diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,89478 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 14908,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.232142857142857e-06,
+      "loss": 23.5423,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.464285714285714e-06,
+      "loss": 21.9737,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.696428571428571e-06,
+      "loss": 23.5032,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.928571428571428e-06,
+      "loss": 22.1821,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1160714285714286e-05,
+      "loss": 22.3239,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3392857142857142e-05,
+      "loss": 23.0868,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5625e-05,
+      "loss": 21.9581,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7857142857142855e-05,
+      "loss": 20.1359,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0089285714285717e-05,
+      "loss": 19.9092,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.232142857142857e-05,
+      "loss": 17.235,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.455357142857143e-05,
+      "loss": 16.2576,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6785714285714284e-05,
+      "loss": 15.1736,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9017857142857146e-05,
+      "loss": 13.7366,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.125e-05,
+      "loss": 12.1639,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.348214285714286e-05,
+      "loss": 10.505,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.571428571428571e-05,
+      "loss": 9.1677,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.794642857142857e-05,
+      "loss": 8.733,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.017857142857143e-05,
+      "loss": 8.6761,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.241071428571429e-05,
+      "loss": 8.6893,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.464285714285714e-05,
+      "loss": 8.2047,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6875e-05,
+      "loss": 7.9739,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.910714285714286e-05,
+      "loss": 8.2525,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.133928571428571e-05,
+      "loss": 7.9486,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.357142857142857e-05,
+      "loss": 7.6669,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.5803571428571434e-05,
+      "loss": 7.0508,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.803571428571429e-05,
+      "loss": 7.0597,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.026785714285714e-05,
+      "loss": 7.48,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.25e-05,
+      "loss": 7.2469,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.473214285714287e-05,
+      "loss": 6.4001,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.696428571428572e-05,
+      "loss": 6.9966,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.919642857142858e-05,
+      "loss": 6.6342,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.142857142857142e-05,
+      "loss": 6.2613,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.366071428571429e-05,
+      "loss": 6.4667,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.589285714285714e-05,
+      "loss": 6.3756,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.8125e-05,
+      "loss": 6.2142,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.035714285714287e-05,
+      "loss": 6.14,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.258928571428572e-05,
+      "loss": 6.0133,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.482142857142858e-05,
+      "loss": 5.7549,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.705357142857142e-05,
+      "loss": 5.9765,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.928571428571429e-05,
+      "loss": 5.4297,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.151785714285714e-05,
+      "loss": 5.7846,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.375e-05,
+      "loss": 5.6314,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.598214285714287e-05,
+      "loss": 5.4479,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.821428571428572e-05,
+      "loss": 5.9607,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010044642857142858,
+      "loss": 5.4253,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010267857142857142,
+      "loss": 5.1347,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010491071428571429,
+      "loss": 5.3237,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010714285714285714,
+      "loss": 5.1514,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000109375,
+      "loss": 5.1648,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011160714285714287,
+      "loss": 5.222,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011383928571428572,
+      "loss": 5.212,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011607142857142858,
+      "loss": 5.1685,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011830357142857142,
+      "loss": 5.1426,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012053571428571429,
+      "loss": 4.8513,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012276785714285715,
+      "loss": 4.779,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000125,
+      "loss": 4.8051,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012723214285714285,
+      "loss": 4.7798,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012946428571428573,
+      "loss": 5.0102,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013169642857142858,
+      "loss": 4.8042,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013392857142857144,
+      "loss": 4.7554,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001361607142857143,
+      "loss": 4.7736,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013839285714285717,
+      "loss": 4.7073,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014062500000000002,
+      "loss": 4.9761,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014285714285714284,
+      "loss": 4.6607,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014508928571428572,
+      "loss": 4.6475,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014732142857142857,
+      "loss": 4.4185,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014955357142857142,
+      "loss": 4.5688,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015178571428571427,
+      "loss": 4.5744,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015401785714285715,
+      "loss": 4.506,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015625,
+      "loss": 4.4237,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015848214285714285,
+      "loss": 4.878,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016071428571428573,
+      "loss": 4.7435,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016294642857142858,
+      "loss": 4.6093,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016517857142857144,
+      "loss": 4.1746,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001674107142857143,
+      "loss": 4.3344,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016964285714285717,
+      "loss": 4.3686,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000171875,
+      "loss": 4.259,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017410714285714284,
+      "loss": 4.5007,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017633928571428572,
+      "loss": 4.0141,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017857142857142857,
+      "loss": 3.9355,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018080357142857142,
+      "loss": 4.3213,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018303571428571427,
+      "loss": 4.4184,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018526785714285715,
+      "loss": 4.1471,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001875,
+      "loss": 4.0371,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018973214285714286,
+      "loss": 4.2251,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019196428571428573,
+      "loss": 3.8151,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019419642857142859,
+      "loss": 4.3449,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019642857142857144,
+      "loss": 4.2025,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001986607142857143,
+      "loss": 4.1503,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020089285714285717,
+      "loss": 4.092,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020312500000000002,
+      "loss": 4.0687,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020535714285714284,
+      "loss": 3.931,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020758928571428572,
+      "loss": 3.9673,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020982142857142857,
+      "loss": 3.9613,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021205357142857142,
+      "loss": 3.8559,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021428571428571427,
+      "loss": 3.9518,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021651785714285715,
+      "loss": 3.8525,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021875,
+      "loss": 3.9416,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022098214285714286,
+      "loss": 3.3147,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022321428571428573,
+      "loss": 3.6974,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022544642857142859,
+      "loss": 3.7526,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022767857142857144,
+      "loss": 3.8085,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002299107142857143,
+      "loss": 3.9479,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023214285714285717,
+      "loss": 3.7228,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000234375,
+      "loss": 3.8529,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023660714285714284,
+      "loss": 3.559,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00023883928571428572,
+      "loss": 3.8216,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024107142857142857,
+      "loss": 3.6117,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024330357142857142,
+      "loss": 3.628,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002455357142857143,
+      "loss": 3.4987,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024776785714285715,
+      "loss": 3.5924,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025,
+      "loss": 3.643,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025223214285714286,
+      "loss": 3.7993,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002544642857142857,
+      "loss": 3.4902,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025669642857142856,
+      "loss": 3.5758,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025892857142857146,
+      "loss": 3.3843,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002611607142857143,
+      "loss": 3.3085,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026339285714285717,
+      "loss": 3.5703,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000265625,
+      "loss": 3.8799,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026785714285714287,
+      "loss": 3.3381,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002700892857142857,
+      "loss": 3.7035,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002723214285714286,
+      "loss": 3.4114,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002745535714285715,
+      "loss": 3.5031,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027678571428571433,
+      "loss": 3.3141,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002790178571428572,
+      "loss": 3.4635,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028125000000000003,
+      "loss": 3.5463,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028348214285714283,
+      "loss": 3.4807,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002857142857142857,
+      "loss": 3.4426,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028794642857142853,
+      "loss": 3.4393,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029017857142857144,
+      "loss": 3.5407,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002924107142857143,
+      "loss": 3.4264,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029464285714285714,
+      "loss": 3.3124,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000296875,
+      "loss": 3.3895,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029910714285714284,
+      "loss": 3.2209,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003013392857142857,
+      "loss": 3.0604,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00030357142857142855,
+      "loss": 3.2821,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00030580357142857145,
+      "loss": 3.2526,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003080357142857143,
+      "loss": 3.2898,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031026785714285716,
+      "loss": 3.3527,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003125,
+      "loss": 3.4513,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031473214285714286,
+      "loss": 3.2924,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003169642857142857,
+      "loss": 3.3233,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031919642857142856,
+      "loss": 3.1742,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 3.163,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003236607142857143,
+      "loss": 3.0874,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00032589285714285717,
+      "loss": 2.9993,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000328125,
+      "loss": 3.3561,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033035714285714287,
+      "loss": 3.1603,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003325892857142857,
+      "loss": 3.189,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003348214285714286,
+      "loss": 3.3543,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003370535714285715,
+      "loss": 3.3649,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00033928571428571433,
+      "loss": 3.1773,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003415178571428572,
+      "loss": 3.2195,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00034375,
+      "loss": 3.4011,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00034598214285714283,
+      "loss": 3.1693,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003482142857142857,
+      "loss": 3.1028,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035044642857142853,
+      "loss": 3.03,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035267857142857144,
+      "loss": 3.042,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003549107142857143,
+      "loss": 3.105,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035714285714285714,
+      "loss": 3.0853,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000359375,
+      "loss": 3.192,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00036160714285714285,
+      "loss": 3.43,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003638392857142857,
+      "loss": 2.9554,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00036607142857142855,
+      "loss": 3.3132,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00036830357142857145,
+      "loss": 3.1903,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003705357142857143,
+      "loss": 3.0269,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037276785714285716,
+      "loss": 3.073,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000375,
+      "loss": 3.1853,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037723214285714286,
+      "loss": 2.9823,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003794642857142857,
+      "loss": 3.0572,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00038169642857142856,
+      "loss": 3.0251,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00038392857142857147,
+      "loss": 3.2262,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003861607142857143,
+      "loss": 3.0564,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00038839285714285717,
+      "loss": 2.9883,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000390625,
+      "loss": 2.9816,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003928571428571429,
+      "loss": 3.1529,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003950892857142857,
+      "loss": 3.0915,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003973214285714286,
+      "loss": 2.7941,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003995535714285715,
+      "loss": 3.0548,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00040178571428571433,
+      "loss": 3.0043,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004040178571428572,
+      "loss": 3.2192,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00040625000000000004,
+      "loss": 3.131,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00040848214285714283,
+      "loss": 2.9293,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004107142857142857,
+      "loss": 2.8549,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00041294642857142854,
+      "loss": 3.0025,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00041517857142857144,
+      "loss": 2.8665,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004174107142857143,
+      "loss": 2.9625,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00041964285714285714,
+      "loss": 3.0174,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000421875,
+      "loss": 3.1326,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00042410714285714285,
+      "loss": 3.022,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004263392857142857,
+      "loss": 3.0861,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00042857142857142855,
+      "loss": 2.8071,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00043080357142857145,
+      "loss": 3.1039,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004330357142857143,
+      "loss": 3.0986,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00043526785714285716,
+      "loss": 3.1278,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004375,
+      "loss": 3.1981,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00043973214285714286,
+      "loss": 2.9265,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004419642857142857,
+      "loss": 2.7521,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00044419642857142856,
+      "loss": 2.9806,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00044642857142857147,
+      "loss": 2.9437,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004486607142857143,
+      "loss": 2.907,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00045089285714285717,
+      "loss": 2.9098,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000453125,
+      "loss": 3.0276,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004553571428571429,
+      "loss": 2.775,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004575892857142857,
+      "loss": 2.8319,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004598214285714286,
+      "loss": 2.8339,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004620535714285715,
+      "loss": 2.8258,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046428571428571433,
+      "loss": 2.9431,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004665178571428572,
+      "loss": 2.9463,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00046875,
+      "loss": 3.1181,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047098214285714283,
+      "loss": 3.0632,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004732142857142857,
+      "loss": 2.93,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047544642857142854,
+      "loss": 2.8725,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00047767857142857144,
+      "loss": 2.6522,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004799107142857143,
+      "loss": 2.7241,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00048214285714285715,
+      "loss": 2.8835,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000484375,
+      "loss": 2.843,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00048660714285714285,
+      "loss": 2.9894,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004888392857142857,
+      "loss": 2.8609,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004910714285714286,
+      "loss": 3.1577,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004933035714285715,
+      "loss": 2.9439,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004955357142857143,
+      "loss": 2.8227,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004977678571428572,
+      "loss": 2.9446,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005,
+      "loss": 2.5928,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005022321428571429,
+      "loss": 2.9863,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005044642857142857,
+      "loss": 2.8898,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005066964285714286,
+      "loss": 2.8106,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005089285714285714,
+      "loss": 2.8724,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005111607142857143,
+      "loss": 2.8162,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005133928571428571,
+      "loss": 2.9513,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000515625,
+      "loss": 2.7848,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005178571428571429,
+      "loss": 2.7283,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005200892857142857,
+      "loss": 2.6332,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005223214285714286,
+      "loss": 2.8901,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005245535714285714,
+      "loss": 2.8605,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005267857142857143,
+      "loss": 2.938,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005290178571428571,
+      "loss": 2.7746,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00053125,
+      "loss": 2.8087,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005334821428571429,
+      "loss": 2.7322,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005357142857142857,
+      "loss": 2.8182,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005379464285714286,
+      "loss": 3.0174,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005401785714285714,
+      "loss": 2.9062,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005424107142857143,
+      "loss": 2.7597,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005446428571428571,
+      "loss": 2.8165,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000546875,
+      "loss": 2.4631,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000549107142857143,
+      "loss": 2.5054,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005513392857142858,
+      "loss": 2.9488,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005535714285714287,
+      "loss": 2.8471,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005558035714285715,
+      "loss": 2.752,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005580357142857144,
+      "loss": 2.9273,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005602678571428572,
+      "loss": 2.8124,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005625000000000001,
+      "loss": 2.7849,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005647321428571429,
+      "loss": 2.672,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005669642857142857,
+      "loss": 2.7658,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005691964285714286,
+      "loss": 2.7637,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005714285714285714,
+      "loss": 2.7049,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005736607142857143,
+      "loss": 2.6897,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005758928571428571,
+      "loss": 2.9739,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000578125,
+      "loss": 2.631,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005803571428571429,
+      "loss": 2.7219,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005825892857142857,
+      "loss": 2.7563,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005848214285714286,
+      "loss": 2.7861,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005870535714285714,
+      "loss": 2.7531,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005892857142857143,
+      "loss": 2.6876,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005915178571428571,
+      "loss": 2.886,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00059375,
+      "loss": 2.6006,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005959821428571429,
+      "loss": 2.7177,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0005982142857142857,
+      "loss": 2.5827,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006004464285714286,
+      "loss": 2.7342,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006026785714285714,
+      "loss": 2.6272,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006049107142857143,
+      "loss": 2.7372,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006071428571428571,
+      "loss": 2.6258,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000609375,
+      "loss": 2.583,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006116071428571429,
+      "loss": 2.6409,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006138392857142857,
+      "loss": 2.5655,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006160714285714286,
+      "loss": 2.6666,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006183035714285714,
+      "loss": 2.6105,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006205357142857143,
+      "loss": 2.4433,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006227678571428571,
+      "loss": 2.6635,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000625,
+      "loss": 2.4914,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006272321428571429,
+      "loss": 2.735,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006294642857142857,
+      "loss": 2.5853,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006316964285714286,
+      "loss": 2.6954,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006339285714285714,
+      "loss": 2.5744,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006361607142857143,
+      "loss": 2.5541,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006383928571428571,
+      "loss": 2.6965,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000640625,
+      "loss": 2.5495,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006428571428571429,
+      "loss": 2.6131,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006450892857142857,
+      "loss": 2.7788,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006473214285714286,
+      "loss": 2.7272,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006495535714285714,
+      "loss": 2.6067,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006517857142857143,
+      "loss": 2.5303,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006540178571428571,
+      "loss": 2.6232,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00065625,
+      "loss": 2.5968,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000658482142857143,
+      "loss": 2.6931,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006607142857142857,
+      "loss": 2.7101,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006629464285714286,
+      "loss": 2.4338,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006651785714285714,
+      "loss": 2.7854,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006674107142857144,
+      "loss": 2.5225,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006696428571428571,
+      "loss": 2.5412,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006718750000000001,
+      "loss": 2.6802,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000674107142857143,
+      "loss": 2.662,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006763392857142858,
+      "loss": 2.5935,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 2.5823,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006808035714285715,
+      "loss": 2.6751,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006830357142857144,
+      "loss": 2.7816,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006852678571428572,
+      "loss": 2.4539,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006875,
+      "loss": 2.632,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006897321428571429,
+      "loss": 2.3925,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006919642857142857,
+      "loss": 2.6127,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006941964285714286,
+      "loss": 2.4367,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006964285714285714,
+      "loss": 2.5866,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0006986607142857143,
+      "loss": 2.5518,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007008928571428571,
+      "loss": 2.7327,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000703125,
+      "loss": 2.6001,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007053571428571429,
+      "loss": 2.836,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007075892857142857,
+      "loss": 2.4117,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007098214285714286,
+      "loss": 2.5975,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007120535714285714,
+      "loss": 2.2806,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007142857142857143,
+      "loss": 2.6254,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007165178571428571,
+      "loss": 2.6465,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00071875,
+      "loss": 2.5212,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007209821428571429,
+      "loss": 2.5722,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007232142857142857,
+      "loss": 2.4648,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007254464285714286,
+      "loss": 2.3957,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007276785714285714,
+      "loss": 2.6217,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007299107142857143,
+      "loss": 2.5823,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007321428571428571,
+      "loss": 2.8008,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000734375,
+      "loss": 2.6494,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007366071428571429,
+      "loss": 2.4883,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007388392857142857,
+      "loss": 2.4256,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007410714285714286,
+      "loss": 2.555,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007433035714285714,
+      "loss": 2.5239,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007455357142857143,
+      "loss": 2.5894,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007477678571428571,
+      "loss": 2.4628,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00075,
+      "loss": 2.5602,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007522321428571429,
+      "loss": 2.5775,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007544642857142857,
+      "loss": 2.5124,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007566964285714286,
+      "loss": 2.4651,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007589285714285714,
+      "loss": 2.5934,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007611607142857143,
+      "loss": 2.6254,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007633928571428571,
+      "loss": 2.5056,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000765625,
+      "loss": 2.5232,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007678571428571429,
+      "loss": 2.5368,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007700892857142857,
+      "loss": 2.4027,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007723214285714286,
+      "loss": 2.6362,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007745535714285714,
+      "loss": 2.4528,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007767857142857143,
+      "loss": 2.4653,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007790178571428571,
+      "loss": 2.5448,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00078125,
+      "loss": 2.6453,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000783482142857143,
+      "loss": 2.4912,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007857142857142857,
+      "loss": 2.4183,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007879464285714287,
+      "loss": 2.5249,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007901785714285714,
+      "loss": 2.7052,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007924107142857144,
+      "loss": 2.6698,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007946428571428572,
+      "loss": 2.524,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0007968750000000001,
+      "loss": 2.7308,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000799107142857143,
+      "loss": 2.587,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008013392857142858,
+      "loss": 2.3502,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008035714285714287,
+      "loss": 2.5218,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008058035714285715,
+      "loss": 2.5631,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008080357142857144,
+      "loss": 2.4422,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008102678571428572,
+      "loss": 2.7221,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008125000000000001,
+      "loss": 2.3997,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008147321428571429,
+      "loss": 2.4444,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008169642857142857,
+      "loss": 2.564,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008191964285714286,
+      "loss": 2.578,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008214285714285714,
+      "loss": 2.4763,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008236607142857143,
+      "loss": 2.6883,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008258928571428571,
+      "loss": 2.7086,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000828125,
+      "loss": 2.5957,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0008303571428571429,
+      "loss": 2.2965,
+      "step": 372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008325892857142857,
+      "loss": 2.4792,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008348214285714286,
+      "loss": 2.4869,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008370535714285714,
+      "loss": 2.5201,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008392857142857143,
+      "loss": 2.5675,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008415178571428571,
+      "loss": 2.5888,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00084375,
+      "loss": 2.6984,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008459821428571429,
+      "loss": 2.5497,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008482142857142857,
+      "loss": 2.5962,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008504464285714286,
+      "loss": 2.6741,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008526785714285714,
+      "loss": 2.3706,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008549107142857143,
+      "loss": 2.3629,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008571428571428571,
+      "loss": 2.5091,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000859375,
+      "loss": 2.5172,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008616071428571429,
+      "loss": 2.5093,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008638392857142857,
+      "loss": 2.2784,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008660714285714286,
+      "loss": 2.461,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008683035714285714,
+      "loss": 2.5657,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008705357142857143,
+      "loss": 2.6151,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008727678571428571,
+      "loss": 2.514,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000875,
+      "loss": 2.4133,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008772321428571429,
+      "loss": 2.3662,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008794642857142857,
+      "loss": 2.4755,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008816964285714286,
+      "loss": 2.5984,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008839285714285714,
+      "loss": 2.3503,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008861607142857143,
+      "loss": 2.5135,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008883928571428571,
+      "loss": 2.524,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000890625,
+      "loss": 2.7288,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008928571428571429,
+      "loss": 2.4293,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008950892857142857,
+      "loss": 2.5342,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008973214285714286,
+      "loss": 2.4639,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0008995535714285714,
+      "loss": 2.8095,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009017857142857143,
+      "loss": 2.6683,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009040178571428571,
+      "loss": 2.5612,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00090625,
+      "loss": 2.519,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000908482142857143,
+      "loss": 2.5996,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009107142857142857,
+      "loss": 2.3855,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009129464285714287,
+      "loss": 2.4945,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009151785714285715,
+      "loss": 2.2645,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009174107142857144,
+      "loss": 2.5823,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009196428571428572,
+      "loss": 2.746,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009218750000000001,
+      "loss": 2.4392,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000924107142857143,
+      "loss": 2.7232,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009263392857142858,
+      "loss": 2.4525,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009285714285714287,
+      "loss": 2.5486,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009308035714285715,
+      "loss": 2.5002,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009330357142857144,
+      "loss": 2.6774,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009352678571428572,
+      "loss": 2.6021,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009375,
+      "loss": 2.5209,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009397321428571429,
+      "loss": 2.3783,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009419642857142857,
+      "loss": 2.6233,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009441964285714286,
+      "loss": 2.6331,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009464285714285714,
+      "loss": 2.5018,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009486607142857143,
+      "loss": 2.5635,
+      "step": 425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009508928571428571,
+      "loss": 2.3492,
+      "step": 426
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000953125,
+      "loss": 2.5252,
+      "step": 427
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009553571428571429,
+      "loss": 2.4401,
+      "step": 428
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009575892857142857,
+      "loss": 2.3581,
+      "step": 429
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009598214285714286,
+      "loss": 2.5013,
+      "step": 430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009620535714285714,
+      "loss": 2.4846,
+      "step": 431
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009642857142857143,
+      "loss": 2.5107,
+      "step": 432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009665178571428571,
+      "loss": 2.4736,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096875,
+      "loss": 2.6173,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009709821428571429,
+      "loss": 2.4479,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009732142857142857,
+      "loss": 2.495,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009754464285714286,
+      "loss": 2.5546,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009776785714285714,
+      "loss": 2.3194,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009799107142857144,
+      "loss": 2.4878,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009821428571428572,
+      "loss": 2.559,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000984375,
+      "loss": 2.503,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000986607142857143,
+      "loss": 2.445,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009888392857142858,
+      "loss": 2.4454,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009910714285714286,
+      "loss": 2.5813,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009933035714285714,
+      "loss": 2.524,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009955357142857144,
+      "loss": 2.4963,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009977678571428572,
+      "loss": 2.4143,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.001,
+      "loss": 2.5412,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999999988199425,
+      "loss": 2.3798,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999999527977009,
+      "loss": 2.4002,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999999893794829,
+      "loss": 2.3573,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999998111908123,
+      "loss": 2.688,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999997049856548,
+      "loss": 2.3953,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999995751793615,
+      "loss": 2.5335,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999994217719382,
+      "loss": 2.4385,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999992447633924,
+      "loss": 2.3901,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999990441537324,
+      "loss": 2.4578,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999988199429677,
+      "loss": 2.4365,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999998572131109,
+      "loss": 2.4015,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999983007181676,
+      "loss": 2.5423,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999980057041568,
+      "loss": 2.2736,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999976870890901,
+      "loss": 2.6316,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999997344872983,
+      "loss": 2.473,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999996979055851,
+      "loss": 2.2187,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999996589637712,
+      "loss": 2.573,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999961766185843,
+      "loss": 2.5121,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999995739998487,
+      "loss": 2.5502,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999995279777441,
+      "loss": 2.4603,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999947959554681,
+      "loss": 2.384,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999942885325908,
+      "loss": 2.5614,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999937575088335,
+      "loss": 2.3648,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999932028842207,
+      "loss": 2.4781,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999926246587792,
+      "loss": 2.4835,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999920228325357,
+      "loss": 2.4749,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999991397405519,
+      "loss": 2.6066,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999907483777587,
+      "loss": 2.4513,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999900757492852,
+      "loss": 2.5356,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999893795201303,
+      "loss": 2.4945,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999886596903267,
+      "loss": 2.4533,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999879162599088,
+      "loss": 2.5906,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999871492289115,
+      "loss": 2.4337,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999863585973706,
+      "loss": 2.3238,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999985544365324,
+      "loss": 2.444,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999847065328099,
+      "loss": 2.5816,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999838450998678,
+      "loss": 2.3925,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999829600665384,
+      "loss": 2.5589,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999820514328636,
+      "loss": 2.5685,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999981119198886,
+      "loss": 2.534,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999801633646499,
+      "loss": 2.4616,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999791839302003,
+      "loss": 2.4395,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999781808955836,
+      "loss": 2.4813,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999771542608465,
+      "loss": 2.4472,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999761040260383,
+      "loss": 2.2543,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999975030191208,
+      "loss": 2.2921,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999739327564068,
+      "loss": 2.6044,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999972811721686,
+      "loss": 2.613,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999716670870986,
+      "loss": 2.3504,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999970498852699,
+      "loss": 2.4304,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999693070185417,
+      "loss": 2.38,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999680915846838,
+      "loss": 2.4455,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999668525511821,
+      "loss": 2.5002,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999965589918095,
+      "loss": 2.321,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999643036854825,
+      "loss": 2.3197,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999962993853405,
+      "loss": 2.5483,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999616604219243,
+      "loss": 2.4202,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999603033911037,
+      "loss": 2.3256,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999589227610072,
+      "loss": 2.6249,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999575185316995,
+      "loss": 2.4616,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999560907032472,
+      "loss": 2.5534,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999546392757178,
+      "loss": 2.3761,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999531642491794,
+      "loss": 2.4552,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999516656237022,
+      "loss": 2.5484,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999501433993567,
+      "loss": 2.5306,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999485975762145,
+      "loss": 2.4762,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999947028154349,
+      "loss": 2.5701,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999945435133834,
+      "loss": 2.4161,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999438185147448,
+      "loss": 2.5787,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999421782971574,
+      "loss": 2.4853,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999405144811497,
+      "loss": 2.5525,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009999388270668002,
+      "loss": 2.3599,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000999937116054188,
+      "loss": 2.4001,
+      "step": 521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999353814433944,
+      "loss": 2.2032,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999933623234501,
+      "loss": 2.403,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999318414275911,
+      "loss": 2.5381,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999300360227484,
+      "loss": 2.5236,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999282070200584,
+      "loss": 2.4406,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999263544196073,
+      "loss": 2.439,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999244782214827,
+      "loss": 2.3281,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999922578425773,
+      "loss": 2.4993,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999920655032568,
+      "loss": 2.5608,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999187080419584,
+      "loss": 2.6759,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999167374540362,
+      "loss": 2.4924,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999914743268894,
+      "loss": 2.4499,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999127254866267,
+      "loss": 2.3728,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999910684107329,
+      "loss": 2.3781,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999086191310974,
+      "loss": 2.2889,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999065305580292,
+      "loss": 2.6553,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999044183882234,
+      "loss": 2.3213,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999022826217793,
+      "loss": 2.4332,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009999001232587977,
+      "loss": 2.4097,
+      "step": 540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999897940299381,
+      "loss": 2.466,
+      "step": 541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998957337436316,
+      "loss": 2.4903,
+      "step": 542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999893503591654,
+      "loss": 2.3516,
+      "step": 543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998912498435535,
+      "loss": 2.5794,
+      "step": 544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998889724994367,
+      "loss": 2.3549,
+      "step": 545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998866715594105,
+      "loss": 2.6153,
+      "step": 546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998843470235838,
+      "loss": 2.6466,
+      "step": 547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998819988920665,
+      "loss": 2.4341,
+      "step": 548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998796271649692,
+      "loss": 2.3067,
+      "step": 549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999877231842404,
+      "loss": 2.6365,
+      "step": 550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998748129244838,
+      "loss": 2.5172,
+      "step": 551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998723704113229,
+      "loss": 2.5352,
+      "step": 552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998699043030368,
+      "loss": 2.3892,
+      "step": 553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998674145997415,
+      "loss": 2.5302,
+      "step": 554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998649013015547,
+      "loss": 2.4239,
+      "step": 555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999862364408595,
+      "loss": 2.5488,
+      "step": 556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998598039209822,
+      "loss": 2.4773,
+      "step": 557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999857219838837,
+      "loss": 2.324,
+      "step": 558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998546121622817,
+      "loss": 2.4954,
+      "step": 559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998519808914392,
+      "loss": 2.5223,
+      "step": 560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998493260264339,
+      "loss": 2.4518,
+      "step": 561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998466475673905,
+      "loss": 2.5147,
+      "step": 562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999843945514436,
