": 14852
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4313028411494513e-09,
+      "loss": 0.8656,
+      "step": 14853
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3948388729090323e-09,
+      "loss": 0.9299,
+      "step": 14854
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3588453640167588e-09,
+      "loss": 0.9096,
+      "step": 14855
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3233223161635e-09,
+      "loss": 0.9279,
+      "step": 14856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2882697310212523e-09,
+      "loss": 0.9104,
+      "step": 14857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2536876102409167e-09,
+      "loss": 0.9751,
+      "step": 14858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2195759554489706e-09,
+      "loss": 0.9641,
+      "step": 14859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1859347682496858e-09,
+      "loss": 0.9253,
+      "step": 14860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1527640502262404e-09,
+      "loss": 0.9783,
+      "step": 14861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1200638029396083e-09,
+      "loss": 0.9083,
+      "step": 14862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.087834027927448e-09,
+      "loss": 0.9197,
+      "step": 14863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0560747267085448e-09,
+      "loss": 0.8837,
+      "step": 14864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0247859007750383e-09,
+      "loss": 0.9031,
+      "step": 14865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.939675516013047e-10,
+      "loss": 0.9396,
+      "step": 14866
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.63619680635075e-10,
+      "loss": 0.9732,
+      "step": 14867
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.337422893052061e-10,
+      "loss": 0.8703,
+      "step": 14868
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.043353790172405e-10,
+      "loss": 0.9039,
+      "step": 14869
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.753989511567363e-10,
+      "loss": 0.7892,
+      "step": 14870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.469330070826065e-10,
+      "loss": 0.9455,
+      "step": 14871
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.189375481360007e-10,
+      "loss": 0.8621,
+      "step": 14872
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.914125756325331e-10,
+      "loss": 0.9057,
+      "step": 14873
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.64358090868944e-10,
+      "loss": 0.88,
+      "step": 14874
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.377740951175494e-10,
+      "loss": 0.938,
+      "step": 14875
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.116605896295703e-10,
+      "loss": 0.904,
+      "step": 14876
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.860175756329134e-10,
+      "loss": 0.9211,
+      "step": 14877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.608450543343914e-10,
+      "loss": 0.922,
+      "step": 14878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.361430269197222e-10,
+      "loss": 0.9272,
+      "step": 14879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.119114945490889e-10,
+      "loss": 0.9063,
+      "step": 14880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.881504583638009e-10,
+      "loss": 0.8372,
+      "step": 14881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.648599194829629e-10,
+      "loss": 1.0141,
+      "step": 14882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.420398790001446e-10,
+      "loss": 0.9048,
+      "step": 14883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.19690337990042e-10,
+      "loss": 0.9276,
+      "step": 14884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.978112975051463e-10,
+      "loss": 0.9817,
+      "step": 14885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.764027585735243e-10,
+      "loss": 0.9359,
+      "step": 14886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.554647222032582e-10,
+      "loss": 0.9776,
+      "step": 14887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3499718937911605e-10,
+      "loss": 0.9552,
+      "step": 14888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1500016106588157e-10,
+      "loss": 0.8688,
+      "step": 14889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.954736382016933e-10,
+      "loss": 0.7797,
+      "step": 14890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.76417621706926e-10,
+      "loss": 0.9173,
+      "step": 14891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.578321124786399e-10,
+      "loss": 1.016,
+      "step": 14892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.397171113894704e-10,
+      "loss": 0.9096,
+      "step": 14893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2207261929428915e-10,
+      "loss": 0.9862,
+      "step": 14894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.048986370213225e-10,
+      "loss": 0.889,
+      "step": 14895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8819516537881286e-10,
+      "loss": 0.9895,
+      "step": 14896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.719622051539084e-10,
+      "loss": 0.8472,
+      "step": 14897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.561997571093322e-10,
+      "loss": 0.9855,
+      "step": 14898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.409078219878236e-10,
+      "loss": 0.9136,
+      "step": 14899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2608640050769682e-10,
+      "loss": 0.9117,
+      "step": 14900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1173549336617193e-10,
+      "loss": 0.8785,
+      "step": 14901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9785510124048502e-10,
+      "loss": 1.0143,
+      "step": 14902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8444522478122673e-10,
+      "loss": 0.891,
+      "step": 14903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.715058646212242e-10,
+      "loss": 0.9238,
+      "step": 14904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5903702136887965e-10,
+      "loss": 0.9468,
+      "step": 14905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4703869561039086e-10,
+      "loss": 0.8979,
+      "step": 14906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3551088790975108e-10,
+      "loss": 0.8586,
+      "step": 14907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2445359881096965e-10,
+      "loss": 0.918,
+      "step": 14908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1386682883363087e-10,
+      "loss": 0.8371,
+      "step": 14909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0375057847511472e-10,
+      "loss": 0.9081,
+      "step": 14910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.410484821281706e-11,
+      "loss": 0.8734,
+      "step": 14911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.492963849970892e-11,
+      "loss": 0.9507,
+      "step": 14912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.622494976766704e-11,
+      "loss": 0.9541,
+      "step": 14913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.79907824263637e-11,
+      "loss": 0.8739,
+      "step": 14914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.022713686215653e-11,
+      "loss": 0.9955,
+      "step": 14915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.293401344252935e-11,
+      "loss": 0.8974,
+      "step": 14916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.611141250832063e-11,
+      "loss": 0.8706,
+      "step": 14917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.975933438260526e-11,
+      "loss": 0.8346,
+      "step": 14918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3877779362923026e-11,
+      "loss": 0.9171,
+      "step": 14919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8466747725719446e-11,
+      "loss": 0.9145,
+      "step": 14920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3526239727456046e-11,
+      "loss": 0.9418,
+      "step": 14921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9056255599059215e-11,
+      "loss": 1.0084,
+      "step": 14922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5056795551471326e-11,
+      "loss": 0.8477,
+      "step": 14923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1527859771209848e-11,
+      "loss": 0.9368,
+      "step": 14924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.469448427028681e-12,
+      "loss": 0.9232,
+      "step": 14925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.881561661036372e-12,
+      "loss": 0.8659,
+      "step": 14926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.764199596467677e-12,
+      "loss": 0.9705,
+      "step": 14927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1173623310222213e-12,
+      "loss": 0.885,
+      "step": 14928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.410499435258402e-13,
+      "loss": 0.8983,
+      "step": 14929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3526248837946184e-13,
+      "loss": 0.9132,
+      "step": 14930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.945,
+      "step": 14931
+    },
+    {
+      "epoch": 1.0,
+      "step": 14931,
+      "total_flos": 4.198532813634101e+23,
+      "train_loss": 0.9905932100872594,
+      "train_runtime": 105405.2791,
+      "train_samples_per_second": 18.132,
+      "train_steps_per_second": 0.142
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 14931,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 400,
+  "total_flos": 4.198532813634101e+23,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}