+      "loss": 2.4376,
+      "step": 563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998412198676978,
+      "loss": 2.4286,
+      "step": 564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998384706273048,
+      "loss": 2.5882,
+      "step": 565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998356977933861,
+      "loss": 2.408,
+      "step": 566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998329013660733,
+      "loss": 2.5254,
+      "step": 567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998300813454982,
+      "loss": 2.582,
+      "step": 568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998272377317937,
+      "loss": 2.6888,
+      "step": 569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999824370525094,
+      "loss": 2.7177,
+      "step": 570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998214797255348,
+      "loss": 2.4789,
+      "step": 571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998185653332522,
+      "loss": 2.5403,
+      "step": 572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998156273483838,
+      "loss": 2.4062,
+      "step": 573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998126657710684,
+      "loss": 2.4395,
+      "step": 574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998096806014458,
+      "loss": 2.2921,
+      "step": 575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999806671839657,
+      "loss": 2.4657,
+      "step": 576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999803639485844,
+      "loss": 2.341,
+      "step": 577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009998005835401493,
+      "loss": 2.5711,
+      "step": 578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999797504002718,
+      "loss": 2.3096,
+      "step": 579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997944008736954,
+      "loss": 2.6531,
+      "step": 580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997912741532273,
+      "loss": 2.3873,
+      "step": 581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999788123841462,
+      "loss": 2.34,
+      "step": 582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997849499385478,
+      "loss": 2.5404,
+      "step": 583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997817524446347,
+      "loss": 2.3745,
+      "step": 584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997785313598732,
+      "loss": 2.2669,
+      "step": 585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999775286684416,
+      "loss": 2.5195,
+      "step": 586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999772018418416,
+      "loss": 2.3256,
+      "step": 587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997687265620274,
+      "loss": 2.4614,
+      "step": 588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997654111154055,
+      "loss": 2.4285,
+      "step": 589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999762072078707,
+      "loss": 2.3919,
+      "step": 590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997587094520892,
+      "loss": 2.3827,
+      "step": 591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997553232357112,
+      "loss": 2.4041,
+      "step": 592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997519134297326,
+      "loss": 2.5058,
+      "step": 593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997484800343143,
+      "loss": 2.5314,
+      "step": 594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997450230496189,
+      "loss": 2.3863,
+      "step": 595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997415424758088,
+      "loss": 2.4247,
+      "step": 596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997380383130487,
+      "loss": 2.5245,
+      "step": 597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999734510561504,
+      "loss": 2.3828,
+      "step": 598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997309592213414,
+      "loss": 2.6199,
+      "step": 599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999727384292728,
+      "loss": 2.4993,
+      "step": 600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997237857758327,
+      "loss": 2.4931,
+      "step": 601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999720163670826,
+      "loss": 2.2391,
+      "step": 602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999716517977878,
+      "loss": 2.4036,
+      "step": 603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997128486971614,
+      "loss": 2.4856,
+      "step": 604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999709155828849,
+      "loss": 2.4575,
+      "step": 605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997054393731154,
+      "loss": 2.344,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009997016993301359,
+      "loss": 2.3933,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996979357000868,
+      "loss": 2.4089,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996941484831463,
+      "loss": 2.5062,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996903376794928,
+      "loss": 2.4642,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996865032893061,
+      "loss": 2.4185,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996826453127674,
+      "loss": 2.472,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996787637500588,
+      "loss": 2.537,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996748586013633,
+      "loss": 2.3347,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996709298668656,
+      "loss": 2.4053,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996669775467508,
+      "loss": 2.3959,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996630016412056,
+      "loss": 2.4919,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996590021504176,
+      "loss": 2.5266,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996549790745757,
+      "loss": 2.4233,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996509324138698,
+      "loss": 2.5241,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996468621684908,
+      "loss": 2.4802,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996427683386308,
+      "loss": 2.3711,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996386509244833,
+      "loss": 2.2191,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996345099262422,
+      "loss": 2.2668,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996303453441036,
+      "loss": 2.3905,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996261571782634,
+      "loss": 2.5099,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996219454289198,
+      "loss": 2.3643,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996177100962712,
+      "loss": 2.4688,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999613451180518,
+      "loss": 2.5914,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996091686818608,
+      "loss": 2.3644,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999604862600502,
+      "loss": 2.3582,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009996005329366447,
+      "loss": 2.4796,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995961796904933,
+      "loss": 2.2328,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995918028622532,
+      "loss": 2.5331,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995874024521314,
+      "loss": 2.4804,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999582978460335,
+      "loss": 2.4227,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995785308870733,
+      "loss": 2.4048,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995740597325562,
+      "loss": 2.3677,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995695649969946,
+      "loss": 2.3796,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995650466806004,
+      "loss": 2.4502,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995605047835874,
+      "loss": 2.3236,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995559393061696,
+      "loss": 2.3907,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995513502485628,
+      "loss": 2.3967,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995467376109834,
+      "loss": 2.3486,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999542101393649,
+      "loss": 2.3072,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995374415967789,
+      "loss": 2.3168,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995327582205927,
+      "loss": 2.4954,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995280512653116,
+      "loss": 2.3361,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995233207311576,
+      "loss": 2.5075,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995185666183543,
+      "loss": 2.358,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995137889271257,
+      "loss": 2.492,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995089876576977,
+      "loss": 2.4674,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009995041628102967,
+      "loss": 2.5461,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994993143851508,
+      "loss": 2.3361,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994944423824881,
+      "loss": 2.37,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994895468025393,
+      "loss": 2.1429,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994846276455353,
+      "loss": 2.5502,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994796849117082,
+      "loss": 2.6197,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999474718601291,
+      "loss": 2.4376,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994697287145189,
+      "loss": 2.4145,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994647152516268,
+      "loss": 2.3605,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994596782128516,
+      "loss": 2.4358,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994546175984308,
+      "loss": 2.54,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994495334086039,
+      "loss": 2.3915,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00099944442564361,
+      "loss": 2.5286,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994392943036908,
+      "loss": 2.3657,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994341393890884,
+      "loss": 2.3461,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000999428960900046,
+      "loss": 2.3233,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994237588368082,
+      "loss": 2.5178,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009994185331996205,
+      "loss": 2.4329,
+      "step": 670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009994132839887294,
+      "loss": 2.4718,
+      "step": 671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999408011204383,
+      "loss": 2.6019,
+      "step": 672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009994027148468299,
+      "loss": 2.3118,
+      "step": 673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993973949163203,
+      "loss": 2.3767,
+      "step": 674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993920514131049,
+      "loss": 2.4421,
+      "step": 675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993866843374364,
+      "loss": 2.4715,
+      "step": 676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999381293689568,
+      "loss": 2.2727,
+      "step": 677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993758794697542,
+      "loss": 2.2731,
+      "step": 678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993704416782505,
+      "loss": 2.4936,
+      "step": 679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993649803153135,
+      "loss": 2.3969,
+      "step": 680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999359495381201,
+      "loss": 2.4038,
+      "step": 681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993539868761718,
+      "loss": 2.4006,
+      "step": 682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993484548004865,
+      "loss": 2.2297,
+      "step": 683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993428991544055,
+      "loss": 2.3028,
+      "step": 684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993373199381913,
+      "loss": 2.4134,
+      "step": 685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993317171521075,
+      "loss": 2.3239,
+      "step": 686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993260907964182,
+      "loss": 2.601,
+      "step": 687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999320440871389,
+      "loss": 2.296,
+      "step": 688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993147673772868,
+      "loss": 2.1944,
+      "step": 689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993090703143796,
+      "loss": 2.4366,
+      "step": 690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009993033496829358,
+      "loss": 2.5024,
+      "step": 691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992976054832257,
+      "loss": 2.4387,
+      "step": 692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992918377155202,
+      "loss": 2.3601,
+      "step": 693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999286046380092,
+      "loss": 2.3782,
+      "step": 694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992802314772143,
+      "loss": 2.2234,
+      "step": 695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992743930071612,
+      "loss": 2.4048,
+      "step": 696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992685309702088,
+      "loss": 2.38,
+      "step": 697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992626453666335,
+      "loss": 2.4981,
+      "step": 698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992567361967132,
+      "loss": 2.4671,
+      "step": 699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999250803460727,
+      "loss": 2.3246,
+      "step": 700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992448471589546,
+      "loss": 2.4781,
+      "step": 701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992388672916772,
+      "loss": 2.372,
+      "step": 702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992328638591772,
+      "loss": 2.3112,
+      "step": 703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999226836861738,
+      "loss": 2.3766,
+      "step": 704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992207862996442,
+      "loss": 2.431,
+      "step": 705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992147121731811,
+      "loss": 2.4958,
+      "step": 706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992086144826355,
+      "loss": 2.3178,
+      "step": 707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009992024932282955,
+      "loss": 2.4129,
+      "step": 708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991963484104495,
+      "loss": 2.3601,
+      "step": 709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991901800293882,
+      "loss": 2.4643,
+      "step": 710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991839880854022,
+      "loss": 2.4375,
+      "step": 711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991777725787841,
+      "loss": 2.3799,
+      "step": 712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991715335098273,
+      "loss": 2.3911,
+      "step": 713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991652708788259,
+      "loss": 2.6848,
+      "step": 714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999158984686076,
+      "loss": 2.4244,
+      "step": 715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999152674931874,
+      "loss": 2.1912,
+      "step": 716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999146341616518,
+      "loss": 2.3118,
+      "step": 717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991399847403067,
+      "loss": 2.3066,
+      "step": 718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991336043035403,
+      "loss": 2.2459,
+      "step": 719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00099912720030652,
+      "loss": 2.3511,
+      "step": 720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999120772749548,
+      "loss": 2.3319,
+      "step": 721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991143216329275,
+      "loss": 2.5849,
+      "step": 722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009991078469569635,
+      "loss": 2.3876,
+      "step": 723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999101348721961,
+      "loss": 2.5111,
+      "step": 724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990948269282273,
+      "loss": 2.5142,
+      "step": 725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00099908828157607,
+      "loss": 2.2518,
+      "step": 726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999081712665798,
+      "loss": 2.4519,
+      "step": 727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990751201977217,
+      "loss": 2.2626,
+      "step": 728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990685041721518,
+      "loss": 2.4701,
+      "step": 729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990618645894008,
+      "loss": 2.3494,
+      "step": 730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990552014497822,
+      "loss": 2.576,
+      "step": 731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990485147536104,
+      "loss": 2.3235,
+      "step": 732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000999041804501201,
+      "loss": 2.2741,
+      "step": 733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990350706928707,
+      "loss": 2.6321,
+      "step": 734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990283133289376,
+      "loss": 2.4886,
+      "step": 735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990215324097206,
+      "loss": 2.359,
+      "step": 736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990147279355394,
+      "loss": 2.4121,
+      "step": 737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990078999067157,
+      "loss": 2.3914,
+      "step": 738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009990010483235715,
+      "loss": 2.4036,
+      "step": 739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989941731864303,
+      "loss": 2.3837,
+      "step": 740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989872744956166,
+      "loss": 2.2752,
+      "step": 741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998980352251456,
+      "loss": 2.2162,
+      "step": 742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989734064542752,
+      "loss": 2.3202,
+      "step": 743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989664371044024,
+      "loss": 2.1788,
+      "step": 744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998959444202166,
+      "loss": 2.2867,
+      "step": 745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989524277478966,
+      "loss": 2.5729,
+      "step": 746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998945387741925,
+      "loss": 2.3818,
+      "step": 747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998938324184584,
+      "loss": 2.5592,
+      "step": 748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989312370762063,
+      "loss": 2.3241,
+      "step": 749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989241264171272,
+      "loss": 2.2014,
+      "step": 750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989169922076818,
+      "loss": 2.4128,
+      "step": 751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998909834448207,
+      "loss": 2.3146,
+      "step": 752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009989026531390408,
+      "loss": 2.4169,
+      "step": 753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998895448280522,
+      "loss": 2.2852,
+      "step": 754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988882198729906,
+      "loss": 2.3217,
+      "step": 755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988809679167881,
+      "loss": 2.3594,
+      "step": 756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988736924122566,
+      "loss": 2.455,
+      "step": 757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988663933597395,
+      "loss": 2.3222,
+      "step": 758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988590707595815,
+      "loss": 2.3205,
+      "step": 759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988517246121281,
+      "loss": 2.487,
+      "step": 760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998844354917726,
+      "loss": 2.5815,
+      "step": 761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988369616767236,
+      "loss": 2.4537,
+      "step": 762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988295448894692,
+      "loss": 2.5039,
+      "step": 763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988221045563132,
+      "loss": 2.4543,
+      "step": 764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988146406776064,
+      "loss": 2.3829,
+      "step": 765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009988071532537018,
+      "loss": 2.4568,
+      "step": 766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987996422849523,
+      "loss": 2.5564,
+      "step": 767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987921077717126,
+      "loss": 2.4589,
+      "step": 768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987845497143385,
+      "loss": 2.433,
+      "step": 769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987769681131865,
+      "loss": 2.456,
+      "step": 770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987693629686146,
+      "loss": 2.54,
+      "step": 771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987617342809819,
+      "loss": 2.3097,
+      "step": 772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998754082050648,
+      "loss": 2.4521,
+      "step": 773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987464062779748,
+      "loss": 2.5182,
+      "step": 774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987387069633241,
+      "loss": 2.3594,
+      "step": 775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987309841070595,
+      "loss": 2.3095,
+      "step": 776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987232377095456,
+      "loss": 2.4472,
+      "step": 777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998715467771148,
+      "loss": 2.3979,
+      "step": 778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009987076742922333,
+      "loss": 2.4797,
+      "step": 779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986998572731698,
+      "loss": 2.4328,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998692016714326,
+      "loss": 2.3976,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986841526160722,
+      "loss": 2.4363,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986762649787796,
+      "loss": 2.203,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986683538028202,
+      "loss": 2.2574,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986604190885683,
+      "loss": 2.339,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986524608363973,
+      "loss": 2.2758,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986444790466838,
+      "loss": 2.3578,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998636473719804,
+      "loss": 2.2804,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998628444856136,
+      "loss": 2.3134,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986203924560587,
+      "loss": 2.4749,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986123165199522,
+      "loss": 2.6303,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009986042170481977,
+      "loss": 2.4027,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985960940411777,
+      "loss": 2.2807,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985879474992752,
+      "loss": 2.5139,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998579777422875,
+      "loss": 2.2886,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998571583812363,
+      "loss": 2.2125,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985633666681256,
+      "loss": 2.2452,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985551259905506,
+      "loss": 2.3474,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985468617800272,
+      "loss": 2.2872,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985385740369456,
+      "loss": 2.4423,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985302627616966,
+      "loss": 2.2966,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985219279546726,
+      "loss": 2.2594,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985135696162675,
+      "loss": 2.4687,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009985051877468753,
+      "loss": 2.3751,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998496782346892,
+      "loss": 2.579,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984883534167141,
+      "loss": 2.2763,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984799009567396,
+      "loss": 2.256,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984714249673674,
+      "loss": 2.3326,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984629254489977,
+      "loss": 2.3652,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984544024020314,
+      "loss": 2.3196,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998445855826871,
+      "loss": 2.1334,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984372857239202,
+      "loss": 2.2752,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984286920935834,
+      "loss": 2.4156,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984200749362658,
+      "loss": 2.3872,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984114342523748,
+      "loss": 2.3195,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009984027700423175,
+      "loss": 2.3163,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009983940823065035,
+      "loss": 2.3873,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009983853710453425,
+      "loss": 2.3818,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000998376636259246,
+      "loss": 2.1708,
+      "step": 819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983678779486263,
+      "loss": 2.6214,
+      "step": 820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983590961138965,
+      "loss": 2.2948,
+      "step": 821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983502907554714,
+      "loss": 2.3442,
+      "step": 822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983414618737665,
+      "loss": 2.3314,
+      "step": 823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983326094691984,
+      "loss": 2.3793,
+      "step": 824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983237335421855,
+      "loss": 2.3442,
+      "step": 825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998314834093146,
+      "loss": 2.3486,
+      "step": 826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009983059111225005,
+      "loss": 2.5259,
+      "step": 827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982969646306703,
+      "loss": 2.3073,
+      "step": 828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982879946180772,
+      "loss": 2.2399,
+      "step": 829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998279001085145,
+      "loss": 2.4036,
+      "step": 830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998269984032298,
+      "loss": 2.3946,
+      "step": 831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982609434599618,
+      "loss": 2.2998,
+      "step": 832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982518793685633,
+      "loss": 2.1829,
+      "step": 833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982427917585304,
+      "loss": 2.2846,
+      "step": 834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982336806302916,
+      "loss": 2.2746,
+      "step": 835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982245459842776,
+      "loss": 2.3133,
+      "step": 836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982153878209193,
+      "loss": 2.3955,
+      "step": 837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009982062061406488,
+      "loss": 2.5602,
+      "step": 838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981970009438996,
+      "loss": 2.3163,
+      "step": 839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981877722311065,
+      "loss": 2.4572,
+      "step": 840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981785200027047,
+      "loss": 2.2981,
+      "step": 841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981692442591313,
+      "loss": 2.2023,
+      "step": 842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981599450008237,
+      "loss": 2.5053,
+      "step": 843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981506222282214,
+      "loss": 2.1842,
+      "step": 844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998141275941764,
+      "loss": 2.2118,
+      "step": 845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981319061418926,
+      "loss": 2.4243,
+      "step": 846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981225128290498,
+      "loss": 2.2583,
+      "step": 847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981130960036789,
+      "loss": 2.4427,
+      "step": 848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009981036556662244,
+      "loss": 2.2371,
+      "step": 849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998094191817132,
+      "loss": 2.299,
+      "step": 850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998084704456848,
+      "loss": 2.3164,
+      "step": 851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009980751935858207,
+      "loss": 2.3594,
+      "step": 852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998065659204499,
+      "loss": 2.4556,
+      "step": 853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009980561013133327,
+      "loss": 2.326,
+      "step": 854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009980465199127728,
+      "loss": 2.3735,
+      "step": 855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998036915003272,
+      "loss": 2.36,
+      "step": 856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009980272865852836,
+      "loss": 2.2679,
+      "step": 857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000998017634659262,
+      "loss": 2.3914,
+      "step": 858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009980079592256626,
+      "loss": 2.3557,
+      "step": 859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979982602849422,
+      "loss": 2.2645,
+      "step": 860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979885378375588,
+      "loss": 2.3938,
+      "step": 861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979787918839712,
+      "loss": 2.1438,
+      "step": 862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979690224246394,
+      "loss": 2.3651,
+      "step": 863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979592294600246,
+      "loss": 2.1261,
+      "step": 864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979494129905891,
+      "loss": 2.218,
+      "step": 865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979395730167961,
+      "loss": 2.3763,
+      "step": 866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979297095391102,
+      "loss": 2.3231,
+      "step": 867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009979198225579969,
+      "loss": 2.5148,
+      "step": 868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997909912073923,
+      "loss": 2.2873,
+      "step": 869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978999780873561,
+      "loss": 2.3982,
+      "step": 870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978900205987653,
+      "loss": 2.192,
+      "step": 871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978800396086205,
+      "loss": 2.3529,
+      "step": 872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978700351173928,
+      "loss": 2.2599,
+      "step": 873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978600071255546,
+      "loss": 2.1525,
+      "step": 874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997849955633579,
+      "loss": 2.3757,
+      "step": 875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997839880641941,
+      "loss": 2.2159,
+      "step": 876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978297821511156,
+      "loss": 2.3039,
+      "step": 877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978196601615795,
+      "loss": 2.4656,
+      "step": 878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009978095146738106,
+      "loss": 2.2246,
+      "step": 879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977993456882878,
+      "loss": 2.5253,
+      "step": 880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977891532054913,
+      "loss": 2.4511,
+      "step": 881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977789372259019,
+      "loss": 2.2358,
+      "step": 882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977686977500019,
+      "loss": 2.2584,
+      "step": 883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977584347782746,
+      "loss": 2.1327,
+      "step": 884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977481483112048,
+      "loss": 2.3668,
+      "step": 885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977378383492774,
+      "loss": 2.3216,
+      "step": 886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977275048929795,
+      "loss": 2.2061,
+      "step": 887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977171479427988,
+      "loss": 2.3254,
+      "step": 888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009977067674992242,
+      "loss": 2.4073,
+      "step": 889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976963635627454,
+      "loss": 2.413,
+      "step": 890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997685936133854,
+      "loss": 2.3674,
+      "step": 891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976754852130416,
+      "loss": 2.3436,
+      "step": 892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976650108008019,
+      "loss": 2.1598,
+      "step": 893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976545128976293,
+      "loss": 2.2881,
+      "step": 894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976439915040192,
+      "loss": 2.1731,
+      "step": 895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976334466204684,
+      "loss": 2.4558,
+      "step": 896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976228782474744,
+      "loss": 2.3862,
+      "step": 897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997612286385536,
+      "loss": 2.1739,
+      "step": 898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009976016710351537,
+      "loss": 2.4422,
+      "step": 899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997591032196828,
+      "loss": 2.3551,
+      "step": 900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975803698710612,
+      "loss": 2.3775,
+      "step": 901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975696840583569,
+      "loss": 2.3145,
+      "step": 902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975589747592193,
+      "loss": 2.5043,
+      "step": 903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975482419741536,
+      "loss": 2.2419,
+      "step": 904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975374857036667,
+      "loss": 2.4027,
+      "step": 905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975267059482665,
+      "loss": 2.253,
+      "step": 906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009975159027084614,
+      "loss": 2.3385,
+      "step": 907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997505075984762,
+      "loss": 2.443,
+      "step": 908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974942257776785,
+      "loss": 2.3418,
+      "step": 909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974833520877235,
+      "loss": 2.3774,
+      "step": 910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974724549154104,
+      "loss": 2.2068,
+      "step": 911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974615342612533,
+      "loss": 2.2128,
+      "step": 912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974505901257677,
+      "loss": 2.2288,
+      "step": 913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974396225094705,
+      "loss": 2.2961,
+      "step": 914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997428631412879,
+      "loss": 2.3733,
+      "step": 915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974176168365124,
+      "loss": 2.3182,
+      "step": 916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009974065787808901,
+      "loss": 2.4056,
+      "step": 917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973955172465335,
+      "loss": 2.3584,
+      "step": 918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973844322339648,
+      "loss": 2.3286,
+      "step": 919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973733237437067,
+      "loss": 2.165,
+      "step": 920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973621917762842,
+      "loss": 2.2143,
+      "step": 921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973510363322227,
+      "loss": 2.3143,
+      "step": 922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973398574120481,
+      "loss": 2.4066,
+      "step": 923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973286550162888,
+      "loss": 2.3695,
+      "step": 924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997317429145473,
+      "loss": 2.4505,
+      "step": 925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009973061798001313,
+      "loss": 2.4108,
+      "step": 926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997294906980794,
+      "loss": 2.4515,
+      "step": 927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972836106879935,
+      "loss": 2.4262,
+      "step": 928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972722909222628,
+      "loss": 2.2147,
+      "step": 929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972609476841367,
+      "loss": 2.2243,
+      "step": 930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972495809741503,
+      "loss": 2.1569,
+      "step": 931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00099723819079284,
+      "loss": 2.544,
+      "step": 932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972267771407438,
+      "loss": 2.3463,
+      "step": 933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009972153400184,
+      "loss": 2.0691,
+      "step": 934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997203879426349,
+      "loss": 2.3984,
+      "step": 935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971923953651313,
+      "loss": 2.415,
+      "step": 936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971808878352893,
+      "loss": 2.4851,
+      "step": 937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971693568373659,
+      "loss": 2.2125,
+      "step": 938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971578023719057,
+      "loss": 2.3701,
+      "step": 939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971462244394538,
+      "loss": 2.3137,
+      "step": 940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971346230405569,
+      "loss": 2.2661,
+      "step": 941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971229981757623,
+      "loss": 2.2694,
+      "step": 942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009971113498456192,
+      "loss": 2.1847,
+      "step": 943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970996780506772,
+      "loss": 2.1702,
+      "step": 944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970879827914873,
+      "loss": 2.3444,
+      "step": 945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970762640686014,
+      "loss": 2.5318,
+      "step": 946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970645218825727,
+      "loss": 2.358,
+      "step": 947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970527562339552,
+      "loss": 2.2887,
+      "step": 948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970409671233049,
+      "loss": 2.3007,
+      "step": 949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000997029154551178,
+      "loss": 2.2667,
+      "step": 950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970173185181318,
+      "loss": 2.2158,
+      "step": 951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009970054590247254,
+      "loss": 2.5185,
+      "step": 952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000996993576071518,
+      "loss": 2.3215,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969816696590711,
+      "loss": 2.378,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969697397879465,
+      "loss": 2.3896,
+      "step": 955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969577864587073,
+      "loss": 2.3573,
+      "step": 956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969458096719178,
+      "loss": 2.1431,
+      "step": 957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969338094281432,
+      "loss": 2.3908,
+      "step": 958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00099692178572795,
+      "loss": 2.1169,
+      "step": 959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009969097385719057,
+      "loss": 2.2252,
+      "step": 960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968976679605793,
+      "loss": 2.3274,
+      "step": 961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968855738945398,
+      "loss": 2.1953,
+      "step": 962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000996873456374359,
+      "loss": 2.3506,
+      "step": 963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968613154006081,
+      "loss": 2.3004,
+      "step": 964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968491509738605,
+      "loss": 2.5567,
+      "step": 965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968369630946907,
+      "loss": 2.3328,
+      "step": 966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968247517636736,
+      "loss": 2.2292,
+      "step": 967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968125169813854,
+      "loss": 2.3255,
+      "step": 968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009968002587484042,
+      "loss": 2.2094,
+      "step": 969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996787977065308,
+      "loss": 2.1451,
+      "step": 970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967756719326771,
+      "loss": 2.2171,
+      "step": 971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967633433510918,
+      "loss": 2.2916,
+      "step": 972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967509913211346,
+      "loss": 2.4659,
+      "step": 973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996738615843388,
+      "loss": 2.2351,
+      "step": 974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967262169184366,
+      "loss": 2.2748,
+      "step": 975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967137945468651,
+      "loss": 2.2577,
+      "step": 976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009967013487292608,
+      "loss": 2.2488,
+      "step": 977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00099668887946621,
+      "loss": 2.5995,
+      "step": 978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966763867583022,
+      "loss": 2.3669,
+      "step": 979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966638706061265,
+      "loss": 2.1459,
+      "step": 980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966513310102742,
+      "loss": 2.2538,
+      "step": 981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996638767971337,
+      "loss": 2.2506,
+      "step": 982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966261814899075,
+      "loss": 2.2923,
+      "step": 983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966135715665804,
+      "loss": 2.2161,
+      "step": 984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009966009382019509,
+      "loss": 2.4293,
+      "step": 985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965882813966146,
+      "loss": 2.3422,
+      "step": 986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965756011511698,
+      "loss": 2.4651,
+      "step": 987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965628974662145,
+      "loss": 2.5525,
+      "step": 988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965501703423484,
+      "loss": 2.3686,
+      "step": 989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965374197801727,
+      "loss": 2.279,
+      "step": 990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965246457802886,
+      "loss": 2.3763,
+      "step": 991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009965118483432995,
+      "loss": 2.2295,
+      "step": 992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964990274698093,
+      "loss": 2.3765,
+      "step": 993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964861831604232,
+      "loss": 2.2321,
+      "step": 994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964733154157474,
+      "loss": 2.1613,
+      "step": 995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964604242363897,
+      "loss": 2.2236,
+      "step": 996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964475096229581,
+      "loss": 2.3728,
+      "step": 997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964345715760623,
+      "loss": 2.2852,
+      "step": 998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964216100963133,
+      "loss": 2.4869,
+      "step": 999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009964086251843226,
+      "loss": 2.2292,
+      "step": 1000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996395616840703,
+      "loss": 2.3115,
+      "step": 1001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996382585066069,
+      "loss": 2.2565,
+      "step": 1002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009963695298610352,
+      "loss": 2.4148,
+      "step": 1003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009963564512262184,
+      "loss": 2.3729,
+      "step": 1004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009963433491622356,
+      "loss": 2.4375,
+      "step": 1005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996330223669705,
+      "loss": 2.3837,
+      "step": 1006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009963170747492468,
+      "loss": 2.24,
+      "step": 1007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009963039024014813,
+      "loss": 2.3397,
+      "step": 1008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00099629070662703,
+      "loss": 2.3639,
+      "step": 1009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996277487426516,
+      "loss": 2.2408,
+      "step": 1010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009962642448005637,
+      "loss": 2.3392,
+      "step": 1011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009962509787497975,
+      "loss": 2.1892,
+      "step": 1012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996237689274844,
+      "loss": 2.3344,
+      "step": 1013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009962243763763305,
+      "loss": 2.3598,
+      "step": 1014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996211040054885,
+      "loss": 2.223,
+      "step": 1015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961976803111376,
+      "loss": 2.3893,
+      "step": 1016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961842971457184,
+      "loss": 2.3589,
+      "step": 1017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961708905592591,
+      "loss": 2.3966,
+      "step": 1018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961574605523932,
+      "loss": 2.2831,
+      "step": 1019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996144007125754,
+      "loss": 2.2908,
+      "step": 1020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961305302799766,
+      "loss": 2.1884,
+      "step": 1021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996117030015697,
+      "loss": 2.2879,
+      "step": 1022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009961035063335531,
+      "loss": 2.3283,
+      "step": 1023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960899592341827,
+      "loss": 2.1948,
+      "step": 1024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960763887182254,
+      "loss": 2.2222,
+      "step": 1025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960627947863216,
+      "loss": 2.4252,
+      "step": 1026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960491774391133,
+      "loss": 2.1544,
+      "step": 1027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000996035536677243,
+      "loss": 2.1946,
+      "step": 1028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960218725013545,
+      "loss": 2.2032,
+      "step": 1029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009960081849120933,
+      "loss": 2.1997,
+      "step": 1030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959944739101048,
+      "loss": 2.0763,
+      "step": 1031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959807394960365,
+      "loss": 2.5151,
+      "step": 1032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959669816705368,
+      "loss": 2.394,
+      "step": 1033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959532004342549,
+      "loss": 2.2155,
+      "step": 1034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959393957878415,
+      "loss": 2.2346,
+      "step": 1035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959255677319482,
+      "loss": 2.1842,
+      "step": 1036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009959117162672275,
+      "loss": 2.1917,
+      "step": 1037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958978413943335,
+      "loss": 2.3191,
+      "step": 1038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958839431139207,
+      "loss": 2.245,
+      "step": 1039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958700214266457,
+      "loss": 2.1343,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958560763331653,
+      "loss": 2.2627,
+      "step": 1041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958421078341377,
+      "loss": 2.2528,
+      "step": 1042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958281159302223,
+      "loss": 2.1711,
+      "step": 1043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958141006220797,
+      "loss": 2.3597,
+      "step": 1044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009958000619103714,
+      "loss": 2.15,
+      "step": 1045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957859997957598,
+      "loss": 2.2333,
+      "step": 1046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000995771914278909,
+      "loss": 2.2507,
+      "step": 1047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957578053604837,
+      "loss": 2.2541,
+      "step": 1048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957436730411498,
+      "loss": 2.299,
+      "step": 1049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957295173215746,
+      "loss": 2.395,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957153382024262,
+      "loss": 2.2061,
+      "step": 1051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009957011356843738,
+      "loss": 2.2979,
+      "step": 1052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956869097680879,
+      "loss": 2.268,
+      "step": 1053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956726604542399,
+      "loss": 2.2633,
+      "step": 1054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956583877435024,
+      "loss": 2.2578,
+      "step": 1055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956440916365492,
+      "loss": 2.2341,
+      "step": 1056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000995629772134055,
+      "loss": 2.1732,
+      "step": 1057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956154292366957,
+      "loss": 2.298,
+      "step": 1058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009956010629451487,
+      "loss": 2.3006,
+      "step": 1059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955866732600915,
+      "loss": 2.3037,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955722601822036,
+      "loss": 2.2313,
+      "step": 1061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955578237121655,
+      "loss": 2.3008,
+      "step": 1062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955433638506586,
+      "loss": 2.2754,
+      "step": 1063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955288805983654,
+      "loss": 2.3511,
+      "step": 1064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009955143739559691,
+      "loss": 2.1907,
+      "step": 1065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954998439241551,
+      "loss": 2.3676,
+      "step": 1066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000995485290503609,
+      "loss": 2.3688,
+      "step": 1067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954707136950176,
+      "loss": 2.4436,
+      "step": 1068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954561134990691,
+      "loss": 2.2257,
+      "step": 1069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954414899164528,
+      "loss": 2.2963,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954268429478589,
+      "loss": 2.3756,
+      "step": 1071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009954121725939784,
+      "loss": 2.1636,
+      "step": 1072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953974788555042,
+      "loss": 2.4184,
+      "step": 1073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953827617331296,
+      "loss": 2.3951,
+      "step": 1074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953680212275496,
+      "loss": 2.3146,
+      "step": 1075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953532573394599,
+      "loss": 2.4526,
+      "step": 1076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953384700695572,
+      "loss": 2.5335,
+      "step": 1077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953236594185395,
+      "loss": 2.2273,
+      "step": 1078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009953088253871062,
+      "loss": 2.3939,
+      "step": 1079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952939679759572,
+      "loss": 2.3018,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952790871857937,
+      "loss": 2.4369,
+      "step": 1081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952641830173186,
+      "loss": 2.2472,
+      "step": 1082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000995249255471235,
+      "loss": 2.3848,
+      "step": 1083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952343045482475,
+      "loss": 2.2123,
+      "step": 1084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952193302490621,
+      "loss": 2.128,
+      "step": 1085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009952043325743853,
+      "loss": 2.1891,
+      "step": 1086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951893115249255,
+      "loss": 2.3795,
+      "step": 1087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951742671013913,
+      "loss": 2.2943,
+      "step": 1088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951591993044928,
+      "loss": 2.2077,
+      "step": 1089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951441081349414,
+      "loss": 2.3875,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951289935934496,
+      "loss": 2.3453,
+      "step": 1091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009951138556807306,
+      "loss": 2.1874,
+      "step": 1092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000995098694397499,
+      "loss": 2.3427,
+      "step": 1093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950835097444705,
+      "loss": 2.3314,
+      "step": 1094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950683017223617,
+      "loss": 2.3255,
+      "step": 1095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950530703318908,
+      "loss": 2.3399,
+      "step": 1096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950378155737763,
+      "loss": 2.1433,
+      "step": 1097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950225374487384,
+      "loss": 2.2476,
+      "step": 1098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009950072359574986,
+      "loss": 2.162,
+      "step": 1099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949919111007788,
+      "loss": 2.313,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949765628793025,
+      "loss": 2.2731,
+      "step": 1101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949611912937941,
+      "loss": 2.4579,
+      "step": 1102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949457963449792,
+      "loss": 2.3204,
+      "step": 1103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949303780335845,
+      "loss": 2.2403,
+      "step": 1104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009949149363603377,
+      "loss": 2.294,
+      "step": 1105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000994899471325968,
+      "loss": 2.2475,
+      "step": 1106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000994883982931205,
+      "loss": 2.3965,
+      "step": 1107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00099486847117678,
+      "loss": 2.4336,
+      "step": 1108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009948529360634252,
+      "loss": 2.2728,
+      "step": 1109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009948373775918735,
+      "loss": 2.2802,
+      "step": 1110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009948217957628598,
+      "loss": 2.2008,
+      "step": 1111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009948061905771196,
+      "loss": 2.2992,
+      "step": 1112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009947905620353889,
+      "loss": 2.2562,
+      "step": 1113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009947749101384062,
+      "loss": 2.2858,
+      "step": 1114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009947592348869096,
+      "loss": 2.3065,
+      "step": 1115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009947435362816395,
+      "loss": 2.2064,
+      "step": 1116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000994727814323337,
+      "loss": 2.2887,
+      "step": 1117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009947120690127437,
+      "loss": 2.0672,
+      "step": 1118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009946963003506029,
+      "loss": 2.2132,
+      "step": 1119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009946805083376592,
+      "loss": 2.2232,
+      "step": 1120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994664692974658,
+      "loss": 2.3181,
+      "step": 1121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009946488542623456,
+      "loss": 2.3335,
+      "step": 1122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099463299220147,
+      "loss": 2.369,
+      "step": 1123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009946171067927794,
+      "loss": 2.3339,
+      "step": 1124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994601198037024,
+      "loss": 2.3387,
+      "step": 1125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945852659349545,
+      "loss": 2.3356,
+      "step": 1126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945693104873232,
+      "loss": 2.2172,
+      "step": 1127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945533316948832,
+      "loss": 2.3603,
+      "step": 1128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945373295583886,
+      "loss": 2.3266,
+      "step": 1129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945213040785945,
+      "loss": 2.3038,
+      "step": 1130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009945052552562579,
+      "loss": 2.2308,
+      "step": 1131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994489183092136,
+      "loss": 2.4183,
+      "step": 1132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009944730875869874,
+      "loss": 2.2835,
+      "step": 1133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009944569687415719,
+      "loss": 2.3431,
+      "step": 1134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009944408265566505,
+      "loss": 2.2903,
+      "step": 1135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994424661032985,
+      "loss": 2.2242,
+      "step": 1136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009944084721713385,
+      "loss": 2.1801,
+      "step": 1137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009943922599724752,
+      "loss": 2.2978,
+      "step": 1138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009943760244371601,
+      "loss": 2.3163,
+      "step": 1139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099435976556616,
+      "loss": 2.2642,
+      "step": 1140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009943434833602418,
+      "loss": 2.3186,
+      "step": 1141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009943271778201745,
+      "loss": 2.4187,
+      "step": 1142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009943108489467278,
+      "loss": 2.264,
+      "step": 1143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994294496740672,
+      "loss": 2.3042,
+      "step": 1144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009942781212027793,
+      "loss": 2.2561,
+      "step": 1145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009942617223338227,
+      "loss": 1.9778,
+      "step": 1146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009942453001345762,
+      "loss": 2.3543,
+      "step": 1147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009942288546058147,
+      "loss": 2.4155,
+      "step": 1148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994212385748315,
+      "loss": 2.2648,
+      "step": 1149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994195893562854,
+      "loss": 2.4095,
+      "step": 1150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009941793780502103,
+      "loss": 2.1914,
+      "step": 1151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009941628392111635,
+      "loss": 2.2744,
+      "step": 1152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009941462770464945,
+      "loss": 2.1323,
+      "step": 1153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009941296915569847,
+      "loss": 2.3603,
+      "step": 1154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000994113082743417,
+      "loss": 2.0419,
+      "step": 1155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009940964506065756,
+      "loss": 2.1786,
+      "step": 1156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009940797951472453,
+      "loss": 2.1401,
+      "step": 1157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009940631163662128,
+      "loss": 2.3773,
+      "step": 1158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009940464142642648,
+      "loss": 2.287,
+      "step": 1159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099402968884219,
+      "loss": 2.2936,
+      "step": 1160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009940129401007777,
+      "loss": 2.1609,
+      "step": 1161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009939961680408185,
+      "loss": 2.3525,
+      "step": 1162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993979372663104,
+      "loss": 2.3478,
+      "step": 1163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009939625539684277,
+      "loss": 2.2688,
+      "step": 1164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009939457119575823,
+      "loss": 2.1884,
+      "step": 1165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009939288466313636,
+      "loss": 2.3299,
+      "step": 1166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009939119579905676,
+      "loss": 2.2883,
+      "step": 1167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009938950460359913,
+      "loss": 2.2085,
+      "step": 1168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009938781107684327,
+      "loss": 2.1552,
+      "step": 1169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993861152188692,
+      "loss": 2.4216,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009938441702975688,
+      "loss": 2.3201,
+      "step": 1171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009938271650958654,
+      "loss": 2.3036,
+      "step": 1172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993810136584384,
+      "loss": 2.4577,
+      "step": 1173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009937930847639285,
+      "loss": 2.2965,
+      "step": 1174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009937760096353039,
+      "loss": 2.378,
+      "step": 1175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993758911199316,
+      "loss": 2.2844,
+      "step": 1176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009937417894567723,
+      "loss": 2.2955,
+      "step": 1177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009937246444084804,
+      "loss": 2.06,
+      "step": 1178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099370747605525,
+      "loss": 2.3439,
+      "step": 1179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936902843978915,
+      "loss": 2.1886,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936730694372162,
+      "loss": 2.2158,
+      "step": 1181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936558311740365,
+      "loss": 2.1643,
+      "step": 1182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936385696091665,
+      "loss": 2.096,
+      "step": 1183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936212847434208,
+      "loss": 2.2298,
+      "step": 1184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009936039765776152,
+      "loss": 2.3184,
+      "step": 1185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993586645112567,
+      "loss": 2.3291,
+      "step": 1186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009935692903490939,
+      "loss": 2.2563,
+      "step": 1187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009935519122880152,
+      "loss": 2.3417,
+      "step": 1188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009935345109301514,
+      "loss": 2.2122,
+      "step": 1189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009935170862763237,
+      "loss": 2.366,
+      "step": 1190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009934996383273544,
+      "loss": 2.2468,
+      "step": 1191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009934821670840675,
+      "loss": 2.2783,
+      "step": 1192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009934646725472873,
+      "loss": 2.3283,
+      "step": 1193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009934471547178398,
+      "loss": 2.2942,
+      "step": 1194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993429613596552,
+      "loss": 1.9919,
+      "step": 1195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009934120491842512,
+      "loss": 2.1572,
+      "step": 1196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009933944614817676,
+      "loss": 2.3536,
+      "step": 1197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009933768504899305,
+      "loss": 2.2815,
+      "step": 1198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009933592162095714,
+      "loss": 2.5475,
+      "step": 1199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009933415586415228,
+      "loss": 2.1929,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993323877786618,
+      "loss": 2.4251,
+      "step": 1201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009933061736456918,
+      "loss": 2.2551,
+      "step": 1202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009932884462195799,
+      "loss": 2.2597,
+      "step": 1203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009932706955091187,
+      "loss": 2.1534,
+      "step": 1204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009932529215151465,
+      "loss": 2.2376,
+      "step": 1205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993235124238502,
+      "loss": 2.2882,
+      "step": 1206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009932173036800254,
+      "loss": 2.2296,
+      "step": 1207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931994598405575,
+      "loss": 2.3333,
+      "step": 1208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931815927209414,
+      "loss": 2.4213,
+      "step": 1209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931637023220196,
+      "loss": 2.0657,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931457886446372,
+      "loss": 2.2818,
+      "step": 1211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931278516896393,
+      "loss": 2.376,
+      "step": 1212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009931098914578728,
+      "loss": 2.3428,
+      "step": 1213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009930919079501855,
+      "loss": 2.4001,
+      "step": 1214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009930739011674262,
+      "loss": 2.2583,
+      "step": 1215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000993055871110445,
+      "loss": 2.4849,
+      "step": 1216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009930378177800926,
+      "loss": 2.2575,
+      "step": 1217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009930197411772213,
+      "loss": 2.083,
+      "step": 1218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009930016413026847,
+      "loss": 2.3595,
+      "step": 1219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009929835181573365,
+      "loss": 2.3382,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009929653717420327,
+      "loss": 2.3643,
+      "step": 1221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099294720205763,
+      "loss": 2.1692,
+      "step": 1222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009929290091049855,
+      "loss": 2.3046,
+      "step": 1223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992910792884958,
+      "loss": 2.2195,
+      "step": 1224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928925533984078,
+      "loss": 2.3266,
+      "step": 1225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928742906461956,
+      "loss": 2.3641,
+      "step": 1226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928560046291834,
+      "loss": 2.2862,
+      "step": 1227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928376953482344,
+      "loss": 2.2493,
+      "step": 1228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928193628042127,
+      "loss": 2.308,
+      "step": 1229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009928010069979838,
+      "loss": 2.2391,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009927826279304143,
+      "loss": 2.2606,
+      "step": 1231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009927642256023711,
+      "loss": 2.1794,
+      "step": 1232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009927458000147237,
+      "loss": 2.3447,
+      "step": 1233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992727351168341,
+      "loss": 2.3893,
+      "step": 1234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009927088790640944,
+      "loss": 2.2514,
+      "step": 1235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009926903837028556,
+      "loss": 2.3748,
+      "step": 1236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009926718650854977,
+      "loss": 2.1742,
+      "step": 1237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009926533232128948,
+      "loss": 2.2299,
+      "step": 1238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992634758085922,
+      "loss": 2.2055,
+      "step": 1239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009926161697054558,
+      "loss": 2.2786,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009925975580723737,
+      "loss": 2.3653,
+      "step": 1241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009925789231875538,
+      "loss": 2.3714,
+      "step": 1242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009925602650518758,
+      "loss": 2.162,
+      "step": 1243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992541583666221,
+      "loss": 2.2801,
+      "step": 1244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009925228790314705,
+      "loss": 2.183,
+      "step": 1245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009925041511485075,
+      "loss": 2.2754,
+      "step": 1246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992485400018216,
+      "loss": 2.1678,
+      "step": 1247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009924666256414811,
+      "loss": 2.3249,
+      "step": 1248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009924478280191888,
+      "loss": 2.2875,
+      "step": 1249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009924290071522267,
+      "loss": 2.2905,
+      "step": 1250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992410163041483,
+      "loss": 2.015,
+      "step": 1251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009923912956878473,
+      "loss": 2.3926,
+      "step": 1252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00099237240509221,
+      "loss": 2.3635,
+      "step": 1253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009923534912554628,
+      "loss": 2.178,
+      "step": 1254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009923345541784988,
+      "loss": 2.3082,
+      "step": 1255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009923155938622115,
+      "loss": 2.2617,
+      "step": 1256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922966103074959,
+      "loss": 2.174,
+      "step": 1257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922776035152484,
+      "loss": 2.138,
+      "step": 1258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922585734863658,
+      "loss": 2.2766,
+      "step": 1259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922395202217466,
+      "loss": 2.1942,
+      "step": 1260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922204437222899,
+      "loss": 2.0847,
+      "step": 1261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009922013439888965,
+      "loss": 2.1721,
+      "step": 1262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009921822210224676,
+      "loss": 2.2367,
+      "step": 1263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009921630748239059,
+      "loss": 2.312,
+      "step": 1264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009921439053941154,
+      "loss": 2.3255,
+      "step": 1265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009921247127340008,
+      "loss": 2.2594,
+      "step": 1266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000992105496844468,
+      "loss": 2.2895,
+      "step": 1267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000992086257726424,
+      "loss": 2.2312,
+      "step": 1268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009920669953807774,
+      "loss": 2.1516,
+      "step": 1269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009920477098084367,
+      "loss": 2.1837,
+      "step": 1270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009920284010103125,
+      "loss": 2.2177,
+      "step": 1271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009920090689873162,
+      "loss": 2.2001,
+      "step": 1272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009919897137403606,
+      "loss": 2.0805,
+      "step": 1273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991970335270359,
+      "loss": 2.1523,
+      "step": 1274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009919509335782261,
+      "loss": 2.2439,
+      "step": 1275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009919315086648782,
+      "loss": 2.4576,
+      "step": 1276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009919120605312315,
+      "loss": 2.2528,
+      "step": 1277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009918925891782043,
+      "loss": 2.3895,
+      "step": 1278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009918730946067158,
+      "loss": 2.1985,
+      "step": 1279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009918535768176859,
+      "loss": 2.3156,
+      "step": 1280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009918340358120363,
+      "loss": 2.2824,
+      "step": 1281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991814471590689,
+      "loss": 2.0896,
+      "step": 1282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009917948841545675,
+      "loss": 2.2745,
+      "step": 1283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009917752735045969,
+      "loss": 2.2473,
+      "step": 1284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009917556396417022,
+      "loss": 2.2913,
+      "step": 1285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009917359825668106,
+      "loss": 2.3006,
+      "step": 1286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009917163022808496,
+      "loss": 2.3597,
+      "step": 1287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009916965987847484,
+      "loss": 2.2717,
+      "step": 1288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009916768720794372,
+      "loss": 2.3676,
+      "step": 1289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009916571221658466,
+      "loss": 2.2685,
+      "step": 1290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009916373490449094,
+      "loss": 2.347,
+      "step": 1291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009916175527175585,
+      "loss": 2.4996,
+      "step": 1292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009915977331847287,
+      "loss": 2.1727,
+      "step": 1293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009915778904473556,
+      "loss": 2.2208,
+      "step": 1294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009915580245063753,
+      "loss": 2.2493,
+      "step": 1295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991538135362726,
+      "loss": 2.1598,
+      "step": 1296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009915182230173464,
+      "loss": 2.1439,
+      "step": 1297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991498287471176,
+      "loss": 2.3734,
+      "step": 1298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009914783287251563,
+      "loss": 2.3487,
+      "step": 1299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009914583467802295,
+      "loss": 2.2696,
+      "step": 1300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009914383416373383,
+      "loss": 2.3044,
+      "step": 1301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991418313297427,
+      "loss": 2.2355,
+      "step": 1302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009913982617614416,
+      "loss": 2.1865,
+      "step": 1303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991378187030328,
+      "loss": 2.3365,
+      "step": 1304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991358089105034,
+      "loss": 2.1264,
+      "step": 1305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009913379679865083,
+      "loss": 2.1661,
+      "step": 1306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009913178236757006,
+      "loss": 2.1944,
+      "step": 1307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009912976561735616,
+      "loss": 2.3792,
+      "step": 1308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009912774654810435,
+      "loss": 2.1901,
+      "step": 1309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009912572515990994,
+      "loss": 2.2569,
+      "step": 1310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991237014528683,
+      "loss": 2.2971,
+      "step": 1311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00099121675427075,
+      "loss": 2.2049,
+      "step": 1312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009911964708262566,
+      "loss": 2.1781,
+      "step": 1313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00099117616419616,
+      "loss": 2.1855,
+      "step": 1314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991155834381419,
+      "loss": 2.2409,
+      "step": 1315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009911354813829933,
+      "loss": 2.2178,
+      "step": 1316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009911151052018432,
+      "loss": 2.3404,
+      "step": 1317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009910947058389308,
+      "loss": 2.2273,
+      "step": 1318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991074283295219,
+      "loss": 2.3686,
+      "step": 1319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009910538375716717,
+      "loss": 2.2647,
+      "step": 1320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009910333686692539,
+      "loss": 2.4883,
+      "step": 1321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000991012876588932,
+      "loss": 2.2444,
+      "step": 1322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009909923613316731,
+      "loss": 2.4121,
+      "step": 1323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009909718228984456,
+      "loss": 2.1947,
+      "step": 1324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990951261290219,
+      "loss": 2.2545,
+      "step": 1325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009909306765079637,
+      "loss": 2.2369,
+      "step": 1326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009909100685526518,
+      "loss": 2.1201,
+      "step": 1327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009908894374252555,
+      "loss": 2.2721,
+      "step": 1328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009908687831267488,
+      "loss": 2.1851,
+      "step": 1329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990848105658107,
+      "loss": 2.1749,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009908274050203056,
+      "loss": 2.3039,
+      "step": 1331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990806681214322,
+      "loss": 2.0586,
+      "step": 1332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009907859342411344,
+      "loss": 2.3698,
+      "step": 1333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009907651641017218,
+      "loss": 2.1322,
+      "step": 1334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009907443707970651,
+      "loss": 2.2162,
+      "step": 1335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009907235543281457,
+      "loss": 2.1654,
+      "step": 1336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009907027146959456,
+      "loss": 2.2693,
+      "step": 1337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009906818519014493,
+      "loss": 2.074,
+      "step": 1338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990660965945641,
+      "loss": 2.2379,
+      "step": 1339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990640056829507,
+      "loss": 2.21,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009906191245540336,
+      "loss": 2.2726,
+      "step": 1341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009905981691202095,
+      "loss": 2.1817,
+      "step": 1342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009905771905290234,
+      "loss": 2.2571,
+      "step": 1343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990556188781466,
+      "loss": 2.1456,
+      "step": 1344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009905351638785282,
+      "loss": 2.1663,
+      "step": 1345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009905141158212027,
+      "loss": 2.2847,
+      "step": 1346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990493044610483,
+      "loss": 2.2844,
+      "step": 1347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009904719502473634,
+      "loss": 2.2998,
+      "step": 1348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009904508327328399,
+      "loss": 2.1491,
+      "step": 1349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009904296920679092,
+      "loss": 2.164,
+      "step": 1350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990408528253569,
+      "loss": 2.3221,
+      "step": 1351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009903873412908189,
+      "loss": 2.2435,
+      "step": 1352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009903661311806585,
+      "loss": 2.0843,
+      "step": 1353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009903448979240888,
+      "loss": 2.1799,
+      "step": 1354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009903236415221123,
+      "loss": 2.2302,
+      "step": 1355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009903023619757326,
+      "loss": 2.1363,
+      "step": 1356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009902810592859535,
+      "loss": 2.261,
+      "step": 1357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009902597334537813,
+      "loss": 2.2939,
+      "step": 1358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990238384480222,
+      "loss": 2.3237,
+      "step": 1359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009902170123662836,
+      "loss": 2.2639,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990195617112975,
+      "loss": 2.1644,
+      "step": 1361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009901741987213058,
+      "loss": 2.2204,
+      "step": 1362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009901527571922872,
+      "loss": 2.2726,
+      "step": 1363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009901312925269315,
+      "loss": 2.3584,
+      "step": 1364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009901098047262514,
+      "loss": 2.2057,
+      "step": 1365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009900882937912616,
+      "loss": 2.3534,
+      "step": 1366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009900667597229772,
+      "loss": 2.4342,
+      "step": 1367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009900452025224148,
+      "loss": 2.3527,
+      "step": 1368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009900236221905916,
+      "loss": 2.1799,
+      "step": 1369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000990002018728527,
+      "loss": 2.4629,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00098998039213724,
+      "loss": 2.1219,
+      "step": 1371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009899587424177518,
+      "loss": 2.1799,
+      "step": 1372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009899370695710841,
+      "loss": 2.425,
+      "step": 1373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00098991537359826,
+      "loss": 2.0044,
+      "step": 1374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009898936545003037,
+      "loss": 2.2686,
+      "step": 1375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009898719122782404,
+      "loss": 2.4419,
+      "step": 1376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009898501469330964,
+      "loss": 2.2881,
+      "step": 1377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009898283584658988,
+      "loss": 2.2303,
+      "step": 1378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989806546877676,
+      "loss": 2.3802,
+      "step": 1379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989784712169458,
+      "loss": 2.1106,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009897628543422752,
+      "loss": 2.1907,
+      "step": 1381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009897409733971596,
+      "loss": 2.1806,
+      "step": 1382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009897190693351438,
+      "loss": 2.2173,
+      "step": 1383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009896971421572613,
+      "loss": 2.4221,
+      "step": 1384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009896751918645479,
+      "loss": 2.3878,
+      "step": 1385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009896532184580393,
+      "loss": 2.2725,
+      "step": 1386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009896312219387728,
+      "loss": 2.2038,
+      "step": 1387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009896092023077864,
+      "loss": 2.1554,
+      "step": 1388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00098958715956612,
+      "loss": 2.3138,
+      "step": 1389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009895650937148137,
+      "loss": 2.1286,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989543004754909,
+      "loss": 2.3925,
+      "step": 1391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009895208926874487,
+      "loss": 2.2913,
+      "step": 1392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009894987575134768,
+      "loss": 2.2508,
+      "step": 1393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009894765992340376,
+      "loss": 2.3656,
+      "step": 1394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009894544178501773,
+      "loss": 2.1312,
+      "step": 1395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989432213362943,
+      "loss": 2.2771,
+      "step": 1396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009894099857733825,
+      "loss": 2.2653,
+      "step": 1397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009893877350825454,
+      "loss": 2.1922,
+      "step": 1398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009893654612914817,
+      "loss": 2.0419,
+      "step": 1399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009893431644012428,
+      "loss": 2.0923,
+      "step": 1400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009893208444128813,
+      "loss": 2.2899,
+      "step": 1401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009892985013274507,
+      "loss": 2.2611,
+      "step": 1402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009892761351460054,
+      "loss": 2.3373,
+      "step": 1403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009892537458696016,
+      "loss": 2.2316,
+      "step": 1404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009892313334992956,
+      "loss": 2.177,
+      "step": 1405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009892088980361458,
+      "loss": 2.1714,
+      "step": 1406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989186439481211,
+      "loss": 2.1676,
+      "step": 1407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989163957835551,
+      "loss": 2.2339,
+      "step": 1408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009891414531002277,
+      "loss": 2.3289,
+      "step": 1409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009891189252763026,
+      "loss": 2.1733,
+      "step": 1410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009890963743648395,
+      "loss": 2.2213,
+      "step": 1411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009890738003669028,
+      "loss": 2.3195,
+      "step": 1412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000989051203283558,
+      "loss": 2.2044,
+      "step": 1413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009890285831158719,
+      "loss": 2.0893,
+      "step": 1414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009890059398649119,
+      "loss": 2.0515,
+      "step": 1415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000988983273531747,
+      "loss": 2.1697,
+      "step": 1416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009889605841174472,
+      "loss": 2.2684,
+      "step": 1417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009889378716230832,
+      "loss": 2.3233,
+      "step": 1418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009889151360497274,
+      "loss": 2.3468,
+      "step": 1419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009888923773984529,
+      "loss": 2.2147,
+      "step": 1420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009888695956703337,
+      "loss": 2.2704,
+      "step": 1421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009888467908664453,
+      "loss": 2.0747,
+      "step": 1422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009888239629878644,
+      "loss": 2.1342,
+      "step": 1423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009888011120356682,
+      "loss": 2.1798,
+      "step": 1424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009887782380109354,
+      "loss": 2.2793,
+      "step": 1425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009887553409147459,
+      "loss": 2.1933,
+      "step": 1426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009887324207481802,
+      "loss": 2.2376,
+      "step": 1427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009887094775123204,
+      "loss": 2.261,
+      "step": 1428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009886865112082492,
+      "loss": 2.1818,
+      "step": 1429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988663521837051,
+      "loss": 2.1586,
+      "step": 1430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009886405093998108,
+      "loss": 2.0999,
+      "step": 1431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009886174738976147,
+      "loss": 2.2551,
+      "step": 1432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009885944153315504,
+      "loss": 2.2353,
+      "step": 1433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988571333702706,
+      "loss": 2.2482,
+      "step": 1434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988548229012171,
+      "loss": 2.1818,
+      "step": 1435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009885251012610363,
+      "loss": 2.1732,
+      "step": 1436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009885019504503932,
+      "loss": 2.1218,
+      "step": 1437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009884787765813347,
+      "loss": 2.196,
+      "step": 1438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009884555796549545,
+      "loss": 2.3598,
+      "step": 1439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988432359672348,
+      "loss": 2.1567,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009884091166346106,
+      "loss": 2.3673,
+      "step": 1441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009883858505428395,
+      "loss": 2.1727,
+      "step": 1442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009883625613981335,
+      "loss": 2.3657,
+      "step": 1443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009883392492015914,
+      "loss": 2.1079,
+      "step": 1444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009883159139543137,
+      "loss": 2.0737,
+      "step": 1445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988292555657402,
+      "loss": 2.1274,
+      "step": 1446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009882691743119587,
+      "loss": 2.1623,
+      "step": 1447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009882457699190875,
+      "loss": 2.174,
+      "step": 1448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009882223424798932,
+      "loss": 2.2547,
+      "step": 1449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009881988919954814,
+      "loss": 2.2086,
+      "step": 1450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009881754184669594,
+      "loss": 2.25,
+      "step": 1451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009881519218954348,
+      "loss": 2.1245,
+      "step": 1452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000988128402282017,
+      "loss": 2.1783,
+      "step": 1453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009881048596278163,
+      "loss": 2.2121,
+      "step": 1454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009880812939339434,
+      "loss": 2.2866,
+      "step": 1455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009880577052015112,
+      "loss": 2.2195,
+      "step": 1456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009880340934316327,
+      "loss": 2.1677,
+      "step": 1457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009880104586254231,
+      "loss": 2.3004,
+      "step": 1458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009879868007839972,
+      "loss": 2.0529,
+      "step": 1459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009879631199084722,
+      "loss": 2.1869,
+      "step": 1460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987939415999966,
+      "loss": 2.2203,
+      "step": 1461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009879156890595972,
+      "loss": 2.2755,
+      "step": 1462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009878919390884854,
+      "loss": 2.1447,
+      "step": 1463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009878681660877525,
+      "loss": 2.3234,
+      "step": 1464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009878443700585202,
+      "loss": 2.2765,
+      "step": 1465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009878205510019116,
+      "loss": 2.2448,
+      "step": 1466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009877967089190513,
+      "loss": 2.2684,
+      "step": 1467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009877728438110645,
+      "loss": 2.1596,
+      "step": 1468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009877489556790776,
+      "loss": 2.3844,
+      "step": 1469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009877250445242186,
+      "loss": 2.2593,
+      "step": 1470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009877011103476157,
+      "loss": 2.0073,
+      "step": 1471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987677153150399,
+      "loss": 2.1959,
+      "step": 1472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987653172933699,
+      "loss": 2.1647,
+      "step": 1473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987629169698648,
+      "loss": 2.1503,
+      "step": 1474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009876051434463786,
+      "loss": 2.3146,
+      "step": 1475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009875810941780251,
+      "loss": 2.2826,
+      "step": 1476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987557021894723,
+      "loss": 2.2838,
+      "step": 1477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009875329265976078,
+      "loss": 2.2218,
+      "step": 1478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009875088082878175,
+      "loss": 2.188,
+      "step": 1479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009874846669664906,
+      "loss": 2.0834,
+      "step": 1480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987460502634766,
+      "loss": 2.1409,
+      "step": 1481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987436315293785,
+      "loss": 2.2404,
+      "step": 1482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009874121049446884,
+      "loss": 2.1094,
+      "step": 1483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00098738787158862,
+      "loss": 2.2898,
+      "step": 1484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987363615226723,
+      "loss": 2.14,
+      "step": 1485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987339335860143,
+      "loss": 2.2915,
+      "step": 1486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009873150334900253,
+      "loss": 2.3103,
+      "step": 1487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009872907081175174,
+      "loss": 2.3454,
+      "step": 1488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009872663597437674,
+      "loss": 2.2464,
+      "step": 1489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009872419883699248,
+      "loss": 2.2078,
+      "step": 1490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009872175939971398,
+      "loss": 2.1471,
+      "step": 1491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987193176626564,
+      "loss": 2.2409,
+      "step": 1492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009871687362593498,
+      "loss": 2.2549,
+      "step": 1493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987144272896651,
+      "loss": 2.2656,
+      "step": 1494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987119786539622,
+      "loss": 2.1622,
+      "step": 1495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000987095277189419,
+      "loss": 2.2135,
+      "step": 1496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009870707448471987,
+      "loss": 2.2253,
+      "step": 1497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009870461895141193,
+      "loss": 2.268,
+      "step": 1498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009870216111913398,
+      "loss": 2.2812,
+      "step": 1499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00098699700988002,
+      "loss": 2.2042,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009869723855813215,
+      "loss": 2.3389,
+      "step": 1501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009869477382964064,
+      "loss": 2.3429,
+      "step": 1502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009869230680264384,
+      "loss": 2.1968,
+      "step": 1503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009868983747725818,
+      "loss": 2.3078,
+      "step": 1504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009868736585360023,
+      "loss": 2.341,
+      "step": 1505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009868489193178662,
+      "loss": 2.2088,
+      "step": 1506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009868241571193417,
+      "loss": 2.2831,
+      "step": 1507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009867993719415973,
+      "loss": 2.3367,
+      "step": 1508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009867745637858031,
+      "loss": 2.3035,
+      "step": 1509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009867497326531303,
+      "loss": 2.2992,
+      "step": 1510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009867248785447506,
+      "loss": 2.2391,
+      "step": 1511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009867000014618372,
+      "loss": 2.3106,
+      "step": 1512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009866751014055647,
+      "loss": 2.3128,
+      "step": 1513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000986650178377108,
+      "loss": 2.2832,
+      "step": 1514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000986625232377644,
+      "loss": 2.0701,
+      "step": 1515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009866002634083498,
+      "loss": 2.1214,
+      "step": 1516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000986575271470404,
+      "loss": 2.298,
+      "step": 1517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009865502565649866,
+      "loss": 2.2693,
+      "step": 1518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009865252186932784,
+      "loss": 2.355,
+      "step": 1519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009865001578564607,
+      "loss": 2.2915,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000986475074055717,
+      "loss": 2.2558,
+      "step": 1521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009864499672922306,
+      "loss": 2.1708,
+      "step": 1522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009864248375671875,
+      "loss": 2.114,
+      "step": 1523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009863996848817733,
+      "loss": 2.3015,
+      "step": 1524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009863745092371756,
+      "loss": 2.2273,
+      "step": 1525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009863493106345824,
+      "loss": 2.2507,
+      "step": 1526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009863240890751832,
+      "loss": 2.2949,
+      "step": 1527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009862988445601688,
+      "loss": 2.2931,
+      "step": 1528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009862735770907304,
+      "loss": 2.2441,
+      "step": 1529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009862482866680611,
+      "loss": 2.1926,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009862229732933545,
+      "loss": 2.4074,
+      "step": 1531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009861976369678052,
+      "loss": 2.2393,
+      "step": 1532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009861722776926094,
+      "loss": 2.2916,
+      "step": 1533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000986146895468964,
+      "loss": 2.2294,
+      "step": 1534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009861214902980673,
+      "loss": 2.2737,
+      "step": 1535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009860960621811184,
+      "loss": 2.0781,
+      "step": 1536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009860706111193172,
+      "loss": 2.4005,
+      "step": 1537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009860451371138655,
+      "loss": 2.3818,
+      "step": 1538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009860196401659657,
+      "loss": 2.3529,
+      "step": 1539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985994120276821,
+      "loss": 2.2668,
+      "step": 1540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009859685774476364,
+      "loss": 2.4158,
+      "step": 1541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009859430116796173,
+      "loss": 2.1429,
+      "step": 1542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009859174229739705,
+      "loss": 2.1287,
+      "step": 1543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985891811331904,
+      "loss": 2.2588,
+      "step": 1544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009858661767546265,
+      "loss": 2.3012,
+      "step": 1545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985840519243348,
+      "loss": 2.288,
+      "step": 1546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009858148387992798,
+      "loss": 2.2334,
+      "step": 1547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985789135423634,
+      "loss": 2.1832,
+      "step": 1548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985763409117624,
+      "loss": 2.2381,
+      "step": 1549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985737659882464,
+      "loss": 2.1884,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009857118877193692,
+      "loss": 2.1353,
+      "step": 1551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009856860926295565,
+      "loss": 2.2253,
+      "step": 1552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009856602746142432,
+      "loss": 2.1592,
+      "step": 1553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985634433674648,
+      "loss": 2.2433,
+      "step": 1554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985608569811991,
+      "loss": 2.3498,
+      "step": 1555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009855826830274925,
+      "loss": 2.0658,
+      "step": 1556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985556773322375,
+      "loss": 2.1734,
+      "step": 1557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985530840697861,
+      "loss": 2.1231,
+      "step": 1558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009855048851551747,
+      "loss": 2.1311,
+      "step": 1559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009854789066955413,
+      "loss": 2.3018,
+      "step": 1560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985452905320187,
+      "loss": 2.1718,
+      "step": 1561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009854268810303394,
+      "loss": 2.204,
+      "step": 1562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009854008338272264,
+      "loss": 2.2799,
+      "step": 1563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000985374763712078,
+      "loss": 2.3272,
+      "step": 1564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009853486706861245,
+      "loss": 2.3339,
+      "step": 1565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009853225547505977,
+      "loss": 2.1665,
+      "step": 1566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009852964159067302,
+      "loss": 2.2348,
+      "step": 1567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009852702541557558,
+      "loss": 2.1604,
+      "step": 1568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009852440694989094,
+      "loss": 2.2729,
+      "step": 1569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000985217861937427,
+      "loss": 2.1959,
+      "step": 1570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009851916314725458,
+      "loss": 1.9806,
+      "step": 1571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009851653781055036,
+      "loss": 2.1666,
+      "step": 1572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00098513910183754,
+      "loss": 2.3541,
+      "step": 1573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009851128026698952,
+      "loss": 2.1958,
+      "step": 1574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009850864806038105,
+      "loss": 2.264,
+      "step": 1575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009850601356405282,
+      "loss": 2.4301,
+      "step": 1576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000985033767781292,
+      "loss": 2.1874,
+      "step": 1577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009850073770273468,
+      "loss": 2.2556,
+      "step": 1578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009849809633799378,
+      "loss": 2.1636,
+      "step": 1579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009849545268403122,
+      "loss": 2.0699,
+      "step": 1580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009849280674097177,
+      "loss": 2.191,
+      "step": 1581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009849015850894032,
+      "loss": 2.2619,
+      "step": 1582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009848750798806185,
+      "loss": 2.2273,
+      "step": 1583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009848485517846154,
+      "loss": 2.3025,
+      "step": 1584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009848220008026455,
+      "loss": 2.1838,
+      "step": 1585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984795426935962,
+      "loss": 2.2766,
+      "step": 1586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009847688301858198,
+      "loss": 2.2021,
+      "step": 1587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009847422105534738,
+      "loss": 2.2927,
+      "step": 1588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984715568040181,
+      "loss": 2.2303,
+      "step": 1589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009846889026471984,
+      "loss": 2.3082,
+      "step": 1590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984662214375785,
+      "loss": 2.3217,
+      "step": 1591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009846355032272006,
+      "loss": 2.3696,
+      "step": 1592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984608769202706,
+      "loss": 2.1713,
+      "step": 1593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984582012303563,
+      "loss": 2.3006,
+      "step": 1594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009845552325310346,
+      "loss": 2.144,
+      "step": 1595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009845284298863849,
+      "loss": 2.2219,
+      "step": 1596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009845016043708791,
+      "loss": 2.2699,
+      "step": 1597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009844747559857835,
+      "loss": 2.3756,
+      "step": 1598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009844478847323651,
+      "loss": 2.1147,
+      "step": 1599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009844209906118924,
+      "loss": 2.2911,
+      "step": 1600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984394073625635,
+      "loss": 2.4476,
+      "step": 1601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009843671337748635,
+      "loss": 2.1421,
+      "step": 1602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009843401710608493,
+      "loss": 2.0757,
+      "step": 1603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009843131854848652,
+      "loss": 2.2115,
+      "step": 1604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984286177048185,
+      "loss": 1.9439,
+      "step": 1605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009842591457520835,
+      "loss": 2.0372,
+      "step": 1606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984232091597837,
+      "loss": 2.1614,
+      "step": 1607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009842050145867217,
+      "loss": 2.2745,
+      "step": 1608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009841779147200165,
+      "loss": 2.2279,
+      "step": 1609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009841507919990003,
+      "loss": 2.2804,
+      "step": 1610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009841236464249532,
+      "loss": 2.2205,
+      "step": 1611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009840964779991566,
+      "loss": 2.3619,
+      "step": 1612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000984069286722893,
+      "loss": 2.127,
+      "step": 1613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009840420725974461,
+      "loss": 2.2953,
+      "step": 1614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009840148356241001,
+      "loss": 2.3291,
+      "step": 1615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009839875758041409,
+      "loss": 2.2837,
+      "step": 1616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009839602931388546,
+      "loss": 2.2378,
+      "step": 1617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009839329876295299,
+      "loss": 2.4156,
+      "step": 1618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009839056592774555,
+      "loss": 2.2161,
+      "step": 1619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009838783080839209,
+      "loss": 2.091,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009838509340502175,
+      "loss": 2.2762,
+      "step": 1621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009838235371776372,
+      "loss": 2.3819,
+      "step": 1622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009837961174674734,
+      "loss": 2.3047,
+      "step": 1623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009837686749210202,
+      "loss": 2.2375,
+      "step": 1624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000983741209539573,
+      "loss": 2.1814,
+      "step": 1625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009837137213244283,
+      "loss": 2.2247,
+      "step": 1626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009836862102768837,
+      "loss": 2.3283,
+      "step": 1627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009836586763982377,
+      "loss": 2.1654,
+      "step": 1628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009836311196897895,
+      "loss": 2.1452,
+      "step": 1629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009836035401528406,
+      "loss": 2.3157,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009835759377886924,
+      "loss": 2.3835,
+      "step": 1631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000983548312598648,
+      "loss": 2.1526,
+      "step": 1632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009835206645840112,
+      "loss": 2.2091,
+      "step": 1633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009834929937460868,
+      "loss": 2.4251,
+      "step": 1634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009834653000861816,
+      "loss": 2.2984,
+      "step": 1635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009834375836056022,
+      "loss": 2.2478,
+      "step": 1636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009834098443056573,
+      "loss": 2.2006,
+      "step": 1637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009833820821876559,
+      "loss": 2.0934,
+      "step": 1638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009833542972529086,
+      "loss": 2.2026,
+      "step": 1639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009833264895027268,
+      "loss": 2.3272,
+      "step": 1640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009832986589384234,
+      "loss": 2.2536,
+      "step": 1641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009832708055613117,
+      "loss": 2.3172,
+      "step": 1642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009832429293727067,
+      "loss": 2.2682,
+      "step": 1643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009832150303739243,
+      "loss": 1.9958,
+      "step": 1644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009831871085662811,
+      "loss": 2.3356,
+      "step": 1645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009831591639510953,
+      "loss": 2.2117,
+      "step": 1646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009831311965296857,
+      "loss": 2.2176,
+      "step": 1647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009831032063033725,
+      "loss": 2.2803,
+      "step": 1648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009830751932734772,
+      "loss": 2.2783,
+      "step": 1649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009830471574413218,
+      "loss": 2.111,
+      "step": 1650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009830190988082297,
+      "loss": 2.2187,
+      "step": 1651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009829910173755252,
+      "loss": 2.3017,
+      "step": 1652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009829629131445341,
+      "loss": 2.2348,
+      "step": 1653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009829347861165828,
+      "loss": 2.2049,
+      "step": 1654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009829066362929992,
+      "loss": 2.1869,
+      "step": 1655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009828784636751115,
+      "loss": 2.211,
+      "step": 1656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009828502682642502,
+      "loss": 2.2332,
+      "step": 1657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009828220500617457,
+      "loss": 2.2078,
+      "step": 1658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009827938090689302,
+      "loss": 2.1376,
+      "step": 1659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009827655452871364,
+      "loss": 2.0788,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009827372587176988,
+      "loss": 2.2147,
+      "step": 1661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009827089493619522,
+      "loss": 2.1304,
+      "step": 1662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009826806172212334,
+      "loss": 2.2943,
+      "step": 1663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009826522622968796,
+      "loss": 2.1334,
+      "step": 1664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982623884590229,
+      "loss": 2.283,
+      "step": 1665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009825954841026208,
+      "loss": 2.16,
+      "step": 1666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009825670608353963,
+      "loss": 2.3893,
+      "step": 1667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009825386147898968,
+      "loss": 2.2324,
+      "step": 1668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009825101459674647,
+      "loss": 2.1159,
+      "step": 1669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009824816543694444,
+      "loss": 2.1323,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009824531399971804,
+      "loss": 2.1498,
+      "step": 1671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009824246028520185,
+      "loss": 2.2653,
+      "step": 1672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982396042935306,
+      "loss": 2.2207,
+      "step": 1673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982367460248391,
+      "loss": 2.1895,
+      "step": 1674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009823388547926227,
+      "loss": 2.2396,
+      "step": 1675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982310226569351,
+      "loss": 2.216,
+      "step": 1676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009822815755799277,
+      "loss": 2.3203,
+      "step": 1677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982252901825705,
+      "loss": 2.1821,
+      "step": 1678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009822242053080358,
+      "loss": 2.3221,
+      "step": 1679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009821954860282755,
+      "loss": 2.094,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009821667439877794,
+      "loss": 2.2463,
+      "step": 1681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982137979187904,
+      "loss": 2.2137,
+      "step": 1682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009821091916300075,
+      "loss": 2.1723,
+      "step": 1683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009820803813154482,
+      "loss": 2.0934,
+      "step": 1684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009820515482455864,
+      "loss": 2.4136,
+      "step": 1685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000982022692421783,
+      "loss": 2.2472,
+      "step": 1686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009819938138454,
+      "loss": 2.074,
+      "step": 1687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009819649125178004,
+      "loss": 2.237,
+      "step": 1688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009819359884403486,
+      "loss": 2.2137,
+      "step": 1689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00098190704161441,
+      "loss": 2.3083,
+      "step": 1690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009818780720413508,
+      "loss": 2.2032,
+      "step": 1691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009818490797225384,
+      "loss": 2.2062,
+      "step": 1692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009818200646593413,
+      "loss": 2.2976,
+      "step": 1693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981791026853129,
+      "loss": 2.2942,
+      "step": 1694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009817619663052724,
+      "loss": 2.2736,
+      "step": 1695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981732883017143,
+      "loss": 2.3266,
+      "step": 1696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009817037769901137,
+      "loss": 2.1289,
+      "step": 1697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009816746482255586,
+      "loss": 2.1367,
+      "step": 1698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009816454967248522,
+      "loss": 2.218,
+      "step": 1699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009816163224893708,
+      "loss": 2.4445,
+      "step": 1700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009815871255204915,
+      "loss": 2.2682,
+      "step": 1701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009815579058195922,
+      "loss": 2.3433,
+      "step": 1702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009815286633880526,
+      "loss": 2.3372,
+      "step": 1703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009814993982272524,
+      "loss": 2.095,
+      "step": 1704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009814701103385735,
+      "loss": 2.2469,
+      "step": 1705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981440799723398,
+      "loss": 2.229,
+      "step": 1706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009814114663831098,
+      "loss": 2.1602,
+      "step": 1707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981382110319093,
+      "loss": 2.1698,
+      "step": 1708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981352731532734,
+      "loss": 2.3125,
+      "step": 1709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981323330025419,
+      "loss": 2.0867,
+      "step": 1710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009812939057985356,
+      "loss": 2.0403,
+      "step": 1711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009812644588534734,
+      "loss": 2.2619,
+      "step": 1712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000981234989191622,
+      "loss": 2.1346,
+      "step": 1713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0009812054968143722,
+      "loss": 2.3065,
+      "step": 1714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009811759817231164,
+      "loss": 2.1993,
+      "step": 1715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009811464439192477,
+      "loss": 2.2575,
+      "step": 1716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009811168834041603,
+      "loss": 2.1721,
+      "step": 1717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009810873001792498,
+      "loss": 2.2499,
+      "step": 1718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009810576942459123,
+      "loss": 2.3661,
+      "step": 1719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009810280656055454,
+      "loss": 2.1631,
+      "step": 1720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009809984142595475,
+      "loss": 2.2641,
+      "step": 1721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009809687402093183,
+      "loss": 2.2157,
+      "step": 1722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009809390434562587,
+      "loss": 2.1692,
+      "step": 1723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00098090932400177,
+      "loss": 2.3663,
+      "step": 1724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009808795818472553,
+      "loss": 2.1669,
+      "step": 1725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009808498169941184,
+      "loss": 2.1959,
+      "step": 1726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009808200294437646,
+      "loss": 2.3036,
+      "step": 1727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009807902191975996,
+      "loss": 2.2023,
+      "step": 1728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009807603862570305,
+      "loss": 2.2106,
+      "step": 1729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009807305306234657,
+      "loss": 2.191,
+      "step": 1730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009807006522983143,
+      "loss": 2.2071,
+      "step": 1731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009806707512829865,
+      "loss": 2.1674,
+      "step": 1732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000980640827578894,
+      "loss": 2.1195,
+      "step": 1733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000980610881187449,
+      "loss": 2.2369,
+      "step": 1734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009805809121100652,
+      "loss": 2.1397,
+      "step": 1735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009805509203481572,
+      "loss": 2.2667,
+      "step": 1736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009805209059031408,
+      "loss": 2.0253,
+      "step": 1737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009804908687764324,
+      "loss": 2.1358,
+      "step": 1738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00098046080896945,
+      "loss": 2.1815,
+      "step": 1739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009804307264836127,
+      "loss": 2.1312,
+      "step": 1740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009804006213203402,
+      "loss": 2.2289,
+      "step": 1741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009803704934810535,
+      "loss": 2.3409,
+      "step": 1742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000980340342967175,
+      "loss": 2.3185,
+      "step": 1743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009803101697801275,
+      "loss": 2.243,
+      "step": 1744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009802799739213354,
+      "loss": 2.2332,
+      "step": 1745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009802497553922243,
+      "loss": 2.105,
+      "step": 1746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009802195141942202,
+      "loss": 2.2044,
+      "step": 1747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009801892503287506,
+      "loss": 2.1479,
+      "step": 1748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009801589637972443,
+      "loss": 2.2955,
+      "step": 1749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009801286546011305,
+      "loss": 2.2875,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009800983227418402,
+      "loss": 2.2988,
+      "step": 1751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000980067968220805,
+      "loss": 2.2993,
+      "step": 1752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009800375910394576,
+      "loss": 2.342,
+      "step": 1753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000980007191199232,
+      "loss": 2.0535,
+      "step": 1754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009799767687015631,
+      "loss": 2.228,
+      "step": 1755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000979946323547887,
+      "loss": 2.1886,
+      "step": 1756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009799158557396406,
+      "loss": 2.1357,
+      "step": 1757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009798853652782623,
+      "loss": 2.4239,
+      "step": 1758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009798548521651911,
+      "loss": 2.1893,
+      "step": 1759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009798243164018674,
+      "loss": 2.2493,
+      "step": 1760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009797937579897325,
+      "loss": 2.2752,
+      "step": 1761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009797631769302287,
+      "loss": 2.134,
+      "step": 1762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009797325732248,
+      "loss": 2.2169,
+      "step": 1763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009797019468748903,
+      "loss": 2.3929,
+      "step": 1764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009796712978819458,
+      "loss": 2.368,
+      "step": 1765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009796406262474126,
+      "loss": 2.185,
+      "step": 1766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009796099319727392,
+      "loss": 2.1591,
+      "step": 1767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000979579215059374,
+      "loss": 2.0898,
+      "step": 1768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009795484755087667,
+      "loss": 2.0898,
+      "step": 1769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009795177133223687,
+      "loss": 2.1754,
+      "step": 1770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000979486928501632,
+      "loss": 2.2381,
+      "step": 1771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009794561210480096,
+      "loss": 2.2341,
+      "step": 1772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009794252909629557,
+      "loss": 2.2321,
+      "step": 1773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009793944382479253,
+      "loss": 2.2375,
+      "step": 1774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009793635629043753,
+      "loss": 2.252,
+      "step": 1775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009793326649337626,
+      "loss": 2.142,
+      "step": 1776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009793017443375457,
+      "loss": 2.3885,
+      "step": 1777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009792708011171845,
+      "loss": 2.117,
+      "step": 1778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009792398352741392,
+      "loss": 2.1627,
+      "step": 1779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009792088468098714,
+      "loss": 2.296,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009791778357258441,
+      "loss": 2.1483,
+      "step": 1781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009791468020235211,
+      "loss": 2.1674,
+      "step": 1782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009791157457043672,
+      "loss": 2.1184,
+      "step": 1783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009790846667698482,
+      "loss": 2.2349,
+      "step": 1784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009790535652214312,
+      "loss": 2.2357,
+      "step": 1785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000979022441060584,
+      "loss": 2.2038,
+      "step": 1786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009789912942887765,
+      "loss": 2.1491,
+      "step": 1787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978960124907478,
+      "loss": 2.1035,
+      "step": 1788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009789289329181603,
+      "loss": 2.0097,
+      "step": 1789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009788977183222956,
+      "loss": 2.4007,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009788664811213573,
+      "loss": 2.2177,
+      "step": 1791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009788352213168196,
+      "loss": 1.9523,
+      "step": 1792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009788039389101587,
+      "loss": 2.3127,
+      "step": 1793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009787726339028505,
+      "loss": 2.2351,
+      "step": 1794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978741306296373,
+      "loss": 2.2489,
+      "step": 1795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978709956092205,
+      "loss": 2.2802,
+      "step": 1796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978678583291826,
+      "loss": 2.3406,
+      "step": 1797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009786471878967173,
+      "loss": 2.1649,
+      "step": 1798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009786157699083605,
+      "loss": 2.3095,
+      "step": 1799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009785843293282387,
+      "loss": 2.3402,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978552866157836,
+      "loss": 2.2165,
+      "step": 1801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009785213803986375,
+      "loss": 2.2299,
+      "step": 1802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009784898720521294,
+      "loss": 2.2446,
+      "step": 1803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009784583411197988,
+      "loss": 2.0921,
+      "step": 1804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009784267876031345,
+      "loss": 2.2179,
+      "step": 1805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009783952115036253,
+      "loss": 2.2255,
+      "step": 1806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009783636128227623,
+      "loss": 2.2505,
+      "step": 1807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009783319915620365,
+      "loss": 2.1805,
+      "step": 1808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009783003477229406,
+      "loss": 2.1285,
+      "step": 1809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009782686813069685,
+      "loss": 2.2003,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009782369923156146,
+      "loss": 2.1907,
+      "step": 1811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009782052807503752,
+      "loss": 2.0948,
+      "step": 1812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009781735466127465,
+      "loss": 2.2492,
+      "step": 1813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000978141789904227,
+      "loss": 2.3052,
+      "step": 1814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009781100106263153,
+      "loss": 2.2287,
+      "step": 1815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009780782087805116,
+      "loss": 2.0809,
+      "step": 1816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009780463843683171,
+      "loss": 2.3765,
+      "step": 1817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009780145373912339,
+      "loss": 2.0866,
+      "step": 1818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009779826678507652,
+      "loss": 2.2385,
+      "step": 1819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009779507757484154,
+      "loss": 2.1728,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009779188610856899,
+      "loss": 2.2649,
+      "step": 1821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009778869238640951,
+      "loss": 2.2089,
+      "step": 1822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009778549640851383,
+      "loss": 2.2894,
+      "step": 1823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009778229817503285,
+      "loss": 2.1769,
+      "step": 1824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009777909768611751,
+      "loss": 2.3213,
+      "step": 1825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977758949419189,
+      "loss": 2.1111,
+      "step": 1826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009777268994258814,
+      "loss": 2.1737,
+      "step": 1827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009776948268827657,
+      "loss": 2.1718,
+      "step": 1828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009776627317913558,
+      "loss": 2.2515,
+      "step": 1829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009776306141531664,
+      "loss": 2.3976,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009775984739697134,
+      "loss": 2.3384,
+      "step": 1831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009775663112425144,
+      "loss": 2.1257,
+      "step": 1832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009775341259730872,
+      "loss": 2.1297,
+      "step": 1833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977501918162951,
+      "loss": 2.3489,
+      "step": 1834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009774696878136263,
+      "loss": 2.0982,
+      "step": 1835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977437434926634,
+      "loss": 2.2293,
+      "step": 1836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977405159503497,
+      "loss": 2.1762,
+      "step": 1837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009773728615457386,
+      "loss": 2.1358,
+      "step": 1838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009773405410548836,
+      "loss": 2.3367,
+      "step": 1839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977308198032457,
+      "loss": 2.3142,
+      "step": 1840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009772758324799859,
+      "loss": 2.2948,
+      "step": 1841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009772434443989979,
+      "loss": 2.1342,
+      "step": 1842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009772110337910218,
+      "loss": 2.2189,
+      "step": 1843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009771786006575877,
+      "loss": 2.2069,
+      "step": 1844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977146145000226,
+      "loss": 2.23,
+      "step": 1845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977113666820469,
+      "loss": 2.0568,
+      "step": 1846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009770811661198497,
+      "loss": 2.2499,
+      "step": 1847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009770486428999025,
+      "loss": 2.1838,
+      "step": 1848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000977016097162162,
+      "loss": 2.1182,
+      "step": 1849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000976983528908165,
+      "loss": 2.232,
+      "step": 1850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009769509381394484,
+      "loss": 2.2679,
+      "step": 1851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009769183248575506,
+      "loss": 2.295,
+      "step": 1852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009768856890640112,
+      "loss": 2.1769,
+      "step": 1853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009768530307603708,
+      "loss": 2.2725,
+      "step": 1854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009768203499481705,
+      "loss": 2.3067,
+      "step": 1855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000976787646628953,
+      "loss": 2.3853,
+      "step": 1856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009767549208042623,
+      "loss": 2.275,
+      "step": 1857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009767221724756429,
+      "loss": 2.2488,
+      "step": 1858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000976689401644641,
+      "loss": 2.2872,
+      "step": 1859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009766566083128027,
+      "loss": 2.1346,
+      "step": 1860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009766237924816764,
+      "loss": 2.3792,
+      "step": 1861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009765909541528113,
+      "loss": 2.2931,
+      "step": 1862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0009765580933277569,
+      "loss": 2.2701,
+      "step": 1863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009765252100080644,
+      "loss": 2.1725,
+      "step": 1864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009764923041952865,
+      "loss": 1.9549,
+      "step": 1865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009764593758909758,
+      "loss": 2.2278,
+      "step": 1866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000976426425096687,
+      "loss": 2.2586,
+      "step": 1867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009763934518139753,
+      "loss": 2.1161,
+      "step": 1868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009763604560443971,
+      "loss": 2.1704,
+      "step": 1869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00097632743778951,
+      "loss": 2.1111,
+      "step": 1870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009762943970508723,
+      "loss": 2.133,
+      "step": 1871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009762613338300438,
+      "loss": 2.3106,
+      "step": 1872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009762282481285852,
+      "loss": 2.2187,
+      "step": 1873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009761951399480578,
+      "loss": 2.0096,
+      "step": 1874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000976162009290025,
+      "loss": 2.2113,
+      "step": 1875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009761288561560502,
+      "loss": 2.2174,
+      "step": 1876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009760956805476986,
+      "loss": 2.355,
+      "step": 1877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009760624824665359,
+      "loss": 2.0829,
+      "step": 1878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009760292619141291,
+      "loss": 2.1023,
+      "step": 1879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009759960188920467,
+      "loss": 2.2708,
+      "step": 1880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009759627534018573,
+      "loss": 2.3044,
+      "step": 1881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009759294654451316,
+      "loss": 2.1656,
+      "step": 1882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009758961550234407,
+      "loss": 2.3059,
+      "step": 1883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009758628221383567,
+      "loss": 2.1184,
+      "step": 1884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009758294667914531,
+      "loss": 2.2409,
+      "step": 1885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009757960889843048,
+      "loss": 2.1409,
+      "step": 1886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009757626887184865,
+      "loss": 2.201,
+      "step": 1887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009757292659955754,
+      "loss": 2.1245,
+      "step": 1888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009756958208171488,
+      "loss": 2.1743,
+      "step": 1889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009756623531847856,
+      "loss": 2.301,
+      "step": 1890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009756288631000655,
+      "loss": 2.4315,
+      "step": 1891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009755953505645691,
+      "loss": 2.1965,
+      "step": 1892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009755618155798787,
+      "loss": 2.2047,
+      "step": 1893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009755282581475768,
+      "loss": 2.148,
+      "step": 1894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009754946782692476,
+      "loss": 2.2747,
+      "step": 1895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009754610759464762,
+      "loss": 2.1043,
+      "step": 1896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009754274511808484,
+      "loss": 2.0387,
+      "step": 1897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009753938039739516,
+      "loss": 2.122,
+      "step": 1898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009753601343273742,
+      "loss": 2.2446,
+      "step": 1899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000975326442242705,
+      "loss": 2.2624,
+      "step": 1900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009752927277215348,
+      "loss": 2.2772,
+      "step": 1901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000975258990765455,
+      "loss": 2.1969,
+      "step": 1902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009752252313760577,
+      "loss": 2.0444,
+      "step": 1903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009751914495549366,
+      "loss": 2.1949,
+      "step": 1904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009751576453036863,
+      "loss": 2.2817,
+      "step": 1905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009751238186239025,
+      "loss": 2.0739,
+      "step": 1906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009750899695171819,
+      "loss": 2.2348,
+      "step": 1907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000975056097985122,
+      "loss": 2.2353,
+      "step": 1908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000975022204029322,
+      "loss": 2.1917,
+      "step": 1909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009749882876513813,
+      "loss": 2.3355,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009749543488529013,
+      "loss": 2.2006,
+      "step": 1911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009749203876354839,
+      "loss": 2.281,
+      "step": 1912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000974886404000732,
+      "loss": 2.0167,
+      "step": 1913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009748523979502495,
+      "loss": 2.3432,
+      "step": 1914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009748183694856421,
+      "loss": 2.2132,
+      "step": 1915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009747843186085155,
+      "loss": 2.1318,
+      "step": 1916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009747502453204773,
+      "loss": 2.1762,
+      "step": 1917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009747161496231359,
+      "loss": 2.2113,
+      "step": 1918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009746820315181005,
+      "loss": 2.149,
+      "step": 1919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009746478910069813,
+      "loss": 2.2289,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009746137280913904,
+      "loss": 2.1075,
+      "step": 1921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00097457954277294,
+      "loss": 2.2259,
+      "step": 1922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009745453350532438,
+      "loss": 2.315,
+      "step": 1923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009745111049339164,
+      "loss": 2.0853,
+      "step": 1924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009744768524165738,
+      "loss": 2.2864,
+      "step": 1925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009744425775028325,
+      "loss": 2.2558,
+      "step": 1926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009744082801943106,
+      "loss": 2.2362,
+      "step": 1927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009743739604926267,
+      "loss": 2.2398,
+      "step": 1928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009743396183994011,
+      "loss": 2.3289,
+      "step": 1929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009743052539162547,
+      "loss": 2.2315,
+      "step": 1930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009742708670448095,
+      "loss": 2.1395,
+      "step": 1931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009742364577866888,
+      "loss": 2.1833,
+      "step": 1932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009742020261435166,
+      "loss": 2.1788,
+      "step": 1933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009741675721169184,
+      "loss": 2.1164,
+      "step": 1934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009741330957085202,
+      "loss": 2.2383,
+      "step": 1935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009740985969199496,
+      "loss": 2.092,
+      "step": 1936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000974064075752835,
+      "loss": 2.1673,
+      "step": 1937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009740295322088057,
+      "loss": 2.3667,
+      "step": 1938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009739949662894925,
+      "loss": 2.1439,
+      "step": 1939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009739603779965267,
+      "loss": 2.0504,
+      "step": 1940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009739257673315412,
+      "loss": 2.2068,
+      "step": 1941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009738911342961697,
+      "loss": 2.1,
+      "step": 1942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009738564788920467,
+      "loss": 2.1247,
+      "step": 1943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009738218011208083,
+      "loss": 2.1469,
+      "step": 1944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009737871009840913,
+      "loss": 2.3928,
+      "step": 1945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009737523784835334,
+      "loss": 2.1022,
+      "step": 1946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009737176336207738,
+      "loss": 2.2129,
+      "step": 1947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009736828663974526,
+      "loss": 2.1299,
+      "step": 1948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009736480768152107,
+      "loss": 2.3683,
+      "step": 1949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009736132648756904,
+      "loss": 2.0489,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009735784305805349,
+      "loss": 2.198,
+      "step": 1951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009735435739313883,
+      "loss": 2.0631,
+      "step": 1952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009735086949298962,
+      "loss": 1.9923,
+      "step": 1953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009734737935777045,
+      "loss": 2.3587,
+      "step": 1954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000973438869876461,
+      "loss": 2.3211,
+      "step": 1955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000973403923827814,
+      "loss": 2.2386,
+      "step": 1956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009733689554334133,
+      "loss": 2.112,
+      "step": 1957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009733339646949093,
+      "loss": 2.1348,
+      "step": 1958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009732989516139537,
+      "loss": 2.3749,
+      "step": 1959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000973263916192199,
+      "loss": 2.2775,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009732288584312992,
+      "loss": 2.2337,
+      "step": 1961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009731937783329089,
+      "loss": 2.1567,
+      "step": 1962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009731586758986841,
+      "loss": 2.2676,
+      "step": 1963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009731235511302818,
+      "loss": 2.1582,
+      "step": 1964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009730884040293598,
+      "loss": 2.1868,
+      "step": 1965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009730532345975771,
+      "loss": 2.1822,
+      "step": 1966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000973018042836594,
+      "loss": 2.1738,
+      "step": 1967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009729828287480713,
+      "loss": 2.2365,
+      "step": 1968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009729475923336716,
+      "loss": 2.0732,
+      "step": 1969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009729123335950579,
+      "loss": 2.3197,
+      "step": 1970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009728770525338944,
+      "loss": 2.2342,
+      "step": 1971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009728417491518465,
+      "loss": 2.1123,
+      "step": 1972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009728064234505807,
+      "loss": 2.28,
+      "step": 1973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009727710754317644,
+      "loss": 2.1683,
+      "step": 1974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009727357050970662,
+      "loss": 2.2084,
+      "step": 1975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009727003124481556,
+      "loss": 2.1234,
+      "step": 1976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009726648974867032,
+      "loss": 2.2569,
+      "step": 1977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009726294602143806,
+      "loss": 2.1148,
+      "step": 1978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009725940006328606,
+      "loss": 2.2374,
+      "step": 1979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000972558518743817,
+      "loss": 2.2877,
+      "step": 1980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009725230145489246,
+      "loss": 2.2083,
+      "step": 1981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009724874880498591,
+      "loss": 2.2546,
+      "step": 1982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009724519392482978,
+      "loss": 2.1991,
+      "step": 1983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009724163681459183,
+      "loss": 2.2452,
+      "step": 1984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009723807747443999,
+      "loss": 2.2619,
+      "step": 1985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009723451590454227,
+      "loss": 2.0068,
+      "step": 1986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009723095210506677,
+      "loss": 2.2078,
+      "step": 1987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000972273860761817,
+      "loss": 2.5492,
+      "step": 1988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009722381781805541,
+      "loss": 2.1911,
+      "step": 1989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009722024733085633,
+      "loss": 2.2556,
+      "step": 1990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009721667461475297,
+      "loss": 2.2353,
+      "step": 1991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00097213099669914,
+      "loss": 2.0831,
+      "step": 1992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009720952249650814,
+      "loss": 2.1246,
+      "step": 1993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009720594309470425,
+      "loss": 2.2725,
+      "step": 1994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000972023614646713,
+      "loss": 2.2892,
+      "step": 1995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009719877760657833,
+      "loss": 2.1082,
+      "step": 1996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009719519152059451,
+      "loss": 2.3227,
+      "step": 1997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009719160320688912,
+      "loss": 2.1455,
+      "step": 1998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009718801266563154,
+      "loss": 2.2122,
+      "step": 1999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009718441989699125,
+      "loss": 2.2479,
+      "step": 2000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009718082490113783,
+      "loss": 2.3095,
+      "step": 2001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009717722767824096,
+      "loss": 2.2695,
+      "step": 2002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009717362822847045,
+      "loss": 2.2184,
+      "step": 2003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009717002655199622,
+      "loss": 2.343,
+      "step": 2004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009716642264898825,
+      "loss": 2.2169,
+      "step": 2005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009716281651961667,
+      "loss": 2.2234,
+      "step": 2006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009715920816405168,
+      "loss": 2.1388,
+      "step": 2007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009715559758246363,
+      "loss": 2.2879,
+      "step": 2008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009715198477502292,
+      "loss": 2.2197,
+      "step": 2009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009714836974190011,
+      "loss": 2.1818,
+      "step": 2010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000971447524832658,
+      "loss": 2.1804,
+      "step": 2011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0009714113299929077,
+      "loss": 2.2634,
+      "step": 2012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009713751129014585,
+      "loss": 2.0231,
+      "step": 2013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00097133887356002,
+      "loss": 2.0624,
+      "step": 2014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009713026119703027,
+      "loss": 2.1805,
+      "step": 2015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009712663281340183,
+      "loss": 2.1688,
+      "step": 2016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009712300220528794,
+      "loss": 2.2487,
+      "step": 2017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009711936937285998,
+      "loss": 2.0753,
+      "step": 2018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009711573431628943,
+      "loss": 2.203,
+      "step": 2019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009711209703574786,
+      "loss": 2.0682,
+      "step": 2020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009710845753140698,
+      "loss": 2.1736,
+      "step": 2021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009710481580343858,
+      "loss": 2.188,
+      "step": 2022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009710117185201452,
+      "loss": 2.0733,
+      "step": 2023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009709752567730685,
+      "loss": 2.3356,
+      "step": 2024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009709387727948766,
+      "loss": 2.1083,
+      "step": 2025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009709022665872916,
+      "loss": 2.138,
+      "step": 2026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009708657381520365,
+      "loss": 2.0808,
+      "step": 2027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009708291874908359,
+      "loss": 2.3269,
+      "step": 2028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009707926146054149,
+      "loss": 2.2357,
+      "step": 2029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009707560194974998,
+      "loss": 2.2366,
+      "step": 2030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000970719402168818,
+      "loss": 2.3763,
+      "step": 2031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000970682762621098,
+      "loss": 2.4651,
+      "step": 2032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000970646100856069,
+      "loss": 2.1206,
+      "step": 2033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009706094168754618,
+      "loss": 2.2678,
+      "step": 2034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009705727106810078,
+      "loss": 2.2388,
+      "step": 2035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009705359822744399,
+      "loss": 2.2405,
+      "step": 2036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009704992316574915,
+      "loss": 2.207,
+      "step": 2037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009704624588318972,
+      "loss": 2.1617,
+      "step": 2038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000970425663799393,
+      "loss": 2.2062,
+      "step": 2039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009703888465617157,
+      "loss": 2.2888,
+      "step": 2040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009703520071206031,
+      "loss": 2.2133,
+      "step": 2041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009703151454777941,
+      "loss": 2.0534,
+      "step": 2042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009702782616350286,
+      "loss": 2.1478,
+      "step": 2043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009702413555940476,
+      "loss": 2.2455,
+      "step": 2044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009702044273565935,
+      "loss": 2.1921,
+      "step": 2045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009701674769244089,
+      "loss": 2.2178,
+      "step": 2046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009701305042992384,
+      "loss": 2.1629,
+      "step": 2047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009700935094828266,
+      "loss": 2.2098,
+      "step": 2048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009700564924769202,
+      "loss": 2.279,
+      "step": 2049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009700194532832664,
+      "loss": 2.191,
+      "step": 2050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009699823919036136,
+      "loss": 2.2185,
+      "step": 2051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000969945308339711,
+      "loss": 2.3442,
+      "step": 2052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009699082025933092,
+      "loss": 2.2042,
+      "step": 2053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009698710746661596,
+      "loss": 2.3353,
+      "step": 2054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009698339245600147,
+      "loss": 2.1457,
+      "step": 2055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000969796752276628,
+      "loss": 2.321,
+      "step": 2056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009697595578177544,
+      "loss": 2.1319,
+      "step": 2057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009697223411851493,
+      "loss": 2.2292,
+      "step": 2058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009696851023805694,
+      "loss": 2.0653,
+      "step": 2059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009696478414057726,
+      "loss": 2.1847,
+      "step": 2060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009696105582625177,
+      "loss": 2.223,
+      "step": 2061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009695732529525644,
+      "loss": 2.2418,
+      "step": 2062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009695359254776738,
+      "loss": 2.3151,
+      "step": 2063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009694985758396076,
+      "loss": 2.0094,
+      "step": 2064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009694612040401292,
+      "loss": 2.1983,
+      "step": 2065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000969423810081002,
+      "loss": 2.2631,
+      "step": 2066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009693863939639918,
+      "loss": 2.2643,
+      "step": 2067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009693489556908641,
+      "loss": 2.2384,
+      "step": 2068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009693114952633863,
+      "loss": 2.0836,
+      "step": 2069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009692740126833269,
+      "loss": 2.0158,
+      "step": 2070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009692365079524549,
+      "loss": 2.3604,
+      "step": 2071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009691989810725406,
+      "loss": 2.1515,
+      "step": 2072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009691614320453552,
+      "loss": 1.9795,
+      "step": 2073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009691238608726716,
+      "loss": 2.1638,
+      "step": 2074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009690862675562626,
+      "loss": 2.266,
+      "step": 2075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009690486520979033,
+      "loss": 2.2092,
+      "step": 2076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009690110144993689,
+      "loss": 2.12,
+      "step": 2077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009689733547624359,
+      "loss": 2.1482,
+      "step": 2078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009689356728888823,
+      "loss": 1.9833,
+      "step": 2079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009688979688804863,
+      "loss": 2.0324,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009688602427390281,
+      "loss": 2.1093,
+      "step": 2081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000968822494466288,
+      "loss": 2.1568,
+      "step": 2082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009687847240640482,
+      "loss": 2.327,
+      "step": 2083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009687469315340912,
+      "loss": 2.241,
+      "step": 2084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009687091168782011,
+      "loss": 2.1582,
+      "step": 2085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009686712800981629,
+      "loss": 2.1422,
+      "step": 2086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009686334211957625,
+      "loss": 2.1139,
+      "step": 2087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009685955401727869,
+      "loss": 2.1478,
+      "step": 2088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009685576370310241,
+      "loss": 2.2605,
+      "step": 2089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009685197117722632,
+      "loss": 2.1617,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009684817643982949,
+      "loss": 2.2567,
+      "step": 2091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009684437949109095,
+      "loss": 2.1287,
+      "step": 2092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009684058033118998,
+      "loss": 2.2269,
+      "step": 2093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009683677896030592,
+      "loss": 2.2377,
+      "step": 2094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009683297537861817,
+      "loss": 2.3719,
+      "step": 2095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000968291695863063,
+      "loss": 2.2688,
+      "step": 2096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009682536158354991,
+      "loss": 1.9915,
+      "step": 2097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009682155137052879,
+      "loss": 2.2221,
+      "step": 2098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009681773894742275,
+      "loss": 2.1646,
+      "step": 2099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009681392431441178,
+      "loss": 2.2179,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009681010747167591,
+      "loss": 2.3892,
+      "step": 2101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009680628841939534,
+      "loss": 2.2703,
+      "step": 2102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009680246715775032,
+      "loss": 2.1138,
+      "step": 2103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009679864368692119,
+      "loss": 2.1323,
+      "step": 2104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009679481800708849,
+      "loss": 2.1099,
+      "step": 2105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009679099011843276,
+      "loss": 2.2259,
+      "step": 2106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009678716002113468,
+      "loss": 2.1801,
+      "step": 2107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009678332771537506,
+      "loss": 2.3471,
+      "step": 2108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009677949320133477,
+      "loss": 2.3351,
+      "step": 2109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009677565647919484,
+      "loss": 2.0826,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009677181754913635,
+      "loss": 2.2957,
+      "step": 2111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009676797641134052,
+      "loss": 2.2052,
+      "step": 2112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009676413306598864,
+      "loss": 2.2123,
+      "step": 2113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009676028751326213,
+      "loss": 2.1309,
+      "step": 2114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009675643975334254,
+      "loss": 2.1326,
+      "step": 2115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009675258978641146,
+      "loss": 2.1042,
+      "step": 2116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009674873761265062,
+      "loss": 2.1525,
+      "step": 2117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009674488323224187,
+      "loss": 2.1948,
+      "step": 2118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009674102664536711,
+      "loss": 2.1228,
+      "step": 2119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009673716785220843,
+      "loss": 2.258,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009673330685294793,
+      "loss": 2.1604,
+      "step": 2121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009672944364776788,
+      "loss": 2.0699,
+      "step": 2122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009672557823685063,
+      "loss": 2.2294,
+      "step": 2123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009672171062037864,
+      "loss": 2.2262,
+      "step": 2124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009671784079853445,
+      "loss": 2.2748,
+      "step": 2125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009671396877150073,
+      "loss": 2.2125,
+      "step": 2126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009671009453946029,
+      "loss": 2.1837,
+      "step": 2127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009670621810259594,
+      "loss": 2.3537,
+      "step": 2128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000967023394610907,
+      "loss": 2.3369,
+      "step": 2129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009669845861512763,
+      "loss": 2.2614,
+      "step": 2130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009669457556488994,
+      "loss": 2.1089,
+      "step": 2131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009669069031056088,
+      "loss": 2.1163,
+      "step": 2132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009668680285232387,
+      "loss": 2.2574,
+      "step": 2133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009668291319036242,
+      "loss": 2.3849,
+      "step": 2134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009667902132486009,
+      "loss": 2.1873,
+      "step": 2135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009667512725600061,
+      "loss": 2.1703,
+      "step": 2136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000966712309839678,
+      "loss": 2.1114,
+      "step": 2137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009666733250894554,
+      "loss": 2.3233,
+      "step": 2138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009666343183111788,
+      "loss": 2.2385,
+      "step": 2139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009665952895066891,
+      "loss": 2.2201,
+      "step": 2140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009665562386778288,
+      "loss": 2.1877,
+      "step": 2141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009665171658264412,
+      "loss": 2.0945,
+      "step": 2142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009664780709543704,
+      "loss": 2.1936,
+      "step": 2143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000966438954063462,
+      "loss": 2.3196,
+      "step": 2144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009663998151555625,
+      "loss": 2.308,
+      "step": 2145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009663606542325187,
+      "loss": 2.1425,
+      "step": 2146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009663214712961798,
+      "loss": 2.156,
+      "step": 2147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009662822663483951,
+      "loss": 2.1604,
+      "step": 2148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009662430393910151,
+      "loss": 2.2614,
+      "step": 2149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009662037904258913,
+      "loss": 2.3041,
+      "step": 2150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009661645194548767,
+      "loss": 2.0098,
+      "step": 2151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009661252264798246,
+      "loss": 2.3078,
+      "step": 2152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00096608591150259,
+      "loss": 2.1959,
+      "step": 2153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009660465745250284,
+      "loss": 2.194,
+      "step": 2154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009660072155489968,
+      "loss": 2.1624,
+      "step": 2155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000965967834576353,
+      "loss": 2.0578,
+      "step": 2156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009659284316089558,
+      "loss": 2.2002,
+      "step": 2157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009658890066486651,
+      "loss": 2.2747,
+      "step": 2158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000965849559697342,
+      "loss": 2.1576,
+      "step": 2159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0009658100907568482,
+      "loss": 2.053,
+      "step": 2160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000965770599829047,
+      "loss": 2.266,
+      "step": 2161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009657310869158025,
+      "loss": 2.2372,
+      "step": 2162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009656915520189795,
+      "loss": 2.147,
+      "step": 2163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009656519951404443,
+      "loss": 2.0337,
+      "step": 2164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009656124162820641,
+      "loss": 2.2236,
+      "step": 2165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009655728154457072,
+      "loss": 2.1081,
+      "step": 2166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009655331926332427,
+      "loss": 2.2702,
+      "step": 2167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009654935478465409,
+      "loss": 2.2623,
+      "step": 2168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009654538810874731,
+      "loss": 2.0911,
+      "step": 2169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009654141923579119,
+      "loss": 2.1751,
+      "step": 2170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009653744816597303,
+      "loss": 2.1843,
+      "step": 2171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000965334748994803,
+      "loss": 2.4233,
+      "step": 2172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009652949943650056,
+      "loss": 2.0904,
+      "step": 2173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009652552177722143,
+      "loss": 2.2248,
+      "step": 2174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009652154192183069,
+      "loss": 2.2191,
+      "step": 2175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009651755987051618,
+      "loss": 2.2974,
+      "step": 2176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009651357562346585,
+      "loss": 2.2395,
+      "step": 2177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000965095891808678,
+      "loss": 2.2023,
+      "step": 2178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000965056005429102,
+      "loss": 2.1329,
+      "step": 2179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009650160970978129,
+      "loss": 2.2712,
+      "step": 2180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009649761668166948,
+      "loss": 2.1013,
+      "step": 2181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009649362145876322,
+      "loss": 2.1407,
+      "step": 2182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000964896240412511,
+      "loss": 2.3025,
+      "step": 2183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009648562442932182,
+      "loss": 2.2842,
+      "step": 2184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009648162262316416,
+      "loss": 2.1747,
+      "step": 2185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009647761862296702,
+      "loss": 2.0874,
+      "step": 2186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009647361242891941,
+      "loss": 2.0288,
+      "step": 2187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009646960404121041,
+      "loss": 2.113,
+      "step": 2188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009646559346002924,
+      "loss": 2.145,
+      "step": 2189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009646158068556521,
+      "loss": 2.055,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000964575657180077,
+      "loss": 2.1779,
+      "step": 2191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009645354855754626,
+      "loss": 1.937,
+      "step": 2192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009644952920437051,
+      "loss": 2.1986,
+      "step": 2193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009644550765867017,
+      "loss": 2.281,
+      "step": 2194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009644148392063505,
+      "loss": 2.255,
+      "step": 2195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009643745799045509,
+      "loss": 2.2675,
+      "step": 2196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009643342986832031,
+      "loss": 2.2081,
+      "step": 2197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009642939955442087,
+      "loss": 2.1077,
+      "step": 2198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00096425367048947,
+      "loss": 2.1926,
+      "step": 2199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009642133235208903,
+      "loss": 2.1144,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009641729546403744,
+      "loss": 2.1933,
+      "step": 2201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009641325638498274,
+      "loss": 2.2524,
+      "step": 2202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009640921511511561,
+      "loss": 2.2218,
+      "step": 2203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000964051716546268,
+      "loss": 2.1859,
+      "step": 2204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009640112600370717,
+      "loss": 2.2569,
+      "step": 2205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009639707816254769,
+      "loss": 2.1563,
+      "step": 2206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009639302813133941,
+      "loss": 2.2551,
+      "step": 2207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009638897591027353,
+      "loss": 2.2698,
+      "step": 2208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000963849214995413,
+      "loss": 2.107,
+      "step": 2209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009638086489933411,
+      "loss": 2.1881,
+      "step": 2210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009637680610984343,
+      "loss": 2.2664,
+      "step": 2211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009637274513126085,
+      "loss": 2.294,
+      "step": 2212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009636868196377806,
+      "loss": 2.0339,
+      "step": 2213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009636461660758684,
+      "loss": 2.1889,
+      "step": 2214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009636054906287911,
+      "loss": 2.1227,
+      "step": 2215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009635647932984683,
+      "loss": 2.1397,
+      "step": 2216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009635240740868214,
+      "loss": 2.1679,
+      "step": 2217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009634833329957721,
+      "loss": 2.1478,
+      "step": 2218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009634425700272437,
+      "loss": 2.0652,
+      "step": 2219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009634017851831603,
+      "loss": 2.2533,
+      "step": 2220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009633609784654469,
+      "loss": 2.342,
+      "step": 2221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009633201498760298,
+      "loss": 2.1809,
+      "step": 2222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000963279299416836,
+      "loss": 2.295,
+      "step": 2223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009632384270897939,
+      "loss": 2.2295,
+      "step": 2224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009631975328968327,
+      "loss": 2.0445,
+      "step": 2225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009631566168398828,
+      "loss": 2.2229,
+      "step": 2226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009631156789208757,
+      "loss": 2.2601,
+      "step": 2227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009630747191417433,
+      "loss": 2.1226,
+      "step": 2228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009630337375044193,
+      "loss": 2.2379,
+      "step": 2229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962992734010838,
+      "loss": 2.26,
+      "step": 2230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962951708662935,
+      "loss": 2.0821,
+      "step": 2231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009629106614626466,
+      "loss": 2.1542,
+      "step": 2232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009628695924119106,
+      "loss": 2.0774,
+      "step": 2233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009628285015126653,
+      "loss": 2.105,
+      "step": 2234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009627873887668503,
+      "loss": 2.3092,
+      "step": 2235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009627462541764065,
+      "loss": 2.3554,
+      "step": 2236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009627050977432752,
+      "loss": 2.122,
+      "step": 2237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009626639194693992,
+      "loss": 2.1242,
+      "step": 2238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009626227193567223,
+      "loss": 2.069,
+      "step": 2239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962581497407189,
+      "loss": 2.1132,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009625402536227454,
+      "loss": 2.0723,
+      "step": 2241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009624989880053381,
+      "loss": 2.3335,
+      "step": 2242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962457700556915,
+      "loss": 2.0845,
+      "step": 2243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009624163912794248,
+      "loss": 2.3092,
+      "step": 2244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009623750601748177,
+      "loss": 2.2015,
+      "step": 2245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009623337072450443,
+      "loss": 2.2152,
+      "step": 2246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009622923324920567,
+      "loss": 2.1298,
+      "step": 2247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962250935917808,
+      "loss": 2.1074,
+      "step": 2248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009622095175242521,
+      "loss": 2.1152,
+      "step": 2249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962168077313344,
+      "loss": 2.31,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009621266152870398,
+      "loss": 2.0721,
+      "step": 2251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009620851314472967,
+      "loss": 2.1671,
+      "step": 2252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009620436257960726,
+      "loss": 2.2875,
+      "step": 2253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000962002098335327,
+      "loss": 2.2148,
+      "step": 2254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009619605490670198,
+      "loss": 2.075,
+      "step": 2255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009619189779931123,
+      "loss": 2.2149,
+      "step": 2256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009618773851155668,
+      "loss": 2.1625,
+      "step": 2257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009618357704363465,
+      "loss": 2.4436,
+      "step": 2258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009617941339574159,
+      "loss": 2.1348,
+      "step": 2259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009617524756807401,
+      "loss": 2.2146,
+      "step": 2260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009617107956082856,
+      "loss": 2.1741,
+      "step": 2261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009616690937420197,
+      "loss": 2.2607,
+      "step": 2262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009616273700839108,
+      "loss": 2.1408,
+      "step": 2263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009615856246359286,
+      "loss": 2.1753,
+      "step": 2264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009615438574000435,
+      "loss": 2.3325,
+      "step": 2265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009615020683782268,
+      "loss": 2.1395,
+      "step": 2266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009614602575724512,
+      "loss": 2.247,
+      "step": 2267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009614184249846903,
+      "loss": 2.2008,
+      "step": 2268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009613765706169186,
+      "loss": 2.26,
+      "step": 2269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009613346944711117,
+      "loss": 2.2669,
+      "step": 2270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009612927965492463,
+      "loss": 2.2299,
+      "step": 2271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009612508768533001,
+      "loss": 1.9762,
+      "step": 2272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009612089353852517,
+      "loss": 2.3081,
+      "step": 2273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000961166972147081,
+      "loss": 2.338,
+      "step": 2274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009611249871407688,
+      "loss": 2.1923,
+      "step": 2275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009610829803682967,
+      "loss": 2.1762,
+      "step": 2276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009610409518316475,
+      "loss": 2.089,
+      "step": 2277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009609989015328051,
+      "loss": 2.3945,
+      "step": 2278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009609568294737546,
+      "loss": 2.3457,
+      "step": 2279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009609147356564816,
+      "loss": 2.2492,
+      "step": 2280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000960872620082973,
+      "loss": 2.2251,
+      "step": 2281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000960830482755217,
+      "loss": 2.253,
+      "step": 2282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009607883236752024,
+      "loss": 2.1062,
+      "step": 2283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009607461428449193,
+      "loss": 2.0854,
+      "step": 2284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009607039402663586,
+      "loss": 2.1244,
+      "step": 2285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009606617159415126,
+      "loss": 2.2304,
+      "step": 2286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009606194698723742,
+      "loss": 2.1913,
+      "step": 2287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009605772020609375,
+      "loss": 2.148,
+      "step": 2288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009605349125091977,
+      "loss": 2.1542,
+      "step": 2289
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000960492601219151,
+      "loss": 2.292,
+      "step": 2290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009604502681927944,
+      "loss": 2.3295,
+      "step": 2291
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009604079134321264,
+      "loss": 2.2105,
+      "step": 2292
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009603655369391461,
+      "loss": 2.2257,
+      "step": 2293
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009603231387158537,
+      "loss": 2.2592,
+      "step": 2294
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009602807187642506,
+      "loss": 2.1963,
+      "step": 2295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009602382770863391,
+      "loss": 2.0009,
+      "step": 2296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009601958136841224,
+      "loss": 2.298,
+      "step": 2297
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009601533285596051,
+      "loss": 2.2212,
+      "step": 2298
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009601108217147925,
+      "loss": 2.1491,
+      "step": 2299
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009600682931516909,
+      "loss": 2.3144,
+      "step": 2300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000960025742872308,
+      "loss": 2.2568,
+      "step": 2301
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000959983170878652,
+      "loss": 2.1625,
+      "step": 2302
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009599405771727326,
+      "loss": 2.247,
+      "step": 2303
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009598979617565602,
+      "loss": 2.2309,
+      "step": 2304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009598553246321464,
+      "loss": 2.172,
+      "step": 2305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009598126658015039,
+      "loss": 2.1155,
+      "step": 2306
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009597699852666459,
+      "loss": 2.3483,
+      "step": 2307
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009597272830295876,
+      "loss": 2.3676,
+      "step": 2308
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000959684559092344,
+      "loss": 2.1489,
+      "step": 2309
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0009596418134569322,
+      "loss": 2.1967,
+      "step": 2310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009595990461253699,
+      "loss": 2.1047,
+      "step": 2311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009595562570996754,
+      "loss": 2.1045,
+      "step": 2312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000959513446381869,
+      "loss": 2.2889,
+      "step": 2313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009594706139739711,
+      "loss": 2.285,
+      "step": 2314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009594277598780037,
+      "loss": 2.3584,
+      "step": 2315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009593848840959893,
+      "loss": 2.2092,
+      "step": 2316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009593419866299521,
+      "loss": 2.1754,
+      "step": 2317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009592990674819168,
+      "loss": 2.2427,
+      "step": 2318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009592561266539092,
+      "loss": 2.3591,
+      "step": 2319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009592131641479562,
+      "loss": 2.2224,
+      "step": 2320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000959170179966086,
+      "loss": 2.1677,
+      "step": 2321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009591271741103273,
+      "loss": 2.3199,
+      "step": 2322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009590841465827101,
+      "loss": 2.2597,
+      "step": 2323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009590410973852654,
+      "loss": 2.1817,
+      "step": 2324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009589980265200253,
+      "loss": 2.0949,
+      "step": 2325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009589549339890229,
+      "loss": 2.2404,
+      "step": 2326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009589118197942921,
+      "loss": 2.1507,
+      "step": 2327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000958868683937868,
+      "loss": 2.077,
+      "step": 2328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009588255264217869,
+      "loss": 2.086,
+      "step": 2329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009587823472480857,
+      "loss": 2.0803,
+      "step": 2330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009587391464188027,
+      "loss": 2.3083,
+      "step": 2331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000958695923935977,
+      "loss": 2.2576,
+      "step": 2332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009586526798016488,
+      "loss": 2.1617,
+      "step": 2333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009586094140178595,
+      "loss": 2.1632,
+      "step": 2334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009585661265866511,
+      "loss": 2.1944,
+      "step": 2335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009585228175100671,
+      "loss": 2.2137,
+      "step": 2336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009584794867901514,
+      "loss": 2.1454,
+      "step": 2337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009584361344289499,
+      "loss": 1.9807,
+      "step": 2338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009583927604285083,
+      "loss": 2.1706,
+      "step": 2339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009583493647908744,
+      "loss": 2.1112,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009583059475180964,
+      "loss": 2.2568,
+      "step": 2341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009582625086122238,
+      "loss": 2.2427,
+      "step": 2342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009582190480753069,
+      "loss": 2.179,
+      "step": 2343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009581755659093972,
+      "loss": 2.2369,
+      "step": 2344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000958132062116547,
+      "loss": 2.2096,
+      "step": 2345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009580885366988101,
+      "loss": 2.2662,
+      "step": 2346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009580449896582406,
+      "loss": 2.2159,
+      "step": 2347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009580014209968945,
+      "loss": 2.1881,
+      "step": 2348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009579578307168278,
+      "loss": 2.2396,
+      "step": 2349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009579142188200986,
+      "loss": 2.227,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000957870585308765,
+      "loss": 2.2571,
+      "step": 2351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000957826930184887,
+      "loss": 2.1626,
+      "step": 2352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009577832534505248,
+      "loss": 2.0419,
+      "step": 2353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009577395551077406,
+      "loss": 2.2919,
+      "step": 2354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009576958351585965,
+      "loss": 2.2576,
+      "step": 2355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009576520936051565,
+      "loss": 2.224,
+      "step": 2356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009576083304494854,
+      "loss": 2.2206,
+      "step": 2357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009575645456936485,
+      "loss": 2.1501,
+      "step": 2358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009575207393397129,
+      "loss": 2.2971,
+      "step": 2359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009574769113897462,
+      "loss": 2.0796,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009574330618458174,
+      "loss": 2.1124,
+      "step": 2361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000957389190709996,
+      "loss": 2.1726,
+      "step": 2362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009573452979843529,
+      "loss": 2.1215,
+      "step": 2363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00095730138367096,
+      "loss": 2.2775,
+      "step": 2364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009572574477718903,
+      "loss": 2.25,
+      "step": 2365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009572134902892175,
+      "loss": 2.3045,
+      "step": 2366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009571695112250164,
+      "loss": 2.2011,
+      "step": 2367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009571255105813631,
+      "loss": 2.0772,
+      "step": 2368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009570814883603345,
+      "loss": 2.273,
+      "step": 2369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009570374445640084,
+      "loss": 2.3304,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000956993379194464,
+      "loss": 2.3825,
+      "step": 2371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000956949292253781,
+      "loss": 2.1604,
+      "step": 2372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009569051837440407,
+      "loss": 2.2148,
+      "step": 2373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000956861053667325,
+      "loss": 2.3169,
+      "step": 2374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009568169020257169,
+      "loss": 2.2125,
+      "step": 2375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009567727288213005,
+      "loss": 2.126,
+      "step": 2376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009567285340561608,
+      "loss": 2.2868,
+      "step": 2377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009566843177323841,
+      "loss": 2.2428,
+      "step": 2378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009566400798520574,
+      "loss": 2.1009,
+      "step": 2379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009565958204172687,
+      "loss": 2.1795,
+      "step": 2380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009565515394301072,
+      "loss": 2.2299,
+      "step": 2381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009565072368926633,
+      "loss": 2.2689,
+      "step": 2382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009564629128070278,
+      "loss": 2.1931,
+      "step": 2383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009564185671752932,
+      "loss": 2.1454,
+      "step": 2384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009563741999995526,
+      "loss": 2.2971,
+      "step": 2385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009563298112819003,
+      "loss": 2.2221,
+      "step": 2386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009562854010244315,
+      "loss": 2.1923,
+      "step": 2387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009562409692292425,
+      "loss": 2.108,
+      "step": 2388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009561965158984304,
+      "loss": 2.2395,
+      "step": 2389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009561520410340936,
+      "loss": 2.2102,
+      "step": 2390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009561075446383316,
+      "loss": 2.355,
+      "step": 2391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009560630267132446,
+      "loss": 2.3184,
+      "step": 2392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000956018487260934,
+      "loss": 2.2833,
+      "step": 2393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000955973926283502,
+      "loss": 2.1483,
+      "step": 2394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000955929343783052,
+      "loss": 2.1245,
+      "step": 2395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009558847397616888,
+      "loss": 2.2988,
+      "step": 2396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009558401142215171,
+      "loss": 2.2722,
+      "step": 2397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000955795467164644,
+      "loss": 2.0649,
+      "step": 2398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009557507985931766,
+      "loss": 2.2042,
+      "step": 2399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009557061085092235,
+      "loss": 2.1168,
+      "step": 2400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000955661396914894,
+      "loss": 2.2347,
+      "step": 2401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009556166638122987,
+      "loss": 2.2651,
+      "step": 2402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000955571909203549,
+      "loss": 2.066,
+      "step": 2403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009555271330907578,
+      "loss": 2.2763,
+      "step": 2404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009554823354760382,
+      "loss": 2.1721,
+      "step": 2405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009554375163615049,
+      "loss": 2.306,
+      "step": 2406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009553926757492735,
+      "loss": 2.0293,
+      "step": 2407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009553478136414605,
+      "loss": 2.1701,
+      "step": 2408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009553029300401835,
+      "loss": 2.094,
+      "step": 2409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009552580249475613,
+      "loss": 2.2122,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009552130983657133,
+      "loss": 2.2849,
+      "step": 2411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009551681502967602,
+      "loss": 2.3328,
+      "step": 2412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009551231807428237,
+      "loss": 2.3589,
+      "step": 2413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009550781897060263,
+      "loss": 2.2101,
+      "step": 2414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009550331771884921,
+      "loss": 2.1546,
+      "step": 2415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009549881431923453,
+      "loss": 2.155,
+      "step": 2416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009549430877197118,
+      "loss": 2.2294,
+      "step": 2417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009548980107727185,
+      "loss": 2.2674,
+      "step": 2418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009548529123534928,
+      "loss": 2.0961,
+      "step": 2419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009548077924641637,
+      "loss": 1.9859,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009547626511068609,
+      "loss": 2.1934,
+      "step": 2421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009547174882837151,
+      "loss": 2.2173,
+      "step": 2422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009546723039968582,
+      "loss": 2.1775,
+      "step": 2423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000954627098248423,
+      "loss": 2.3404,
+      "step": 2424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009545818710405432,
+      "loss": 2.0268,
+      "step": 2425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009545366223753538,
+      "loss": 2.2976,
+      "step": 2426
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009544913522549904,
+      "loss": 2.2577,
+      "step": 2427
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00095444606068159,
+      "loss": 2.0998,
+      "step": 2428
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009544007476572905,
+      "loss": 2.1634,
+      "step": 2429
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009543554131842308,
+      "loss": 2.3296,
+      "step": 2430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009543100572645506,
+      "loss": 2.2386,
+      "step": 2431
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009542646799003911,
+      "loss": 2.2897,
+      "step": 2432
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009542192810938939,
+      "loss": 2.1825,
+      "step": 2433
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009541738608472021,
+      "loss": 2.1269,
+      "step": 2434
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009541284191624598,
+      "loss": 2.2057,
+      "step": 2435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009540829560418114,
+      "loss": 2.0516,
+      "step": 2436
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009540374714874036,
+      "loss": 2.1214,
+      "step": 2437
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009539919655013829,
+      "loss": 2.2829,
+      "step": 2438
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009539464380858973,
+      "loss": 2.2225,
+      "step": 2439
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000953900889243096,
+      "loss": 2.2047,
+      "step": 2440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009538553189751287,
+      "loss": 2.234,
+      "step": 2441
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009538097272841469,
+      "loss": 2.332,
+      "step": 2442
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009537641141723021,
+      "loss": 2.3692,
+      "step": 2443
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009537184796417479,
+      "loss": 2.0825,
+      "step": 2444
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009536728236946379,
+      "loss": 2.1444,
+      "step": 2445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009536271463331272,
+      "loss": 2.3368,
+      "step": 2446
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009535814475593723,
+      "loss": 2.1714,
+      "step": 2447
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009535357273755296,
+      "loss": 2.2291,
+      "step": 2448
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009534899857837579,
+      "loss": 2.1858,
+      "step": 2449
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009534442227862159,
+      "loss": 2.1696,
+      "step": 2450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009533984383850638,
+      "loss": 2.2069,
+      "step": 2451
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009533526325824626,
+      "loss": 2.125,
+      "step": 2452
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009533068053805747,
+      "loss": 2.3197,
+      "step": 2453
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009532609567815632,
+      "loss": 2.2439,
+      "step": 2454
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009532150867875922,
+      "loss": 2.2292,
+      "step": 2455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009531691954008268,
+      "loss": 2.1666,
+      "step": 2456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009531232826234332,
+      "loss": 2.2572,
+      "step": 2457
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009530773484575785,
+      "loss": 2.1827,
+      "step": 2458
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0009530313929054313,
+      "loss": 2.1437,
+      "step": 2459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009529854159691604,
+      "loss": 2.1928,
+      "step": 2460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009529394176509361,
+      "loss": 2.1778,
+      "step": 2461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009528933979529297,
+      "loss": 2.2576,
+      "step": 2462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009528473568773134,
+      "loss": 2.1505,
+      "step": 2463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009528012944262605,
+      "loss": 2.0947,
+      "step": 2464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009527552106019453,
+      "loss": 2.1299,
+      "step": 2465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009527091054065429,
+      "loss": 2.1405,
+      "step": 2466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009526629788422296,
+      "loss": 2.0644,
+      "step": 2467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009526168309111827,
+      "loss": 2.2866,
+      "step": 2468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009525706616155805,
+      "loss": 2.0454,
+      "step": 2469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009525244709576024,
+      "loss": 2.0195,
+      "step": 2470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009524782589394285,
+      "loss": 2.164,
+      "step": 2471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009524320255632402,
+      "loss": 2.0237,
+      "step": 2472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009523857708312201,
+      "loss": 2.2025,
+      "step": 2473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000952339494745551,
+      "loss": 2.0993,
+      "step": 2474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009522931973084176,
+      "loss": 2.1378,
+      "step": 2475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009522468785220053,
+      "loss": 2.3178,
+      "step": 2476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009522005383885002,
+      "loss": 2.3013,
+      "step": 2477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009521541769100896,
+      "loss": 2.1711,
+      "step": 2478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009521077940889624,
+      "loss": 2.1946,
+      "step": 2479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009520613899273074,
+      "loss": 2.2705,
+      "step": 2480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009520149644273153,
+      "loss": 2.2762,
+      "step": 2481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009519685175911771,
+      "loss": 2.0732,
+      "step": 2482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009519220494210858,
+      "loss": 2.1911,
+      "step": 2483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009518755599192346,
+      "loss": 2.1065,
+      "step": 2484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009518290490878175,
+      "loss": 2.1682,
+      "step": 2485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009517825169290304,
+      "loss": 2.152,
+      "step": 2486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009517359634450695,
+      "loss": 2.0236,
+      "step": 2487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009516893886381322,
+      "loss": 2.1425,
+      "step": 2488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009516427925104171,
+      "loss": 2.1027,
+      "step": 2489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009515961750641235,
+      "loss": 2.1765,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009515495363014521,
+      "loss": 2.0821,
+      "step": 2491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009515028762246041,
+      "loss": 2.1284,
+      "step": 2492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009514561948357818,
+      "loss": 2.1542,
+      "step": 2493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009514094921371891,
+      "loss": 2.3055,
+      "step": 2494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009513627681310304,
+      "loss": 2.1197,
+      "step": 2495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009513160228195108,
+      "loss": 2.1003,
+      "step": 2496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009512692562048372,
+      "loss": 2.2796,
+      "step": 2497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009512224682892169,
+      "loss": 2.2749,
+      "step": 2498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009511756590748583,
+      "loss": 2.1035,
+      "step": 2499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009511288285639712,
+      "loss": 2.2372,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000951081976758766,
+      "loss": 2.105,
+      "step": 2501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009510351036614539,
+      "loss": 2.195,
+      "step": 2502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009509882092742479,
+      "loss": 2.0302,
+      "step": 2503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009509412935993614,
+      "loss": 2.2242,
+      "step": 2504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009508943566390085,
+      "loss": 2.2484,
+      "step": 2505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009508473983954052,
+      "loss": 2.1262,
+      "step": 2506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000950800418870768,
+      "loss": 2.0884,
+      "step": 2507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009507534180673142,
+      "loss": 2.0694,
+      "step": 2508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009507063959872625,
+      "loss": 2.1318,
+      "step": 2509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009506593526328324,
+      "loss": 2.1971,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009506122880062447,
+      "loss": 2.1978,
+      "step": 2511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009505652021097205,
+      "loss": 2.1199,
+      "step": 2512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009505180949454827,
+      "loss": 2.2719,
+      "step": 2513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009504709665157547,
+      "loss": 2.275,
+      "step": 2514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009504238168227613,
+      "loss": 2.3151,
+      "step": 2515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009503766458687277,
+      "loss": 2.1754,
+      "step": 2516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009503294536558808,
+      "loss": 2.1683,
+      "step": 2517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009502822401864483,
+      "loss": 2.2248,
+      "step": 2518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009502350054626583,
+      "loss": 2.1199,
+      "step": 2519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009501877494867408,
+      "loss": 2.1279,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009501404722609261,
+      "loss": 2.2677,
+      "step": 2521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000950093173787446,
+      "loss": 2.2224,
+      "step": 2522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009500458540685331,
+      "loss": 2.1686,
+      "step": 2523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009499985131064207,
+      "loss": 2.1621,
+      "step": 2524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009499511509033438,
+      "loss": 2.0805,
+      "step": 2525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009499037674615379,
+      "loss": 2.1207,
+      "step": 2526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009498563627832394,
+      "loss": 2.1151,
+      "step": 2527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009498089368706862,
+      "loss": 2.1853,
+      "step": 2528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009497614897261167,
+      "loss": 2.1415,
+      "step": 2529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009497140213517706,
+      "loss": 2.124,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009496665317498883,
+      "loss": 2.4022,
+      "step": 2531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009496190209227118,
+      "loss": 2.3737,
+      "step": 2532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009495714888724836,
+      "loss": 2.3093,
+      "step": 2533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009495239356014471,
+      "loss": 2.1396,
+      "step": 2534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009494763611118471,
+      "loss": 2.0724,
+      "step": 2535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009494287654059293,
+      "loss": 2.2468,
+      "step": 2536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009493811484859402,
+      "loss": 2.1742,
+      "step": 2537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009493335103541274,
+      "loss": 2.1769,
+      "step": 2538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009492858510127397,
+      "loss": 2.0884,
+      "step": 2539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009492381704640265,
+      "loss": 2.1374,
+      "step": 2540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009491904687102386,
+      "loss": 2.1831,
+      "step": 2541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009491427457536277,
+      "loss": 2.4379,
+      "step": 2542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009490950015964461,
+      "loss": 2.3632,
+      "step": 2543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009490472362409479,
+      "loss": 2.0161,
+      "step": 2544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009489994496893874,
+      "loss": 2.2203,
+      "step": 2545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009489516419440203,
+      "loss": 2.0127,
+      "step": 2546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009489038130071031,
+      "loss": 2.233,
+      "step": 2547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009488559628808938,
+      "loss": 2.0633,
+      "step": 2548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009488080915676508,
+      "loss": 2.2782,
+      "step": 2549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009487601990696337,
+      "loss": 2.0402,
+      "step": 2550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009487122853891033,
+      "loss": 2.17,
+      "step": 2551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000948664350528321,
+      "loss": 2.0886,
+      "step": 2552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009486163944895496,
+      "loss": 2.0108,
+      "step": 2553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009485684172750528,
+      "loss": 2.2444,
+      "step": 2554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009485204188870952,
+      "loss": 2.0479,
+      "step": 2555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009484723993279422,
+      "loss": 2.2688,
+      "step": 2556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009484243585998608,
+      "loss": 2.1854,
+      "step": 2557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009483762967051183,
+      "loss": 2.0809,
+      "step": 2558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009483282136459834,
+      "loss": 2.2224,
+      "step": 2559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000948280109424726,
+      "loss": 2.2181,
+      "step": 2560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009482319840436165,
+      "loss": 2.3115,
+      "step": 2561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009481838375049265,
+      "loss": 2.2234,
+      "step": 2562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009481356698109287,
+      "loss": 2.2361,
+      "step": 2563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009480874809638968,
+      "loss": 2.0314,
+      "step": 2564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009480392709661052,
+      "loss": 2.188,
+      "step": 2565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009479910398198299,
+      "loss": 2.0122,
+      "step": 2566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000947942787527347,
+      "loss": 2.1462,
+      "step": 2567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009478945140909346,
+      "loss": 2.1851,
+      "step": 2568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000947846219512871,
+      "loss": 2.1578,
+      "step": 2569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009477979037954361,
+      "loss": 2.1477,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009477495669409102,
+      "loss": 2.3281,
+      "step": 2571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009477012089515751,
+      "loss": 2.1006,
+      "step": 2572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009476528298297134,
+      "loss": 2.0519,
+      "step": 2573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009476044295776087,
+      "loss": 2.0748,
+      "step": 2574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009475560081975456,
+      "loss": 2.2086,
+      "step": 2575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009475075656918096,
+      "loss": 2.0655,
+      "step": 2576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009474591020626875,
+      "loss": 2.0723,
+      "step": 2577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009474106173124665,
+      "loss": 2.1264,
+      "step": 2578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009473621114434357,
+      "loss": 2.3308,
+      "step": 2579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009473135844578844,
+      "loss": 2.1882,
+      "step": 2580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009472650363581032,
+      "loss": 2.1592,
+      "step": 2581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009472164671463838,
+      "loss": 2.1571,
+      "step": 2582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009471678768250187,
+      "loss": 2.1508,
+      "step": 2583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009471192653963015,
+      "loss": 2.2164,
+      "step": 2584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009470706328625268,
+      "loss": 2.18,
+      "step": 2585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009470219792259898,
+      "loss": 2.1487,
+      "step": 2586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009469733044889877,
+      "loss": 2.1252,
+      "step": 2587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009469246086538175,
+      "loss": 2.2145,
+      "step": 2588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009468758917227781,
+      "loss": 2.1634,
+      "step": 2589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000946827153698169,
+      "loss": 2.185,
+      "step": 2590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009467783945822906,
+      "loss": 2.1023,
+      "step": 2591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009467296143774445,
+      "loss": 2.1118,
+      "step": 2592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009466808130859334,
+      "loss": 1.9771,
+      "step": 2593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009466319907100607,
+      "loss": 2.3356,
+      "step": 2594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009465831472521308,
+      "loss": 2.202,
+      "step": 2595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009465342827144494,
+      "loss": 2.2131,
+      "step": 2596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000946485397099323,
+      "loss": 2.1457,
+      "step": 2597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009464364904090591,
+      "loss": 2.1541,
+      "step": 2598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009463875626459662,
+      "loss": 2.3895,
+      "step": 2599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009463386138123537,
+      "loss": 2.1873,
+      "step": 2600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009462896439105323,
+      "loss": 1.9643,
+      "step": 2601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009462406529428133,
+      "loss": 2.209,
+      "step": 2602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009461916409115094,
+      "loss": 2.1078,
+      "step": 2603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009461426078189339,
+      "loss": 2.1811,
+      "step": 2604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009460935536674013,
+      "loss": 2.1993,
+      "step": 2605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000946044478459227,
+      "loss": 2.1412,
+      "step": 2606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009459953821967278,
+      "loss": 2.1663,
+      "step": 2607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0009459462648822208,
+      "loss": 2.1662,
+      "step": 2608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009458971265180245,
+      "loss": 2.1201,
+      "step": 2609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009458479671064584,
+      "loss": 2.1745,
+      "step": 2610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000945798786649843,
+      "loss": 2.2038,
+      "step": 2611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009457495851504998,
+      "loss": 2.1165,
+      "step": 2612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009457003626107509,
+      "loss": 2.1217,
+      "step": 2613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00094565111903292,
+      "loss": 2.1298,
+      "step": 2614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009456018544193314,
+      "loss": 2.1292,
+      "step": 2615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009455525687723106,
+      "loss": 2.097,
+      "step": 2616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009455032620941839,
+      "loss": 2.3171,
+      "step": 2617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009454539343872788,
+      "loss": 2.1137,
+      "step": 2618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009454045856539235,
+      "loss": 2.2344,
+      "step": 2619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009453552158964475,
+      "loss": 2.189,
+      "step": 2620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009453058251171811,
+      "loss": 2.298,
+      "step": 2621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009452564133184557,
+      "loss": 2.1984,
+      "step": 2622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009452069805026038,
+      "loss": 2.2062,
+      "step": 2623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009451575266719583,
+      "loss": 2.12,
+      "step": 2624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009451080518288541,
+      "loss": 2.1888,
+      "step": 2625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009450585559756261,
+      "loss": 2.2057,
+      "step": 2626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009450090391146109,
+      "loss": 2.1537,
+      "step": 2627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009449595012481455,
+      "loss": 2.2407,
+      "step": 2628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009449099423785685,
+      "loss": 2.0659,
+      "step": 2629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000944860362508219,
+      "loss": 2.264,
+      "step": 2630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009448107616394374,
+      "loss": 2.2325,
+      "step": 2631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000944761139774565,
+      "loss": 2.1964,
+      "step": 2632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009447114969159438,
+      "loss": 2.1674,
+      "step": 2633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009446618330659174,
+      "loss": 2.2391,
+      "step": 2634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00094461214822683,
+      "loss": 2.1023,
+      "step": 2635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009445624424010266,
+      "loss": 2.2125,
+      "step": 2636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009445127155908535,
+      "loss": 2.0906,
+      "step": 2637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009444629677986581,
+      "loss": 2.0949,
+      "step": 2638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009444131990267886,
+      "loss": 2.0518,
+      "step": 2639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000944363409277594,
+      "loss": 2.0643,
+      "step": 2640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009443135985534245,
+      "loss": 2.237,
+      "step": 2641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009442637668566315,
+      "loss": 2.0891,
+      "step": 2642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009442139141895669,
+      "loss": 2.1345,
+      "step": 2643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009441640405545842,
+      "loss": 2.2053,
+      "step": 2644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009441141459540372,
+      "loss": 2.0433,
+      "step": 2645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009440642303902812,
+      "loss": 2.1274,
+      "step": 2646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009440142938656724,
+      "loss": 2.1925,
+      "step": 2647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009439643363825679,
+      "loss": 2.1145,
+      "step": 2648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009439143579433256,
+      "loss": 2.2018,
+      "step": 2649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009438643585503049,
+      "loss": 2.2209,
+      "step": 2650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009438143382058655,
+      "loss": 2.0315,
+      "step": 2651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009437642969123688,
+      "loss": 2.0304,
+      "step": 2652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009437142346721768,
+      "loss": 2.1636,
+      "step": 2653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009436641514876526,
+      "loss": 2.1924,
+      "step": 2654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009436140473611601,
+      "loss": 2.049,
+      "step": 2655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009435639222950644,
+      "loss": 2.167,
+      "step": 2656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009435137762917316,
+      "loss": 2.1405,
+      "step": 2657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009434636093535285,
+      "loss": 2.1523,
+      "step": 2658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009434134214828233,
+      "loss": 2.3637,
+      "step": 2659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009433632126819849,
+      "loss": 2.1924,
+      "step": 2660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009433129829533832,
+      "loss": 2.2268,
+      "step": 2661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009432627322993891,
+      "loss": 2.0733,
+      "step": 2662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000943212460722375,
+      "loss": 2.0506,
+      "step": 2663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009431621682247133,
+      "loss": 2.3345,
+      "step": 2664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009431118548087783,
+      "loss": 2.057,
+      "step": 2665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009430615204769446,
+      "loss": 2.086,
+      "step": 2666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009430111652315883,
+      "loss": 2.1511,
+      "step": 2667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009429607890750862,
+      "loss": 2.1878,
+      "step": 2668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009429103920098161,
+      "loss": 2.2547,
+      "step": 2669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009428599740381571,
+      "loss": 2.0976,
+      "step": 2670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009428095351624888,
+      "loss": 2.1051,
+      "step": 2671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009427590753851921,
+      "loss": 2.0733,
+      "step": 2672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009427085947086489,
+      "loss": 2.181,
+      "step": 2673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009426580931352419,
+      "loss": 2.1914,
+      "step": 2674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009426075706673549,
+      "loss": 2.1241,
+      "step": 2675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009425570273073728,
+      "loss": 2.1326,
+      "step": 2676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009425064630576811,
+      "loss": 2.0848,
+      "step": 2677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000942455877920667,
+      "loss": 2.0962,
+      "step": 2678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009424052718987178,
+      "loss": 2.1982,
+      "step": 2679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009423546449942224,
+      "loss": 2.0595,
+      "step": 2680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009423039972095706,
+      "loss": 2.2405,
+      "step": 2681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009422533285471528,
+      "loss": 2.2235,
+      "step": 2682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000942202639009361,
+      "loss": 2.3599,
+      "step": 2683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009421519285985876,
+      "loss": 2.0219,
+      "step": 2684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009421011973172265,
+      "loss": 2.3809,
+      "step": 2685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000942050445167672,
+      "loss": 2.1284,
+      "step": 2686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00094199967215232,
+      "loss": 2.2304,
+      "step": 2687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000941948878273567,
+      "loss": 2.1148,
+      "step": 2688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009418980635338108,
+      "loss": 2.3031,
+      "step": 2689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009418472279354495,
+      "loss": 2.0866,
+      "step": 2690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009417963714808831,
+      "loss": 2.1838,
+      "step": 2691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000941745494172512,
+      "loss": 2.1966,
+      "step": 2692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009416945960127376,
+      "loss": 2.4274,
+      "step": 2693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009416436770039626,
+      "loss": 2.1898,
+      "step": 2694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009415927371485904,
+      "loss": 2.2346,
+      "step": 2695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009415417764490255,
+      "loss": 2.2314,
+      "step": 2696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009414907949076732,
+      "loss": 2.1396,
+      "step": 2697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009414397925269402,
+      "loss": 2.0313,
+      "step": 2698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009413887693092339,
+      "loss": 2.2347,
+      "step": 2699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009413377252569625,
+      "loss": 2.1272,
+      "step": 2700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009412866603725357,
+      "loss": 2.3109,
+      "step": 2701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009412355746583637,
+      "loss": 2.0265,
+      "step": 2702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009411844681168577,
+      "loss": 2.1141,
+      "step": 2703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009411333407504303,
+      "loss": 2.3072,
+      "step": 2704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009410821925614948,
+      "loss": 2.2275,
+      "step": 2705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009410310235524655,
+      "loss": 2.2256,
+      "step": 2706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009409798337257576,
+      "loss": 2.2549,
+      "step": 2707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009409286230837874,
+      "loss": 2.2014,
+      "step": 2708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009408773916289723,
+      "loss": 2.0575,
+      "step": 2709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009408261393637304,
+      "loss": 2.1677,
+      "step": 2710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009407748662904809,
+      "loss": 2.2406,
+      "step": 2711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009407235724116442,
+      "loss": 2.2713,
+      "step": 2712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009406722577296413,
+      "loss": 2.1988,
+      "step": 2713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009406209222468944,
+      "loss": 2.1985,
+      "step": 2714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009405695659658267,
+      "loss": 2.103,
+      "step": 2715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009405181888888623,
+      "loss": 2.1967,
+      "step": 2716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009404667910184264,
+      "loss": 2.1203,
+      "step": 2717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009404153723569451,
+      "loss": 2.2681,
+      "step": 2718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009403639329068451,
+      "loss": 2.1762,
+      "step": 2719
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009403124726705552,
+      "loss": 2.3119,
+      "step": 2720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009402609916505037,
+      "loss": 2.15,
+      "step": 2721
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009402094898491211,
+      "loss": 2.1176,
+      "step": 2722
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009401579672688382,
+      "loss": 2.023,
+      "step": 2723
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009401064239120869,
+      "loss": 2.2929,
+      "step": 2724
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009400548597813005,
+      "loss": 2.1564,
+      "step": 2725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009400032748789127,
+      "loss": 2.0227,
+      "step": 2726
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009399516692073583,
+      "loss": 2.1335,
+      "step": 2727
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009399000427690735,
+      "loss": 2.1445,
+      "step": 2728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009398483955664949,
+      "loss": 2.043,
+      "step": 2729
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009397967276020606,
+      "loss": 2.0617,
+      "step": 2730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009397450388782096,
+      "loss": 2.1406,
+      "step": 2731
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009396933293973812,
+      "loss": 2.3089,
+      "step": 2732
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009396415991620166,
+      "loss": 2.1648,
+      "step": 2733
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009395898481745575,
+      "loss": 2.0928,
+      "step": 2734
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009395380764374468,
+      "loss": 2.0956,
+      "step": 2735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000939486283953128,
+      "loss": 2.3046,
+      "step": 2736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009394344707240457,
+      "loss": 2.0876,
+      "step": 2737
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009393826367526462,
+      "loss": 2.1991,
+      "step": 2738
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009393307820413755,
+      "loss": 2.1642,
+      "step": 2739
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009392789065926817,
+      "loss": 2.3006,
+      "step": 2740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009392270104090132,
+      "loss": 2.3813,
+      "step": 2741
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009391750934928198,
+      "loss": 2.0179,
+      "step": 2742
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009391231558465521,
+      "loss": 2.053,
+      "step": 2743
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009390711974726616,
+      "loss": 2.1312,
+      "step": 2744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009390192183736008,
+      "loss": 2.1575,
+      "step": 2745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009389672185518232,
+      "loss": 2.068,
+      "step": 2746
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009389151980097836,
+      "loss": 2.1635,
+      "step": 2747
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009388631567499372,
+      "loss": 2.2074,
+      "step": 2748
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009388110947747406,
+      "loss": 2.1751,
+      "step": 2749
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009387590120866512,
+      "loss": 2.1396,
+      "step": 2750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009387069086881272,
+      "loss": 2.2086,
+      "step": 2751
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009386547845816284,
+      "loss": 2.068,
+      "step": 2752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009386026397696152,
+      "loss": 2.2962,
+      "step": 2753
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009385504742545485,
+      "loss": 2.1888,
+      "step": 2754
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000938498288038891,
+      "loss": 2.0844,
+      "step": 2755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009384460811251059,
+      "loss": 2.1668,
+      "step": 2756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0009383938535156575,
+      "loss": 2.0738,
+      "step": 2757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000938341605213011,
+      "loss": 2.3256,
+      "step": 2758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009382893362196328,
+      "loss": 2.185,
+      "step": 2759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009382370465379898,
+      "loss": 2.1899,
+      "step": 2760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009381847361705505,
+      "loss": 2.0104,
+      "step": 2761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000938132405119784,
+      "loss": 2.221,
+      "step": 2762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009380800533881604,
+      "loss": 2.2149,
+      "step": 2763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000938027680978151,
+      "loss": 1.8894,
+      "step": 2764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009379752878922274,
+      "loss": 2.3176,
+      "step": 2765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009379228741328633,
+      "loss": 2.2138,
+      "step": 2766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009378704397025324,
+      "loss": 2.2284,
+      "step": 2767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009378179846037096,
+      "loss": 1.9698,
+      "step": 2768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009377655088388712,
+      "loss": 2.0955,
+      "step": 2769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009377130124104942,
+      "loss": 1.9963,
+      "step": 2770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009376604953210563,
+      "loss": 2.2398,
+      "step": 2771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009376079575730365,
+      "loss": 2.1705,
+      "step": 2772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000937555399168915,
+      "loss": 2.1956,
+      "step": 2773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009375028201111721,
+      "loss": 2.3562,
+      "step": 2774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009374502204022901,
+      "loss": 2.1772,
+      "step": 2775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009373976000447517,
+      "loss": 2.1374,
+      "step": 2776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009373449590410408,
+      "loss": 2.2338,
+      "step": 2777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000937292297393642,
+      "loss": 2.1304,
+      "step": 2778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009372396151050411,
+      "loss": 2.0711,
+      "step": 2779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000937186912177725,
+      "loss": 2.1827,
+      "step": 2780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009371341886141811,
+      "loss": 2.3494,
+      "step": 2781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009370814444168982,
+      "loss": 2.3197,
+      "step": 2782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009370286795883662,
+      "loss": 2.2211,
+      "step": 2783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009369758941310753,
+      "loss": 2.0575,
+      "step": 2784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009369230880475174,
+      "loss": 2.2317,
+      "step": 2785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000936870261340185,
+      "loss": 2.1892,
+      "step": 2786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009368174140115715,
+      "loss": 2.3344,
+      "step": 2787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009367645460641715,
+      "loss": 2.246,
+      "step": 2788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009367116575004806,
+      "loss": 2.1935,
+      "step": 2789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009366587483229951,
+      "loss": 2.1708,
+      "step": 2790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009366058185342126,
+      "loss": 2.0925,
+      "step": 2791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009365528681366313,
+      "loss": 2.1116,
+      "step": 2792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009364998971327507,
+      "loss": 2.3542,
+      "step": 2793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009364469055250712,
+      "loss": 2.269,
+      "step": 2794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009363938933160942,
+      "loss": 2.1075,
+      "step": 2795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009363408605083216,
+      "loss": 2.1897,
+      "step": 2796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009362878071042572,
+      "loss": 2.2601,
+      "step": 2797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000936234733106405,
+      "loss": 2.1768,
+      "step": 2798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009361816385172699,
+      "loss": 2.2561,
+      "step": 2799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009361285233393587,
+      "loss": 2.0853,
+      "step": 2800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009360753875751781,
+      "loss": 2.2514,
+      "step": 2801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009360222312272364,
+      "loss": 2.2666,
+      "step": 2802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009359690542980426,
+      "loss": 2.2481,
+      "step": 2803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009359158567901069,
+      "loss": 2.326,
+      "step": 2804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009358626387059403,
+      "loss": 2.2884,
+      "step": 2805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009358094000480548,
+      "loss": 2.1319,
+      "step": 2806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009357561408189634,
+      "loss": 2.1689,
+      "step": 2807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009357028610211802,
+      "loss": 2.054,
+      "step": 2808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009356495606572197,
+      "loss": 2.2906,
+      "step": 2809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009355962397295983,
+      "loss": 2.0999,
+      "step": 2810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009355428982408328,
+      "loss": 2.4023,
+      "step": 2811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009354895361934405,
+      "loss": 2.191,
+      "step": 2812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009354361535899409,
+      "loss": 2.1407,
+      "step": 2813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009353827504328533,
+      "loss": 2.0792,
+      "step": 2814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009353293267246988,
+      "loss": 2.2282,
+      "step": 2815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000935275882467999,
+      "loss": 2.2156,
+      "step": 2816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009352224176652764,
+      "loss": 2.132,
+      "step": 2817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009351689323190549,
+      "loss": 2.0185,
+      "step": 2818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009351154264318589,
+      "loss": 2.0795,
+      "step": 2819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009350619000062144,
+      "loss": 2.0448,
+      "step": 2820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009350083530446476,
+      "loss": 2.2838,
+      "step": 2821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009349547855496861,
+      "loss": 2.246,
+      "step": 2822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009349011975238586,
+      "loss": 2.1658,
+      "step": 2823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009348475889696943,
+      "loss": 2.0257,
+      "step": 2824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000934793959889724,
+      "loss": 2.2832,
+      "step": 2825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009347403102864787,
+      "loss": 2.0861,
+      "step": 2826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009346866401624909,
+      "loss": 2.2474,
+      "step": 2827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009346329495202943,
+      "loss": 2.1314,
+      "step": 2828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000934579238362423,
+      "loss": 2.223,
+      "step": 2829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000934525506691412,
+      "loss": 2.1384,
+      "step": 2830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000934471754509798,
+      "loss": 2.0924,
+      "step": 2831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009344179818201179,
+      "loss": 2.1337,
+      "step": 2832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009343641886249102,
+      "loss": 2.1265,
+      "step": 2833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009343103749267139,
+      "loss": 2.3531,
+      "step": 2834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000934256540728069,
+      "loss": 2.2399,
+      "step": 2835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009342026860315169,
+      "loss": 2.124,
+      "step": 2836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009341488108395994,
+      "loss": 2.1328,
+      "step": 2837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009340949151548597,
+      "loss": 2.1827,
+      "step": 2838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009340409989798415,
+      "loss": 2.1632,
+      "step": 2839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009339870623170903,
+      "loss": 2.2633,
+      "step": 2840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009339331051691516,
+      "loss": 2.1483,
+      "step": 2841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009338791275385722,
+      "loss": 2.1731,
+      "step": 2842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009338251294279006,
+      "loss": 2.2626,
+      "step": 2843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000933771110839685,
+      "loss": 2.2856,
+      "step": 2844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009337170717764754,
+      "loss": 2.2479,
+      "step": 2845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009336630122408228,
+      "loss": 2.1839,
+      "step": 2846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009336089322352785,
+      "loss": 2.1872,
+      "step": 2847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009335548317623957,
+      "loss": 1.9858,
+      "step": 2848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009335007108247276,
+      "loss": 2.2411,
+      "step": 2849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009334465694248292,
+      "loss": 2.1315,
+      "step": 2850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009333924075652558,
+      "loss": 2.1854,
+      "step": 2851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009333382252485643,
+      "loss": 2.0009,
+      "step": 2852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009332840224773118,
+      "loss": 2.1672,
+      "step": 2853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009332297992540571,
+      "loss": 2.0885,
+      "step": 2854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009331755555813597,
+      "loss": 2.1059,
+      "step": 2855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009331212914617797,
+      "loss": 2.1443,
+      "step": 2856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009330670068978789,
+      "loss": 2.102,
+      "step": 2857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009330127018922195,
+      "loss": 2.2492,
+      "step": 2858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009329583764473646,
+      "loss": 2.2958,
+      "step": 2859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009329040305658787,
+      "loss": 2.0579,
+      "step": 2860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000932849664250327,
+      "loss": 2.0255,
+      "step": 2861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009327952775032757,
+      "loss": 2.2099,
+      "step": 2862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009327408703272922,
+      "loss": 2.1673,
+      "step": 2863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009326864427249444,
+      "loss": 2.1375,
+      "step": 2864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009326319946988015,
+      "loss": 2.2344,
+      "step": 2865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009325775262514334,
+      "loss": 2.1658,
+      "step": 2866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009325230373854114,
+      "loss": 2.3464,
+      "step": 2867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009324685281033073,
+      "loss": 2.1521,
+      "step": 2868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009324139984076942,
+      "loss": 2.1281,
+      "step": 2869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009323594483011458,
+      "loss": 2.3042,
+      "step": 2870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009323048777862373,
+      "loss": 2.1441,
+      "step": 2871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009322502868655445,
+      "loss": 2.0991,
+      "step": 2872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009321956755416438,
+      "loss": 2.1924,
+      "step": 2873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009321410438171136,
+      "loss": 2.1604,
+      "step": 2874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009320863916945323,
+      "loss": 2.1407,
+      "step": 2875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009320317191764796,
+      "loss": 2.0923,
+      "step": 2876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009319770262655363,
+      "loss": 1.9895,
+      "step": 2877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009319223129642839,
+      "loss": 2.0607,
+      "step": 2878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009318675792753051,
+      "loss": 2.1808,
+      "step": 2879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009318128252011833,
+      "loss": 2.0534,
+      "step": 2880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009317580507445032,
+      "loss": 2.0813,
+      "step": 2881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009317032559078501,
+      "loss": 2.1804,
+      "step": 2882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009316484406938106,
+      "loss": 2.1843,
+      "step": 2883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009315936051049721,
+      "loss": 2.1675,
+      "step": 2884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009315387491439229,
+      "loss": 2.1563,
+      "step": 2885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009314838728132523,
+      "loss": 2.1232,
+      "step": 2886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009314289761155507,
+      "loss": 2.1204,
+      "step": 2887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009313740590534092,
+      "loss": 2.1728,
+      "step": 2888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009313191216294201,
+      "loss": 2.165,
+      "step": 2889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009312641638461767,
+      "loss": 2.0696,
+      "step": 2890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009312091857062727,
+      "loss": 2.1268,
+      "step": 2891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009311541872123038,
+      "loss": 2.2525,
+      "step": 2892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009310991683668657,
+      "loss": 2.076,
+      "step": 2893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009310441291725554,
+      "loss": 2.1106,
+      "step": 2894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000930989069631971,
+      "loss": 2.2132,
+      "step": 2895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009309339897477113,
+      "loss": 2.1806,
+      "step": 2896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009308788895223763,
+      "loss": 2.0176,
+      "step": 2897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009308237689585667,
+      "loss": 2.0821,
+      "step": 2898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009307686280588847,
+      "loss": 2.0748,
+      "step": 2899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009307134668259327,
+      "loss": 2.2115,
+      "step": 2900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009306582852623146,
+      "loss": 2.2178,
+      "step": 2901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009306030833706352,
+      "loss": 2.1302,
+      "step": 2902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009305478611534997,
+      "loss": 2.1831,
+      "step": 2903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009304926186135152,
+      "loss": 2.2993,
+      "step": 2904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009304373557532892,
+      "loss": 1.996,
+      "step": 2905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009303820725754299,
+      "loss": 2.2986,
+      "step": 2906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0009303267690825473,
+      "loss": 2.1486,
+      "step": 2907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009302714452772515,
+      "loss": 2.1017,
+      "step": 2908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009302161011621539,
+      "loss": 2.1249,
+      "step": 2909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009301607367398671,
+      "loss": 2.1438,
+      "step": 2910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009301053520130043,
+      "loss": 2.1383,
+      "step": 2911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009300499469841796,
+      "loss": 2.1289,
+      "step": 2912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009299945216560086,
+      "loss": 2.1024,
+      "step": 2913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009299390760311074,
+      "loss": 2.2271,
+      "step": 2914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009298836101120928,
+      "loss": 2.0922,
+      "step": 2915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009298281239015836,
+      "loss": 2.1641,
+      "step": 2916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009297726174021982,
+      "loss": 2.1185,
+      "step": 2917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000929717090616557,
+      "loss": 2.2864,
+      "step": 2918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009296615435472809,
+      "loss": 2.1556,
+      "step": 2919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000929605976196992,
+      "loss": 2.2104,
+      "step": 2920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009295503885683128,
+      "loss": 2.2107,
+      "step": 2921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009294947806638676,
+      "loss": 2.1961,
+      "step": 2922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009294391524862811,
+      "loss": 2.1858,
+      "step": 2923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009293835040381789,
+      "loss": 2.1444,
+      "step": 2924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000929327835322188,
+      "loss": 2.1447,
+      "step": 2925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000929272146340936,
+      "loss": 2.1274,
+      "step": 2926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009292164370970513,
+      "loss": 2.1869,
+      "step": 2927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009291607075931638,
+      "loss": 2.3981,
+      "step": 2928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000929104957831904,
+      "loss": 2.1693,
+      "step": 2929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009290491878159034,
+      "loss": 2.1688,
+      "step": 2930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009289933975477944,
+      "loss": 2.3064,
+      "step": 2931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009289375870302106,
+      "loss": 2.2659,
+      "step": 2932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009288817562657862,
+      "loss": 2.18,
+      "step": 2933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009288259052571565,
+      "loss": 2.1075,
+      "step": 2934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009287700340069582,
+      "loss": 2.0366,
+      "step": 2935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000928714142517828,
+      "loss": 2.2082,
+      "step": 2936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009286582307924046,
+      "loss": 2.1496,
+      "step": 2937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009286022988333267,
+      "loss": 2.0011,
+      "step": 2938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009285463466432348,
+      "loss": 2.2994,
+      "step": 2939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009284903742247698,
+      "loss": 2.2032,
+      "step": 2940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009284343815805737,
+      "loss": 2.1113,
+      "step": 2941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009283783687132896,
+      "loss": 2.1706,
+      "step": 2942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009283223356255613,
+      "loss": 2.1569,
+      "step": 2943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009282662823200339,
+      "loss": 2.1604,
+      "step": 2944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009282102087993529,
+      "loss": 2.1114,
+      "step": 2945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009281541150661654,
+      "loss": 2.1411,
+      "step": 2946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000928098001123119,
+      "loss": 2.1316,
+      "step": 2947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009280418669728625,
+      "loss": 2.182,
+      "step": 2948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009279857126180455,
+      "loss": 2.1693,
+      "step": 2949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009279295380613187,
+      "loss": 2.2419,
+      "step": 2950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009278733433053336,
+      "loss": 2.2401,
+      "step": 2951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009278171283527427,
+      "loss": 2.3149,
+      "step": 2952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009277608932061995,
+      "loss": 2.0469,
+      "step": 2953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009277046378683585,
+      "loss": 2.1115,
+      "step": 2954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009276483623418748,
+      "loss": 2.2365,
+      "step": 2955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009275920666294051,
+      "loss": 2.1127,
+      "step": 2956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009275357507336066,
+      "loss": 2.2253,
+      "step": 2957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009274794146571373,
+      "loss": 2.0462,
+      "step": 2958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009274230584026568,
+      "loss": 2.2521,
+      "step": 2959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009273666819728248,
+      "loss": 2.1894,
+      "step": 2960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009273102853703026,
+      "loss": 2.1355,
+      "step": 2961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009272538685977524,
+      "loss": 2.1925,
+      "step": 2962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000927197431657837,
+      "loss": 2.0744,
+      "step": 2963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009271409745532205,
+      "loss": 2.0725,
+      "step": 2964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009270844972865675,
+      "loss": 2.2665,
+      "step": 2965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009270279998605442,
+      "loss": 2.1103,
+      "step": 2966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009269714822778172,
+      "loss": 2.1451,
+      "step": 2967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009269149445410546,
+      "loss": 2.1403,
+      "step": 2968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009268583866529245,
+      "loss": 2.1956,
+      "step": 2969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009268018086160973,
+      "loss": 2.0717,
+      "step": 2970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009267452104332427,
+      "loss": 2.0522,
+      "step": 2971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009266885921070333,
+      "loss": 2.1777,
+      "step": 2972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009266319536401408,
+      "loss": 2.0625,
+      "step": 2973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009265752950352392,
+      "loss": 2.209,
+      "step": 2974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009265186162950025,
+      "loss": 2.1532,
+      "step": 2975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009264619174221064,
+      "loss": 2.1796,
+      "step": 2976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000926405198419227,
+      "loss": 2.1496,
+      "step": 2977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009263484592890417,
+      "loss": 2.2094,
+      "step": 2978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009262917000342286,
+      "loss": 2.3801,
+      "step": 2979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000926234920657467,
+      "loss": 2.086,
+      "step": 2980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009261781211614368,
+      "loss": 2.1573,
+      "step": 2981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009261213015488192,
+      "loss": 2.0854,
+      "step": 2982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009260644618222964,
+      "loss": 2.1912,
+      "step": 2983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009260076019845511,
+      "loss": 2.0838,
+      "step": 2984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009259507220382673,
+      "loss": 2.2239,
+      "step": 2985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009258938219861299,
+      "loss": 2.0813,
+      "step": 2986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009258369018308248,
+      "loss": 2.258,
+      "step": 2987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009257799615750384,
+      "loss": 2.1926,
+      "step": 2988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009257230012214588,
+      "loss": 2.0996,
+      "step": 2989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009256660207727745,
+      "loss": 2.1194,
+      "step": 2990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009256090202316751,
+      "loss": 2.1846,
+      "step": 2991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009255519996008512,
+      "loss": 2.2845,
+      "step": 2992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009254949588829944,
+      "loss": 2.1623,
+      "step": 2993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009254378980807969,
+      "loss": 2.0396,
+      "step": 2994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009253808171969523,
+      "loss": 2.156,
+      "step": 2995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000925323716234155,
+      "loss": 2.2978,
+      "step": 2996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009252665951951,
+      "loss": 2.1548,
+      "step": 2997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009252094540824838,
+      "loss": 2.2819,
+      "step": 2998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009251522928990035,
+      "loss": 2.0437,
+      "step": 2999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009250951116473574,
+      "loss": 2.2229,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009250379103302442,
+      "loss": 2.233,
+      "step": 3001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009249806889503645,
+      "loss": 2.2276,
+      "step": 3002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009249234475104187,
+      "loss": 2.1481,
+      "step": 3003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009248661860131091,
+      "loss": 2.1277,
+      "step": 3004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009248089044611385,
+      "loss": 2.3889,
+      "step": 3005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009247516028572106,
+      "loss": 2.3282,
+      "step": 3006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009246942812040302,
+      "loss": 2.1618,
+      "step": 3007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009246369395043033,
+      "loss": 2.2016,
+      "step": 3008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009245795777607362,
+      "loss": 2.1696,
+      "step": 3009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009245221959760366,
+      "loss": 2.108,
+      "step": 3010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009244647941529132,
+      "loss": 2.365,
+      "step": 3011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009244073722940752,
+      "loss": 2.1312,
+      "step": 3012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009243499304022333,
+      "loss": 2.0652,
+      "step": 3013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009242924684800989,
+      "loss": 2.0028,
+      "step": 3014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009242349865303842,
+      "loss": 2.2769,
+      "step": 3015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009241774845558025,
+      "loss": 2.312,
+      "step": 3016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009241199625590682,
+      "loss": 2.1576,
+      "step": 3017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009240624205428962,
+      "loss": 2.233,
+      "step": 3018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009240048585100028,
+      "loss": 2.1005,
+      "step": 3019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000923947276463105,
+      "loss": 2.1761,
+      "step": 3020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009238896744049208,
+      "loss": 2.2418,
+      "step": 3021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009238320523381691,
+      "loss": 2.2336,
+      "step": 3022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009237744102655698,
+      "loss": 2.0434,
+      "step": 3023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000923716748189844,
+      "loss": 2.231,
+      "step": 3024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009236590661137132,
+      "loss": 2.2934,
+      "step": 3025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009236013640399001,
+      "loss": 2.1904,
+      "step": 3026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009235436419711285,
+      "loss": 2.1576,
+      "step": 3027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009234858999101231,
+      "loss": 2.2402,
+      "step": 3028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009234281378596093,
+      "loss": 2.194,
+      "step": 3029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009233703558223136,
+      "loss": 1.9999,
+      "step": 3030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009233125538009636,
+      "loss": 2.0472,
+      "step": 3031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009232547317982874,
+      "loss": 2.2478,
+      "step": 3032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009231968898170147,
+      "loss": 2.2223,
+      "step": 3033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009231390278598754,
+      "loss": 2.3744,
+      "step": 3034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009230811459296012,
+      "loss": 2.1829,
+      "step": 3035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009230232440289237,
+      "loss": 2.0798,
+      "step": 3036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009229653221605763,
+      "loss": 2.1593,
+      "step": 3037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000922907380327293,
+      "loss": 2.2129,
+      "step": 3038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009228494185318088,
+      "loss": 2.0975,
+      "step": 3039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009227914367768598,
+      "loss": 2.1295,
+      "step": 3040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009227334350651826,
+      "loss": 2.2133,
+      "step": 3041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000922675413399515,
+      "loss": 2.0648,
+      "step": 3042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000922617371782596,
+      "loss": 2.1743,
+      "step": 3043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009225593102171652,
+      "loss": 2.2226,
+      "step": 3044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009225012287059631,
+      "loss": 2.2085,
+      "step": 3045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009224431272517314,
+      "loss": 2.1209,
+      "step": 3046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009223850058572128,
+      "loss": 2.2258,
+      "step": 3047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009223268645251504,
+      "loss": 2.0689,
+      "step": 3048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009222687032582887,
+      "loss": 2.098,
+      "step": 3049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009222105220593733,
+      "loss": 2.2541,
+      "step": 3050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009221523209311502,
+      "loss": 2.2195,
+      "step": 3051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009220940998763668,
+      "loss": 2.1116,
+      "step": 3052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009220358588977712,
+      "loss": 2.1291,
+      "step": 3053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009219775979981124,
+      "loss": 2.216,
+      "step": 3054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0009219193171801408,
+      "loss": 2.0599,
+      "step": 3055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000921861016446607,
+      "loss": 2.1094,
+      "step": 3056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009218026958002631,
+      "loss": 2.191,
+      "step": 3057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000921744355243862,
+      "loss": 2.3359,
+      "step": 3058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009216859947801573,
+      "loss": 2.1019,
+      "step": 3059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009216276144119041,
+      "loss": 2.121,
+      "step": 3060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009215692141418579,
+      "loss": 2.3451,
+      "step": 3061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009215107939727752,
+      "loss": 2.1928,
+      "step": 3062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009214523539074137,
+      "loss": 2.162,
+      "step": 3063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009213938939485319,
+      "loss": 2.1013,
+      "step": 3064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009213354140988892,
+      "loss": 2.0657,
+      "step": 3065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009212769143612462,
+      "loss": 2.1551,
+      "step": 3066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009212183947383639,
+      "loss": 2.1296,
+      "step": 3067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009211598552330047,
+      "loss": 2.2685,
+      "step": 3068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009211012958479319,
+      "loss": 2.1218,
+      "step": 3069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009210427165859095,
+      "loss": 2.0977,
+      "step": 3070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009209841174497024,
+      "loss": 2.071,
+      "step": 3071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009209254984420771,
+      "loss": 2.258,
+      "step": 3072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009208668595658002,
+      "loss": 2.1413,
+      "step": 3073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009208082008236396,
+      "loss": 1.9855,
+      "step": 3074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009207495222183644,
+      "loss": 2.1735,
+      "step": 3075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000920690823752744,
+      "loss": 2.0987,
+      "step": 3076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009206321054295492,
+      "loss": 2.0803,
+      "step": 3077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009205733672515519,
+      "loss": 2.176,
+      "step": 3078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009205146092215245,
+      "loss": 2.243,
+      "step": 3079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009204558313422402,
+      "loss": 2.0918,
+      "step": 3080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000920397033616474,
+      "loss": 2.3005,
+      "step": 3081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009203382160470011,
+      "loss": 2.1769,
+      "step": 3082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009202793786365975,
+      "loss": 2.1583,
+      "step": 3083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009202205213880409,
+      "loss": 2.1371,
+      "step": 3084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009201616443041093,
+      "loss": 2.2172,
+      "step": 3085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009201027473875818,
+      "loss": 2.136,
+      "step": 3086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009200438306412387,
+      "loss": 2.1713,
+      "step": 3087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009199848940678605,
+      "loss": 2.1363,
+      "step": 3088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009199259376702298,
+      "loss": 2.1535,
+      "step": 3089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000919866961451129,
+      "loss": 2.1385,
+      "step": 3090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009198079654133421,
+      "loss": 2.1234,
+      "step": 3091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009197489495596539,
+      "loss": 2.1467,
+      "step": 3092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009196899138928499,
+      "loss": 2.2616,
+      "step": 3093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009196308584157167,
+      "loss": 2.0461,
+      "step": 3094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009195717831310421,
+      "loss": 2.1434,
+      "step": 3095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009195126880416145,
+      "loss": 2.208,
+      "step": 3096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009194535731502231,
+      "loss": 2.1236,
+      "step": 3097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009193944384596587,
+      "loss": 2.2456,
+      "step": 3098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009193352839727121,
+      "loss": 1.9745,
+      "step": 3099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009192761096921758,
+      "loss": 2.0586,
+      "step": 3100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000919216915620843,
+      "loss": 2.0589,
+      "step": 3101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009191577017615077,
+      "loss": 2.0159,
+      "step": 3102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009190984681169649,
+      "loss": 2.0559,
+      "step": 3103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009190392146900106,
+      "loss": 2.1667,
+      "step": 3104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009189799414834419,
+      "loss": 2.0695,
+      "step": 3105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009189206485000563,
+      "loss": 1.9747,
+      "step": 3106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009188613357426527,
+      "loss": 2.1657,
+      "step": 3107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009188020032140309,
+      "loss": 2.0332,
+      "step": 3108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009187426509169914,
+      "loss": 2.234,
+      "step": 3109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009186832788543358,
+      "loss": 2.0693,
+      "step": 3110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009186238870288665,
+      "loss": 2.1324,
+      "step": 3111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009185644754433872,
+      "loss": 2.1585,
+      "step": 3112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009185050441007021,
+      "loss": 2.2121,
+      "step": 3113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009184455930036163,
+      "loss": 2.169,
+      "step": 3114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009183861221549363,
+      "loss": 2.0791,
+      "step": 3115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009183266315574692,
+      "loss": 2.1318,
+      "step": 3116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009182671212140233,
+      "loss": 2.1636,
+      "step": 3117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000918207591127407,
+      "loss": 2.2436,
+      "step": 3118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000918148041300431,
+      "loss": 2.033,
+      "step": 3119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009180884717359057,
+      "loss": 2.1645,
+      "step": 3120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000918028882436643,
+      "loss": 1.9822,
+      "step": 3121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009179692734054559,
+      "loss": 2.1053,
+      "step": 3122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009179096446451577,
+      "loss": 2.1534,
+      "step": 3123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009178499961585634,
+      "loss": 2.183,
+      "step": 3124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009177903279484883,
+      "loss": 2.2122,
+      "step": 3125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009177306400177487,
+      "loss": 2.0871,
+      "step": 3126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009176709323691625,
+      "loss": 2.0328,
+      "step": 3127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009176112050055478,
+      "loss": 2.0655,
+      "step": 3128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009175514579297238,
+      "loss": 2.1282,
+      "step": 3129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009174916911445109,
+      "loss": 2.1566,
+      "step": 3130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009174319046527298,
+      "loss": 2.0176,
+      "step": 3131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009173720984572031,
+      "loss": 2.3065,
+      "step": 3132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009173122725607534,
+      "loss": 2.0916,
+      "step": 3133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009172524269662047,
+      "loss": 2.0887,
+      "step": 3134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000917192561676382,
+      "loss": 2.3124,
+      "step": 3135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000917132676694111,
+      "loss": 2.187,
+      "step": 3136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009170727720222183,
+      "loss": 2.2203,
+      "step": 3137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009170128476635316,
+      "loss": 2.0546,
+      "step": 3138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009169529036208797,
+      "loss": 2.1517,
+      "step": 3139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009168929398970918,
+      "loss": 2.0052,
+      "step": 3140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009168329564949984,
+      "loss": 2.0444,
+      "step": 3141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009167729534174307,
+      "loss": 2.0912,
+      "step": 3142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009167129306672214,
+      "loss": 2.1456,
+      "step": 3143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009166528882472033,
+      "loss": 2.2451,
+      "step": 3144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009165928261602109,
+      "loss": 1.9771,
+      "step": 3145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009165327444090789,
+      "loss": 2.2499,
+      "step": 3146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009164726429966437,
+      "loss": 2.1832,
+      "step": 3147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009164125219257418,
+      "loss": 2.0879,
+      "step": 3148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009163523811992112,
+      "loss": 2.0886,
+      "step": 3149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000916292220819891,
+      "loss": 2.2499,
+      "step": 3150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009162320407906203,
+      "loss": 2.2417,
+      "step": 3151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009161718411142403,
+      "loss": 2.2266,
+      "step": 3152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009161116217935924,
+      "loss": 2.0838,
+      "step": 3153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009160513828315188,
+      "loss": 2.0504,
+      "step": 3154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009159911242308633,
+      "loss": 2.3639,
+      "step": 3155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009159308459944701,
+      "loss": 2.3888,
+      "step": 3156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009158705481251844,
+      "loss": 2.1918,
+      "step": 3157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009158102306258526,
+      "loss": 2.2767,
+      "step": 3158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009157498934993216,
+      "loss": 2.2534,
+      "step": 3159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009156895367484394,
+      "loss": 2.0922,
+      "step": 3160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009156291603760553,
+      "loss": 2.0128,
+      "step": 3161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009155687643850189,
+      "loss": 2.1663,
+      "step": 3162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009155083487781812,
+      "loss": 2.1439,
+      "step": 3163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000915447913558394,
+      "loss": 2.4474,
+      "step": 3164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009153874587285099,
+      "loss": 2.188,
+      "step": 3165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009153269842913823,
+      "loss": 2.2082,
+      "step": 3166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000915266490249866,
+      "loss": 2.2975,
+      "step": 3167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009152059766068163,
+      "loss": 2.1781,
+      "step": 3168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009151454433650899,
+      "loss": 2.3324,
+      "step": 3169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009150848905275437,
+      "loss": 2.172,
+      "step": 3170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009150243180970361,
+      "loss": 2.187,
+      "step": 3171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009149637260764263,
+      "loss": 2.0682,
+      "step": 3172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009149031144685743,
+      "loss": 2.0156,
+      "step": 3173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009148424832763413,
+      "loss": 2.1979,
+      "step": 3174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009147818325025889,
+      "loss": 2.2716,
+      "step": 3175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009147211621501803,
+      "loss": 2.1329,
+      "step": 3176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009146604722219792,
+      "loss": 2.4255,
+      "step": 3177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00091459976272085,
+      "loss": 2.3211,
+      "step": 3178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009145390336496586,
+      "loss": 2.17,
+      "step": 3179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009144782850112716,
+      "loss": 2.2569,
+      "step": 3180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009144175168085563,
+      "loss": 2.1161,
+      "step": 3181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009143567290443812,
+      "loss": 2.1209,
+      "step": 3182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009142959217216157,
+      "loss": 2.2493,
+      "step": 3183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009142350948431299,
+      "loss": 2.2867,
+      "step": 3184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000914174248411795,
+      "loss": 2.1431,
+      "step": 3185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000914113382430483,
+      "loss": 2.1569,
+      "step": 3186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009140524969020672,
+      "loss": 2.2193,
+      "step": 3187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009139915918294213,
+      "loss": 2.1764,
+      "step": 3188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009139306672154201,
+      "loss": 2.0263,
+      "step": 3189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009138697230629397,
+      "loss": 2.0103,
+      "step": 3190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009138087593748565,
+      "loss": 2.2751,
+      "step": 3191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009137477761540483,
+      "loss": 2.3201,
+      "step": 3192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009136867734033936,
+      "loss": 2.1068,
+      "step": 3193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009136257511257718,
+      "loss": 2.0281,
+      "step": 3194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009135647093240632,
+      "loss": 2.2104,
+      "step": 3195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009135036480011493,
+      "loss": 2.2544,
+      "step": 3196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009134425671599124,
+      "loss": 2.038,
+      "step": 3197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009133814668032356,
+      "loss": 2.184,
+      "step": 3198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009133203469340028,
+      "loss": 2.1202,
+      "step": 3199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009132592075550991,
+      "loss": 2.0666,
+      "step": 3200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009131980486694107,
+      "loss": 2.136,
+      "step": 3201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000913136870279824,
+      "loss": 2.2345,
+      "step": 3202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009130756723892268,
+      "loss": 2.1838,
+      "step": 3203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009130144550005082,
+      "loss": 2.0338,
+      "step": 3204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0009129532181165574,
+      "loss": 2.1775,
+      "step": 3205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009128919617402649,
+      "loss": 2.2627,
+      "step": 3206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009128306858745225,
+      "loss": 2.2133,
+      "step": 3207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009127693905222223,
+      "loss": 2.163,
+      "step": 3208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009127080756862575,
+      "loss": 2.1942,
+      "step": 3209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009126467413695227,
+      "loss": 2.1946,
+      "step": 3210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009125853875749125,
+      "loss": 2.1275,
+      "step": 3211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009125240143053233,
+      "loss": 2.2122,
+      "step": 3212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000912462621563652,
+      "loss": 2.1466,
+      "step": 3213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009124012093527964,
+      "loss": 2.2213,
+      "step": 3214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009123397776756553,
+      "loss": 2.207,
+      "step": 3215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009122783265351285,
+      "loss": 2.2219,
+      "step": 3216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009122168559341164,
+      "loss": 2.3221,
+      "step": 3217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009121553658755208,
+      "loss": 2.122,
+      "step": 3218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009120938563622442,
+      "loss": 2.1585,
+      "step": 3219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00091203232739719,
+      "loss": 2.2272,
+      "step": 3220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009119707789832623,
+      "loss": 2.0899,
+      "step": 3221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009119092111233664,
+      "loss": 2.1452,
+      "step": 3222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009118476238204086,
+      "loss": 2.1971,
+      "step": 3223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009117860170772956,
+      "loss": 2.2222,
+      "step": 3224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009117243908969358,
+      "loss": 2.1707,
+      "step": 3225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009116627452822379,
+      "loss": 2.2011,
+      "step": 3226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009116010802361116,
+      "loss": 2.1305,
+      "step": 3227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009115393957614679,
+      "loss": 2.3635,
+      "step": 3228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009114776918612184,
+      "loss": 2.2365,
+      "step": 3229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009114159685382755,
+      "loss": 2.2331,
+      "step": 3230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009113542257955527,
+      "loss": 2.3177,
+      "step": 3231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009112924636359645,
+      "loss": 2.3433,
+      "step": 3232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009112306820624262,
+      "loss": 2.1324,
+      "step": 3233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000911168881077854,
+      "loss": 2.227,
+      "step": 3234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000911107060685165,
+      "loss": 2.1446,
+      "step": 3235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009110452208872775,
+      "loss": 2.1844,
+      "step": 3236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009109833616871101,
+      "loss": 1.9954,
+      "step": 3237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009109214830875829,
+      "loss": 2.0556,
+      "step": 3238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009108595850916168,
+      "loss": 2.1246,
+      "step": 3239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009107976677021335,
+      "loss": 2.0088,
+      "step": 3240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009107357309220555,
+      "loss": 2.1243,
+      "step": 3241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009106737747543065,
+      "loss": 2.18,
+      "step": 3242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009106117992018108,
+      "loss": 2.2752,
+      "step": 3243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000910549804267494,
+      "loss": 2.0548,
+      "step": 3244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009104877899542823,
+      "loss": 2.0497,
+      "step": 3245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009104257562651029,
+      "loss": 2.1744,
+      "step": 3246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000910363703202884,
+      "loss": 2.0611,
+      "step": 3247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009103016307705545,
+      "loss": 2.2908,
+      "step": 3248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009102395389710446,
+      "loss": 2.1623,
+      "step": 3249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000910177427807285,
+      "loss": 2.1952,
+      "step": 3250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009101152972822075,
+      "loss": 2.2333,
+      "step": 3251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000910053147398745,
+      "loss": 2.2359,
+      "step": 3252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009099909781598308,
+      "loss": 1.9904,
+      "step": 3253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009099287895683995,
+      "loss": 2.2502,
+      "step": 3254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009098665816273868,
+      "loss": 2.1563,
+      "step": 3255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009098043543397287,
+      "loss": 2.2066,
+      "step": 3256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000909742107708363,
+      "loss": 2.1269,
+      "step": 3257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009096798417362273,
+      "loss": 2.1577,
+      "step": 3258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009096175564262609,
+      "loss": 2.3112,
+      "step": 3259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009095552517814038,
+      "loss": 2.1981,
+      "step": 3260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009094929278045972,
+      "loss": 1.9747,
+      "step": 3261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009094305844987827,
+      "loss": 2.1761,
+      "step": 3262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009093682218669028,
+      "loss": 2.3204,
+      "step": 3263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009093058399119016,
+      "loss": 2.0914,
+      "step": 3264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009092434386367234,
+      "loss": 2.1058,
+      "step": 3265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009091810180443139,
+      "loss": 2.21,
+      "step": 3266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009091185781376193,
+      "loss": 2.1127,
+      "step": 3267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000909056118919587,
+      "loss": 2.1608,
+      "step": 3268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009089936403931651,
+      "loss": 2.1653,
+      "step": 3269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009089311425613029,
+      "loss": 2.1906,
+      "step": 3270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009088686254269504,
+      "loss": 1.9996,
+      "step": 3271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009088060889930584,
+      "loss": 2.1875,
+      "step": 3272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000908743533262579,
+      "loss": 2.1468,
+      "step": 3273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009086809582384647,
+      "loss": 2.2206,
+      "step": 3274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009086183639236694,
+      "loss": 2.1925,
+      "step": 3275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009085557503211477,
+      "loss": 2.2324,
+      "step": 3276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009084931174338551,
+      "loss": 2.1145,
+      "step": 3277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009084304652647478,
+      "loss": 2.1861,
+      "step": 3278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009083677938167833,
+      "loss": 2.2612,
+      "step": 3279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009083051030929199,
+      "loss": 2.1332,
+      "step": 3280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009082423930961166,
+      "loss": 2.1181,
+      "step": 3281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009081796638293336,
+      "loss": 2.1368,
+      "step": 3282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009081169152955318,
+      "loss": 1.971,
+      "step": 3283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009080541474976729,
+      "loss": 2.2046,
+      "step": 3284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00090799136043872,
+      "loss": 2.1929,
+      "step": 3285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009079285541216365,
+      "loss": 2.1789,
+      "step": 3286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009078657285493873,
+      "loss": 2.1594,
+      "step": 3287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009078028837249377,
+      "loss": 2.1678,
+      "step": 3288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009077400196512542,
+      "loss": 2.0531,
+      "step": 3289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000907677136331304,
+      "loss": 2.0106,
+      "step": 3290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009076142337680555,
+      "loss": 2.1744,
+      "step": 3291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009075513119644778,
+      "loss": 2.2337,
+      "step": 3292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009074883709235409,
+      "loss": 2.1254,
+      "step": 3293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009074254106482159,
+      "loss": 2.0643,
+      "step": 3294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009073624311414745,
+      "loss": 2.0695,
+      "step": 3295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009072994324062896,
+      "loss": 2.1515,
+      "step": 3296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009072364144456348,
+      "loss": 2.2043,
+      "step": 3297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009071733772624846,
+      "loss": 1.9375,
+      "step": 3298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009071103208598148,
+      "loss": 2.1147,
+      "step": 3299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009070472452406016,
+      "loss": 2.1893,
+      "step": 3300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009069841504078224,
+      "loss": 2.1364,
+      "step": 3301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009069210363644553,
+      "loss": 2.1942,
+      "step": 3302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009068579031134794,
+      "loss": 2.2215,
+      "step": 3303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000906794750657875,
+      "loss": 2.3375,
+      "step": 3304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009067315790006228,
+      "loss": 2.1155,
+      "step": 3305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009066683881447047,
+      "loss": 2.2583,
+      "step": 3306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009066051780931035,
+      "loss": 2.2063,
+      "step": 3307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009065419488488029,
+      "loss": 1.8982,
+      "step": 3308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009064787004147873,
+      "loss": 1.9814,
+      "step": 3309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009064154327940424,
+      "loss": 2.2414,
+      "step": 3310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009063521459895542,
+      "loss": 2.2396,
+      "step": 3311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009062888400043106,
+      "loss": 1.9907,
+      "step": 3312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009062255148412991,
+      "loss": 2.1738,
+      "step": 3313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009061621705035093,
+      "loss": 2.1696,
+      "step": 3314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009060988069939309,
+      "loss": 2.0504,
+      "step": 3315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009060354243155548,
+      "loss": 2.1062,
+      "step": 3316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009059720224713732,
+      "loss": 2.1031,
+      "step": 3317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009059086014643785,
+      "loss": 2.0813,
+      "step": 3318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009058451612975642,
+      "loss": 2.1091,
+      "step": 3319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009057817019739251,
+      "loss": 2.3124,
+      "step": 3320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009057182234964564,
+      "loss": 2.1065,
+      "step": 3321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009056547258681546,
+      "loss": 2.1595,
+      "step": 3322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009055912090920169,
+      "loss": 2.1071,
+      "step": 3323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009055276731710413,
+      "loss": 2.2102,
+      "step": 3324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000905464118108227,
+      "loss": 2.2253,
+      "step": 3325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009054005439065739,
+      "loss": 2.1692,
+      "step": 3326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009053369505690827,
+      "loss": 2.181,
+      "step": 3327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009052733380987554,
+      "loss": 2.0571,
+      "step": 3328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009052097064985947,
+      "loss": 2.1478,
+      "step": 3329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009051460557716038,
+      "loss": 2.2094,
+      "step": 3330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009050823859207875,
+      "loss": 2.0928,
+      "step": 3331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000905018696949151,
+      "loss": 1.9587,
+      "step": 3332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009049549888597004,
+      "loss": 2.0894,
+      "step": 3333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009048912616554432,
+      "loss": 2.1,
+      "step": 3334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009048275153393875,
+      "loss": 2.2448,
+      "step": 3335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009047637499145418,
+      "loss": 2.1803,
+      "step": 3336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009046999653839165,
+      "loss": 2.149,
+      "step": 3337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000904636161750522,
+      "loss": 2.152,
+      "step": 3338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009045723390173704,
+      "loss": 2.3102,
+      "step": 3339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009045084971874737,
+      "loss": 1.9628,
+      "step": 3340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009044446362638459,
+      "loss": 2.0605,
+      "step": 3341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000904380756249501,
+      "loss": 2.0391,
+      "step": 3342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009043168571474547,
+      "loss": 2.1914,
+      "step": 3343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009042529389607228,
+      "loss": 2.2894,
+      "step": 3344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009041890016923225,
+      "loss": 2.2172,
+      "step": 3345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009041250453452719,
+      "loss": 2.2279,
+      "step": 3346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009040610699225897,
+      "loss": 2.0926,
+      "step": 3347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009039970754272959,
+      "loss": 2.2563,
+      "step": 3348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009039330618624111,
+      "loss": 2.2448,
+      "step": 3349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009038690292309568,
+      "loss": 2.1841,
+      "step": 3350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009038049775359556,
+      "loss": 2.1454,
+      "step": 3351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009037409067804308,
+      "loss": 2.2014,
+      "step": 3352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009036768169674067,
+      "loss": 2.2792,
+      "step": 3353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0009036127080999086,
+      "loss": 2.0724,
+      "step": 3354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009035485801809624,
+      "loss": 2.2031,
+      "step": 3355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009034844332135952,
+      "loss": 2.0534,
+      "step": 3356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000903420267200835,
+      "loss": 2.1609,
+      "step": 3357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009033560821457102,
+      "loss": 2.2417,
+      "step": 3358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009032918780512509,
+      "loss": 2.173,
+      "step": 3359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009032276549204875,
+      "loss": 2.0587,
+      "step": 3360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009031634127564515,
+      "loss": 2.1403,
+      "step": 3361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009030991515621752,
+      "loss": 2.1798,
+      "step": 3362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009030348713406921,
+      "loss": 2.0982,
+      "step": 3363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009029705720950361,
+      "loss": 2.2128,
+      "step": 3364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009029062538282425,
+      "loss": 2.1597,
+      "step": 3365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009028419165433471,
+      "loss": 2.132,
+      "step": 3366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009027775602433868,
+      "loss": 2.0845,
+      "step": 3367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009027131849313995,
+      "loss": 2.1735,
+      "step": 3368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009026487906104237,
+      "loss": 2.1788,
+      "step": 3369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000902584377283499,
+      "loss": 2.1042,
+      "step": 3370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009025199449536659,
+      "loss": 2.2236,
+      "step": 3371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009024554936239658,
+      "loss": 2.2108,
+      "step": 3372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009023910232974408,
+      "loss": 2.1627,
+      "step": 3373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009023265339771343,
+      "loss": 2.1144,
+      "step": 3374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009022620256660899,
+      "loss": 2.3048,
+      "step": 3375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009021974983673531,
+      "loss": 2.19,
+      "step": 3376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009021329520839692,
+      "loss": 2.1648,
+      "step": 3377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009020683868189854,
+      "loss": 2.1289,
+      "step": 3378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009020038025754489,
+      "loss": 2.0638,
+      "step": 3379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009019391993564086,
+      "loss": 2.1821,
+      "step": 3380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009018745771649137,
+      "loss": 2.1995,
+      "step": 3381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009018099360040146,
+      "loss": 2.2689,
+      "step": 3382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009017452758767625,
+      "loss": 2.0169,
+      "step": 3383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009016805967862095,
+      "loss": 2.0864,
+      "step": 3384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009016158987354086,
+      "loss": 2.1472,
+      "step": 3385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009015511817274137,
+      "loss": 2.1368,
+      "step": 3386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009014864457652796,
+      "loss": 2.1776,
+      "step": 3387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009014216908520619,
+      "loss": 2.1885,
+      "step": 3388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009013569169908173,
+      "loss": 2.1662,
+      "step": 3389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009012921241846033,
+      "loss": 2.1296,
+      "step": 3390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000901227312436478,
+      "loss": 2.1065,
+      "step": 3391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000901162481749501,
+      "loss": 2.0534,
+      "step": 3392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009010976321267324,
+      "loss": 2.323,
+      "step": 3393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000901032763571233,
+      "loss": 2.0716,
+      "step": 3394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009009678760860649,
+      "loss": 2.1235,
+      "step": 3395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009009029696742909,
+      "loss": 2.1749,
+      "step": 3396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009008380443389751,
+      "loss": 2.043,
+      "step": 3397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009007731000831815,
+      "loss": 2.0463,
+      "step": 3398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009007081369099762,
+      "loss": 2.1548,
+      "step": 3399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009006431548224252,
+      "loss": 2.233,
+      "step": 3400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000900578153823596,
+      "loss": 2.0681,
+      "step": 3401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009005131339165567,
+      "loss": 2.319,
+      "step": 3402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009004480951043763,
+      "loss": 2.2742,
+      "step": 3403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009003830373901251,
+      "loss": 2.2113,
+      "step": 3404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009003179607768736,
+      "loss": 2.2683,
+      "step": 3405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009002528652676939,
+      "loss": 2.2047,
+      "step": 3406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009001877508656585,
+      "loss": 2.2718,
+      "step": 3407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009001226175738409,
+      "loss": 2.0884,
+      "step": 3408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0009000574653953156,
+      "loss": 2.0544,
+      "step": 3409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000899992294333158,
+      "loss": 2.1833,
+      "step": 3410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008999271043904442,
+      "loss": 2.0762,
+      "step": 3411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008998618955702515,
+      "loss": 2.1465,
+      "step": 3412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008997966678756576,
+      "loss": 2.3122,
+      "step": 3413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008997314213097416,
+      "loss": 2.3278,
+      "step": 3414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008996661558755833,
+      "loss": 2.0092,
+      "step": 3415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008996008715762634,
+      "loss": 1.9508,
+      "step": 3416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008995355684148634,
+      "loss": 2.1836,
+      "step": 3417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008994702463944657,
+      "loss": 2.256,
+      "step": 3418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008994049055181536,
+      "loss": 2.1006,
+      "step": 3419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008993395457890116,
+      "loss": 2.1613,
+      "step": 3420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008992741672101248,
+      "loss": 2.1764,
+      "step": 3421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000899208769784579,
+      "loss": 2.1302,
+      "step": 3422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008991433535154612,
+      "loss": 2.1098,
+      "step": 3423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008990779184058593,
+      "loss": 2.3151,
+      "step": 3424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008990124644588618,
+      "loss": 2.1401,
+      "step": 3425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008989469916775583,
+      "loss": 2.2291,
+      "step": 3426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008988815000650395,
+      "loss": 2.1129,
+      "step": 3427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008988159896243966,
+      "loss": 2.1345,
+      "step": 3428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008987504603587218,
+      "loss": 2.256,
+      "step": 3429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008986849122711082,
+      "loss": 2.235,
+      "step": 3430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008986193453646501,
+      "loss": 2.227,
+      "step": 3431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000898553759642442,
+      "loss": 2.0867,
+      "step": 3432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008984881551075801,
+      "loss": 2.015,
+      "step": 3433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008984225317631607,
+      "loss": 2.1227,
+      "step": 3434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008983568896122815,
+      "loss": 1.9689,
+      "step": 3435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008982912286580412,
+      "loss": 2.1442,
+      "step": 3436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008982255489035389,
+      "loss": 2.1253,
+      "step": 3437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000898159850351875,
+      "loss": 2.1268,
+      "step": 3438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008980941330061505,
+      "loss": 2.265,
+      "step": 3439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008980283968694673,
+      "loss": 2.0626,
+      "step": 3440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008979626419449285,
+      "loss": 2.2422,
+      "step": 3441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008978968682356379,
+      "loss": 2.0801,
+      "step": 3442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008978310757446999,
+      "loss": 2.2293,
+      "step": 3443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008977652644752203,
+      "loss": 2.27,
+      "step": 3444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008976994344303057,
+      "loss": 1.9851,
+      "step": 3445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000897633585613063,
+      "loss": 2.2775,
+      "step": 3446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008975677180266007,
+      "loss": 2.1328,
+      "step": 3447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008975018316740278,
+      "loss": 2.2119,
+      "step": 3448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008974359265584544,
+      "loss": 2.1489,
+      "step": 3449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008973700026829912,
+      "loss": 2.2147,
+      "step": 3450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008973040600507501,
+      "loss": 2.188,
+      "step": 3451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008972380986648436,
+      "loss": 2.0043,
+      "step": 3452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008971721185283856,
+      "loss": 2.0878,
+      "step": 3453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00089710611964449,
+      "loss": 2.0538,
+      "step": 3454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008970401020162724,
+      "loss": 2.0969,
+      "step": 3455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008969740656468488,
+      "loss": 2.1382,
+      "step": 3456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008969080105393365,
+      "loss": 2.2648,
+      "step": 3457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008968419366968534,
+      "loss": 2.0895,
+      "step": 3458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008967758441225182,
+      "loss": 2.1364,
+      "step": 3459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008967097328194508,
+      "loss": 2.172,
+      "step": 3460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008966436027907716,
+      "loss": 2.2164,
+      "step": 3461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008965774540396023,
+      "loss": 2.1381,
+      "step": 3462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008965112865690651,
+      "loss": 2.2411,
+      "step": 3463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008964451003822834,
+      "loss": 2.1122,
+      "step": 3464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008963788954823811,
+      "loss": 2.0558,
+      "step": 3465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008963126718724836,
+      "loss": 2.0494,
+      "step": 3466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008962464295557165,
+      "loss": 2.129,
+      "step": 3467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008961801685352066,
+      "loss": 2.2477,
+      "step": 3468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008961138888140817,
+      "loss": 2.2547,
+      "step": 3469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008960475903954703,
+      "loss": 2.2233,
+      "step": 3470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000895981273282502,
+      "loss": 2.0659,
+      "step": 3471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008959149374783068,
+      "loss": 2.1741,
+      "step": 3472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008958485829860163,
+      "loss": 2.3291,
+      "step": 3473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008957822098087622,
+      "loss": 2.279,
+      "step": 3474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008957158179496776,
+      "loss": 2.1866,
+      "step": 3475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008956494074118965,
+      "loss": 2.1791,
+      "step": 3476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008955829781985534,
+      "loss": 2.074,
+      "step": 3477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008955165303127841,
+      "loss": 1.9848,
+      "step": 3478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000895450063757725,
+      "loss": 2.1923,
+      "step": 3479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008953835785365134,
+      "loss": 2.0772,
+      "step": 3480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008953170746522876,
+      "loss": 2.0011,
+      "step": 3481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000895250552108187,
+      "loss": 2.1721,
+      "step": 3482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008951840109073512,
+      "loss": 2.08,
+      "step": 3483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008951174510529214,
+      "loss": 2.0717,
+      "step": 3484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000895050872548039,
+      "loss": 2.215,
+      "step": 3485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008949842753958472,
+      "loss": 2.0509,
+      "step": 3486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008949176595994892,
+      "loss": 2.1971,
+      "step": 3487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008948510251621096,
+      "loss": 2.1885,
+      "step": 3488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008947843720868534,
+      "loss": 2.2137,
+      "step": 3489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008947177003768671,
+      "loss": 2.1868,
+      "step": 3490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008946510100352976,
+      "loss": 2.1882,
+      "step": 3491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008945843010652927,
+      "loss": 2.0558,
+      "step": 3492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008945175734700016,
+      "loss": 2.1257,
+      "step": 3493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008944508272525737,
+      "loss": 2.0554,
+      "step": 3494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008943840624161596,
+      "loss": 2.1932,
+      "step": 3495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008943172789639109,
+      "loss": 2.1006,
+      "step": 3496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008942504768989797,
+      "loss": 2.1052,
+      "step": 3497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008941836562245194,
+      "loss": 2.1454,
+      "step": 3498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000894116816943684,
+      "loss": 2.1219,
+      "step": 3499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008940499590596285,
+      "loss": 2.0212,
+      "step": 3500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008939830825755088,
+      "loss": 2.3407,
+      "step": 3501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008939161874944816,
+      "loss": 2.0454,
+      "step": 3502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0008938492738197044,
+      "loss": 2.0602,
+      "step": 3503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008937823415543357,
+      "loss": 2.3329,
+      "step": 3504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008937153907015349,
+      "loss": 2.1288,
+      "step": 3505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008936484212644625,
+      "loss": 2.208,
+      "step": 3506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008935814332462791,
+      "loss": 2.1365,
+      "step": 3507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008935144266501469,
+      "loss": 2.1534,
+      "step": 3508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000893447401479229,
+      "loss": 2.1297,
+      "step": 3509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008933803577366888,
+      "loss": 2.2264,
+      "step": 3510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008933132954256913,
+      "loss": 2.1183,
+      "step": 3511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008932462145494016,
+      "loss": 2.2175,
+      "step": 3512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008931791151109862,
+      "loss": 2.192,
+      "step": 3513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008931119971136123,
+      "loss": 2.2222,
+      "step": 3514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008930448605604483,
+      "loss": 2.259,
+      "step": 3515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000892977705454663,
+      "loss": 2.0845,
+      "step": 3516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008929105317994261,
+      "loss": 2.265,
+      "step": 3517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008928433395979088,
+      "loss": 2.0487,
+      "step": 3518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008927761288532822,
+      "loss": 2.2872,
+      "step": 3519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008927088995687192,
+      "loss": 2.2924,
+      "step": 3520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008926416517473929,
+      "loss": 2.0875,
+      "step": 3521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000892574385392478,
+      "loss": 2.1644,
+      "step": 3522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000892507100507149,
+      "loss": 2.2112,
+      "step": 3523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008924397970945824,
+      "loss": 2.1617,
+      "step": 3524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008923724751579548,
+      "loss": 2.1601,
+      "step": 3525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008923051347004441,
+      "loss": 2.2048,
+      "step": 3526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008922377757252288,
+      "loss": 2.1143,
+      "step": 3527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008921703982354885,
+      "loss": 1.9467,
+      "step": 3528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008921030022344035,
+      "loss": 2.2586,
+      "step": 3529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008920355877251551,
+      "loss": 2.1138,
+      "step": 3530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008919681547109255,
+      "loss": 2.1207,
+      "step": 3531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008919007031948975,
+      "loss": 2.1042,
+      "step": 3532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008918332331802552,
+      "loss": 2.0736,
+      "step": 3533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000891765744670183,
+      "loss": 2.2401,
+      "step": 3534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008916982376678668,
+      "loss": 2.0993,
+      "step": 3535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000891630712176493,
+      "loss": 2.1234,
+      "step": 3536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008915631681992491,
+      "loss": 2.0512,
+      "step": 3537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008914956057393231,
+      "loss": 2.0611,
+      "step": 3538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008914280247999043,
+      "loss": 2.2737,
+      "step": 3539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008913604253841824,
+      "loss": 2.1856,
+      "step": 3540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008912928074953485,
+      "loss": 2.1823,
+      "step": 3541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008912251711365942,
+      "loss": 2.1885,
+      "step": 3542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008911575163111121,
+      "loss": 2.3193,
+      "step": 3543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008910898430220957,
+      "loss": 2.0821,
+      "step": 3544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008910221512727394,
+      "loss": 1.9031,
+      "step": 3545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008909544410662384,
+      "loss": 1.9778,
+      "step": 3546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008908867124057887,
+      "loss": 2.0895,
+      "step": 3547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000890818965294587,
+      "loss": 2.0674,
+      "step": 3548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008907511997358318,
+      "loss": 2.0179,
+      "step": 3549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000890683415732721,
+      "loss": 2.2772,
+      "step": 3550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008906156132884547,
+      "loss": 2.2082,
+      "step": 3551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008905477924062332,
+      "loss": 2.2645,
+      "step": 3552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008904799530892577,
+      "loss": 2.3015,
+      "step": 3553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008904120953407305,
+      "loss": 2.009,
+      "step": 3554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008903442191638544,
+      "loss": 2.2948,
+      "step": 3555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008902763245618337,
+      "loss": 2.2301,
+      "step": 3556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008902084115378728,
+      "loss": 2.2329,
+      "step": 3557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008901404800951777,
+      "loss": 2.1529,
+      "step": 3558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008900725302369547,
+      "loss": 2.3202,
+      "step": 3559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008900045619664112,
+      "loss": 2.1274,
+      "step": 3560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008899365752867554,
+      "loss": 2.1635,
+      "step": 3561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008898685702011966,
+      "loss": 2.1768,
+      "step": 3562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008898005467129448,
+      "loss": 2.2447,
+      "step": 3563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008897325048252108,
+      "loss": 2.1137,
+      "step": 3564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008896644445412061,
+      "loss": 2.148,
+      "step": 3565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008895963658641438,
+      "loss": 2.183,
+      "step": 3566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008895282687972369,
+      "loss": 2.2839,
+      "step": 3567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008894601533437,
+      "loss": 2.2644,
+      "step": 3568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000889392019506748,
+      "loss": 2.1342,
+      "step": 3569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008893238672895974,
+      "loss": 2.0446,
+      "step": 3570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000889255696695465,
+      "loss": 2.013,
+      "step": 3571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008891875077275684,
+      "loss": 2.1872,
+      "step": 3572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008891193003891266,
+      "loss": 2.1023,
+      "step": 3573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008890510746833589,
+      "loss": 2.2065,
+      "step": 3574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008889828306134857,
+      "loss": 2.1746,
+      "step": 3575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008889145681827284,
+      "loss": 2.1187,
+      "step": 3576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008888462873943093,
+      "loss": 2.1719,
+      "step": 3577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008887779882514509,
+      "loss": 2.0919,
+      "step": 3578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008887096707573776,
+      "loss": 2.1252,
+      "step": 3579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008886413349153139,
+      "loss": 2.1219,
+      "step": 3580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008885729807284854,
+      "loss": 2.2299,
+      "step": 3581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008885046082001187,
+      "loss": 2.2126,
+      "step": 3582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000888436217333441,
+      "loss": 2.031,
+      "step": 3583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008883678081316806,
+      "loss": 2.1221,
+      "step": 3584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008882993805980666,
+      "loss": 2.1252,
+      "step": 3585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008882309347358287,
+      "loss": 2.1045,
+      "step": 3586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000888162470548198,
+      "loss": 2.0826,
+      "step": 3587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000888093988038406,
+      "loss": 2.1866,
+      "step": 3588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008880254872096853,
+      "loss": 2.1159,
+      "step": 3589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008879569680652693,
+      "loss": 2.1799,
+      "step": 3590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008878884306083921,
+      "loss": 2.1537,
+      "step": 3591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008878198748422891,
+      "loss": 2.1484,
+      "step": 3592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008877513007701961,
+      "loss": 2.1396,
+      "step": 3593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008876827083953501,
+      "loss": 2.1781,
+      "step": 3594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008876140977209885,
+      "loss": 2.1373,
+      "step": 3595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008875454687503503,
+      "loss": 2.2202,
+      "step": 3596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008874768214866746,
+      "loss": 2.2336,
+      "step": 3597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008874081559332019,
+      "loss": 2.1287,
+      "step": 3598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008873394720931734,
+      "loss": 2.0561,
+      "step": 3599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008872707699698311,
+      "loss": 2.1785,
+      "step": 3600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008872020495664177,
+      "loss": 2.2085,
+      "step": 3601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008871333108861772,
+      "loss": 2.1298,
+      "step": 3602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008870645539323541,
+      "loss": 2.2551,
+      "step": 3603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008869957787081941,
+      "loss": 2.1644,
+      "step": 3604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008869269852169432,
+      "loss": 2.0775,
+      "step": 3605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008868581734618489,
+      "loss": 2.1555,
+      "step": 3606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000886789343446159,
+      "loss": 2.2048,
+      "step": 3607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008867204951731227,
+      "loss": 2.2377,
+      "step": 3608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008866516286459898,
+      "loss": 2.146,
+      "step": 3609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008865827438680107,
+      "loss": 2.0595,
+      "step": 3610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008865138408424371,
+      "loss": 1.9159,
+      "step": 3611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008864449195725214,
+      "loss": 2.1566,
+      "step": 3612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000886375980061517,
+      "loss": 2.3094,
+      "step": 3613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008863070223126774,
+      "loss": 2.1483,
+      "step": 3614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008862380463292583,
+      "loss": 2.235,
+      "step": 3615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008861690521145152,
+      "loss": 2.2682,
+      "step": 3616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008861000396717045,
+      "loss": 2.2457,
+      "step": 3617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008860310090040842,
+      "loss": 2.0531,
+      "step": 3618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008859619601149125,
+      "loss": 2.1746,
+      "step": 3619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008858928930074487,
+      "loss": 2.1026,
+      "step": 3620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000885823807684953,
+      "loss": 1.9318,
+      "step": 3621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008857547041506861,
+      "loss": 2.194,
+      "step": 3622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008856855824079102,
+      "loss": 2.1605,
+      "step": 3623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008856164424598878,
+      "loss": 2.3502,
+      "step": 3624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008855472843098826,
+      "loss": 2.1619,
+      "step": 3625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008854781079611589,
+      "loss": 2.2487,
+      "step": 3626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000885408913416982,
+      "loss": 2.193,
+      "step": 3627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008853397006806182,
+      "loss": 2.1782,
+      "step": 3628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008852704697553342,
+      "loss": 1.9242,
+      "step": 3629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008852012206443982,
+      "loss": 2.1748,
+      "step": 3630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008851319533510787,
+      "loss": 2.0318,
+      "step": 3631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008850626678786453,
+      "loss": 2.083,
+      "step": 3632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008849933642303685,
+      "loss": 2.0123,
+      "step": 3633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008849240424095194,
+      "loss": 2.1277,
+      "step": 3634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008848547024193705,
+      "loss": 2.2756,
+      "step": 3635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008847853442631944,
+      "loss": 2.2404,
+      "step": 3636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008847159679442653,
+      "loss": 2.1238,
+      "step": 3637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008846465734658577,
+      "loss": 2.1488,
+      "step": 3638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008845771608312474,
+      "loss": 2.1478,
+      "step": 3639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008845077300437107,
+      "loss": 2.1116,
+      "step": 3640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008844382811065248,
+      "loss": 2.0598,
+      "step": 3641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000884368814022968,
+      "loss": 2.1887,
+      "step": 3642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008842993287963192,
+      "loss": 1.9926,
+      "step": 3643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008842298254298584,
+      "loss": 2.1744,
+      "step": 3644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008841603039268661,
+      "loss": 2.1615,
+      "step": 3645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008840907642906242,
+      "loss": 2.1337,
+      "step": 3646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008840212065244149,
+      "loss": 2.1572,
+      "step": 3647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008839516306315215,
+      "loss": 2.1198,
+      "step": 3648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008838820366152281,
+      "loss": 2.1134,
+      "step": 3649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008838124244788199,
+      "loss": 2.1041,
+      "step": 3650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008837427942255824,
+      "loss": 2.2892,
+      "step": 3651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0008836731458588028,
+      "loss": 2.1311,
+      "step": 3652
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008836034793817682,
+      "loss": 2.1957,
+      "step": 3653
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008835337947977672,
+      "loss": 2.0492,
+      "step": 3654
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008834640921100893,
+      "loss": 2.2283,
+      "step": 3655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008833943713220243,
+      "loss": 2.1871,
+      "step": 3656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008833246324368633,
+      "loss": 2.0784,
+      "step": 3657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008832548754578981,
+      "loss": 2.1081,
+      "step": 3658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008831851003884215,
+      "loss": 2.0329,
+      "step": 3659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008831153072317268,
+      "loss": 2.1834,
+      "step": 3660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008830454959911086,
+      "loss": 2.0411,
+      "step": 3661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008829756666698621,
+      "loss": 2.0931,
+      "step": 3662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008829058192712835,
+      "loss": 2.1576,
+      "step": 3663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008828359537986696,
+      "loss": 2.2093,
+      "step": 3664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008827660702553184,
+      "loss": 2.0958,
+      "step": 3665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008826961686445283,
+      "loss": 2.0761,
+      "step": 3666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008826262489695991,
+      "loss": 2.2369,
+      "step": 3667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008825563112338309,
+      "loss": 2.1902,
+      "step": 3668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008824863554405251,
+      "loss": 2.2634,
+      "step": 3669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008824163815929838,
+      "loss": 2.2183,
+      "step": 3670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008823463896945097,
+      "loss": 2.2444,
+      "step": 3671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008822763797484067,
+      "loss": 2.1066,
+      "step": 3672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008822063517579797,
+      "loss": 2.0334,
+      "step": 3673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008821363057265339,
+      "loss": 1.9815,
+      "step": 3674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008820662416573757,
+      "loss": 1.9429,
+      "step": 3675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008819961595538122,
+      "loss": 2.3612,
+      "step": 3676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008819260594191513,
+      "loss": 2.4218,
+      "step": 3677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008818559412567024,
+      "loss": 2.0973,
+      "step": 3678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008817858050697749,
+      "loss": 2.2482,
+      "step": 3679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008817156508616793,
+      "loss": 2.1639,
+      "step": 3680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008816454786357272,
+      "loss": 2.0089,
+      "step": 3681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008815752883952309,
+      "loss": 2.1268,
+      "step": 3682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008815050801435034,
+      "loss": 2.2352,
+      "step": 3683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008814348538838588,
+      "loss": 2.1142,
+      "step": 3684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000881364609619612,
+      "loss": 2.1652,
+      "step": 3685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008812943473540785,
+      "loss": 2.0396,
+      "step": 3686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008812240670905749,
+      "loss": 2.215,
+      "step": 3687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008811537688324187,
+      "loss": 2.1028,
+      "step": 3688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008810834525829282,
+      "loss": 2.2941,
+      "step": 3689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008810131183454223,
+      "loss": 2.1829,
+      "step": 3690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000880942766123221,
+      "loss": 1.9685,
+      "step": 3691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000880872395919645,
+      "loss": 2.0573,
+      "step": 3692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008808020077380163,
+      "loss": 2.0054,
+      "step": 3693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008807316015816569,
+      "loss": 2.2404,
+      "step": 3694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008806611774538904,
+      "loss": 2.0814,
+      "step": 3695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000880590735358041,
+      "loss": 2.1579,
+      "step": 3696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008805202752974335,
+      "loss": 2.2416,
+      "step": 3697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000880449797275394,
+      "loss": 2.1323,
+      "step": 3698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008803793012952491,
+      "loss": 2.0548,
+      "step": 3699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008803087873603266,
+      "loss": 1.9143,
+      "step": 3700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008802382554739546,
+      "loss": 2.2521,
+      "step": 3701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008801677056394626,
+      "loss": 2.2101,
+      "step": 3702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008800971378601806,
+      "loss": 2.1671,
+      "step": 3703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008800265521394395,
+      "loss": 2.1763,
+      "step": 3704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008799559484805714,
+      "loss": 2.1524,
+      "step": 3705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008798853268869087,
+      "loss": 2.0468,
+      "step": 3706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008798146873617848,
+      "loss": 2.1964,
+      "step": 3707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008797440299085344,
+      "loss": 2.2717,
+      "step": 3708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008796733545304924,
+      "loss": 2.0421,
+      "step": 3709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008796026612309948,
+      "loss": 2.1025,
+      "step": 3710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879531950013379,
+      "loss": 2.3561,
+      "step": 3711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879461220880982,
+      "loss": 2.1004,
+      "step": 3712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879390473837143,
+      "loss": 2.3031,
+      "step": 3713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879319708885201,
+      "loss": 2.1442,
+      "step": 3714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008792489260284965,
+      "loss": 2.1019,
+      "step": 3715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008791781252703706,
+      "loss": 2.1242,
+      "step": 3716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879107306614165,
+      "loss": 2.1915,
+      "step": 3717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000879036470063223,
+      "loss": 2.1918,
+      "step": 3718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008789656156208876,
+      "loss": 2.0523,
+      "step": 3719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008788947432905039,
+      "loss": 2.2095,
+      "step": 3720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008788238530754169,
+      "loss": 2.034,
+      "step": 3721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008787529449789729,
+      "loss": 2.1192,
+      "step": 3722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008786820190045189,
+      "loss": 2.0995,
+      "step": 3723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008786110751554026,
+      "loss": 2.2334,
+      "step": 3724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008785401134349731,
+      "loss": 2.0456,
+      "step": 3725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008784691338465795,
+      "loss": 2.326,
+      "step": 3726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008783981363935727,
+      "loss": 2.1575,
+      "step": 3727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008783271210793034,
+      "loss": 2.0704,
+      "step": 3728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008782560879071242,
+      "loss": 2.1472,
+      "step": 3729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008781850368803876,
+      "loss": 2.1903,
+      "step": 3730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008781139680024476,
+      "loss": 1.9612,
+      "step": 3731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008780428812766587,
+      "loss": 2.1491,
+      "step": 3732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008779717767063766,
+      "loss": 2.2596,
+      "step": 3733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008779006542949573,
+      "loss": 2.1802,
+      "step": 3734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008778295140457582,
+      "loss": 2.1745,
+      "step": 3735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000877758355962137,
+      "loss": 2.0555,
+      "step": 3736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008776871800474529,
+      "loss": 1.9852,
+      "step": 3737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008776159863050651,
+      "loss": 2.1122,
+      "step": 3738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008775447747383345,
+      "loss": 2.0618,
+      "step": 3739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008774735453506222,
+      "loss": 2.1445,
+      "step": 3740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008774022981452905,
+      "loss": 2.1084,
+      "step": 3741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008773310331257025,
+      "loss": 2.1511,
+      "step": 3742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000877259750295222,
+      "loss": 2.1888,
+      "step": 3743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008771884496572137,
+      "loss": 2.2582,
+      "step": 3744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008771171312150431,
+      "loss": 2.2996,
+      "step": 3745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008770457949720767,
+      "loss": 2.2331,
+      "step": 3746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008769744409316818,
+      "loss": 2.0908,
+      "step": 3747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008769030690972262,
+      "loss": 2.1047,
+      "step": 3748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000876831679472079,
+      "loss": 2.1399,
+      "step": 3749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00087676027205961,
+      "loss": 2.1077,
+      "step": 3750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008766888468631897,
+      "loss": 2.1935,
+      "step": 3751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008766174038861896,
+      "loss": 2.0919,
+      "step": 3752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000876545943131982,
+      "loss": 2.1168,
+      "step": 3753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008764744646039399,
+      "loss": 2.1213,
+      "step": 3754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008764029683054373,
+      "loss": 2.1152,
+      "step": 3755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008763314542398489,
+      "loss": 2.2513,
+      "step": 3756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008762599224105505,
+      "loss": 2.2452,
+      "step": 3757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008761883728209185,
+      "loss": 2.1486,
+      "step": 3758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008761168054743303,
+      "loss": 2.1631,
+      "step": 3759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008760452203741638,
+      "loss": 2.2836,
+      "step": 3760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000875973617523798,
+      "loss": 2.1289,
+      "step": 3761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008759019969266131,
+      "loss": 2.049,
+      "step": 3762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008758303585859892,
+      "loss": 2.122,
+      "step": 3763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008757587025053084,
+      "loss": 2.1231,
+      "step": 3764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008756870286879526,
+      "loss": 2.2028,
+      "step": 3765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000875615337137305,
+      "loss": 2.0997,
+      "step": 3766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008755436278567497,
+      "loss": 2.1686,
+      "step": 3767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008754719008496716,
+      "loss": 2.1085,
+      "step": 3768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008754001561194564,
+      "loss": 2.1889,
+      "step": 3769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008753283936694904,
+      "loss": 2.1316,
+      "step": 3770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008752566135031612,
+      "loss": 2.2245,
+      "step": 3771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008751848156238567,
+      "loss": 2.2825,
+      "step": 3772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008751130000349663,
+      "loss": 2.1773,
+      "step": 3773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008750411667398795,
+      "loss": 2.0072,
+      "step": 3774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008749693157419872,
+      "loss": 2.1513,
+      "step": 3775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000874897447044681,
+      "loss": 2.2818,
+      "step": 3776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008748255606513531,
+      "loss": 2.2043,
+      "step": 3777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008747536565653966,
+      "loss": 2.0488,
+      "step": 3778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008746817347902059,
+      "loss": 2.0326,
+      "step": 3779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008746097953291756,
+      "loss": 2.2233,
+      "step": 3780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008745378381857014,
+      "loss": 2.1714,
+      "step": 3781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008744658633631802,
+      "loss": 1.9576,
+      "step": 3782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008743938708650088,
+      "loss": 2.0437,
+      "step": 3783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008743218606945858,
+      "loss": 2.08,
+      "step": 3784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008742498328553101,
+      "loss": 2.0627,
+      "step": 3785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008741777873505818,
+      "loss": 2.1787,
+      "step": 3786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008741057241838014,
+      "loss": 1.9683,
+      "step": 3787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008740336433583705,
+      "loss": 2.2466,
+      "step": 3788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008739615448776914,
+      "loss": 1.9505,
+      "step": 3789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008738894287451674,
+      "loss": 2.0268,
+      "step": 3790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008738172949642026,
+      "loss": 2.1137,
+      "step": 3791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008737451435382018,
+      "loss": 2.1737,
+      "step": 3792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008736729744705707,
+      "loss": 2.1225,
+      "step": 3793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000873600787764716,
+      "loss": 2.1717,
+      "step": 3794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008735285834240448,
+      "loss": 2.0747,
+      "step": 3795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008734563614519655,
+      "loss": 2.1633,
+      "step": 3796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008733841218518872,
+      "loss": 2.1586,
+      "step": 3797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008733118646272195,
+      "loss": 2.1359,
+      "step": 3798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008732395897813735,
+      "loss": 2.2065,
+      "step": 3799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008731672973177605,
+      "loss": 2.2657,
+      "step": 3800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0008730949872397929,
+      "loss": 2.1676,
+      "step": 3801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008730226595508838,
+      "loss": 2.1505,
+      "step": 3802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008729503142544475,
+      "loss": 2.239,
+      "step": 3803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008728779513538985,
+      "loss": 2.1696,
+      "step": 3804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008728055708526528,
+      "loss": 2.1067,
+      "step": 3805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000872733172754127,
+      "loss": 2.1288,
+      "step": 3806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000872660757061738,
+      "loss": 2.1158,
+      "step": 3807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008725883237789045,
+      "loss": 2.2146,
+      "step": 3808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008725158729090451,
+      "loss": 2.0519,
+      "step": 3809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00087244340445558,
+      "loss": 2.1276,
+      "step": 3810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008723709184219295,
+      "loss": 2.1221,
+      "step": 3811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008722984148115155,
+      "loss": 2.0658,
+      "step": 3812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00087222589362776,
+      "loss": 2.2469,
+      "step": 3813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008721533548740866,
+      "loss": 2.1738,
+      "step": 3814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008720807985539187,
+      "loss": 2.0825,
+      "step": 3815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008720082246706817,
+      "loss": 1.951,
+      "step": 3816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008719356332278008,
+      "loss": 2.1631,
+      "step": 3817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008718630242287028,
+      "loss": 2.0061,
+      "step": 3818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008717903976768149,
+      "loss": 2.081,
+      "step": 3819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008717177535755652,
+      "loss": 2.0197,
+      "step": 3820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008716450919283827,
+      "loss": 2.3863,
+      "step": 3821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008715724127386971,
+      "loss": 2.2738,
+      "step": 3822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008714997160099392,
+      "loss": 2.096,
+      "step": 3823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008714270017455405,
+      "loss": 2.0773,
+      "step": 3824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000871354269948933,
+      "loss": 2.1719,
+      "step": 3825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00087128152062355,
+      "loss": 2.161,
+      "step": 3826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008712087537728255,
+      "loss": 2.1663,
+      "step": 3827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000871135969400194,
+      "loss": 2.1536,
+      "step": 3828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008710631675090913,
+      "loss": 2.1876,
+      "step": 3829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008709903481029539,
+      "loss": 2.1547,
+      "step": 3830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008709175111852187,
+      "loss": 2.0711,
+      "step": 3831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008708446567593241,
+      "loss": 2.2231,
+      "step": 3832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008707717848287089,
+      "loss": 2.2672,
+      "step": 3833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008706988953968128,
+      "loss": 2.2332,
+      "step": 3834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008706259884670764,
+      "loss": 2.0524,
+      "step": 3835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008705530640429409,
+      "loss": 2.176,
+      "step": 3836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008704801221278487,
+      "loss": 2.3452,
+      "step": 3837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008704071627252428,
+      "loss": 2.3011,
+      "step": 3838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008703341858385668,
+      "loss": 2.0566,
+      "step": 3839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008702611914712657,
+      "loss": 1.9528,
+      "step": 3840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000870188179626785,
+      "loss": 2.2745,
+      "step": 3841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008701151503085708,
+      "loss": 2.2525,
+      "step": 3842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008700421035200702,
+      "loss": 2.0334,
+      "step": 3843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008699690392647317,
+      "loss": 2.137,
+      "step": 3844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008698959575460035,
+      "loss": 2.1853,
+      "step": 3845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008698228583673356,
+      "loss": 2.2287,
+      "step": 3846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008697497417321782,
+      "loss": 2.1592,
+      "step": 3847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008696766076439825,
+      "loss": 2.124,
+      "step": 3848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008696034561062011,
+      "loss": 2.1187,
+      "step": 3849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008695302871222865,
+      "loss": 2.1115,
+      "step": 3850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008694571006956926,
+      "loss": 2.1038,
+      "step": 3851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000869383896829874,
+      "loss": 2.2448,
+      "step": 3852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008693106755282859,
+      "loss": 2.0008,
+      "step": 3853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008692374367943845,
+      "loss": 2.2844,
+      "step": 3854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000869164180631627,
+      "loss": 2.187,
+      "step": 3855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008690909070434714,
+      "loss": 2.0268,
+      "step": 3856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000869017616033376,
+      "loss": 2.0995,
+      "step": 3857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008689443076048005,
+      "loss": 2.2018,
+      "step": 3858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008688709817612054,
+      "loss": 2.0365,
+      "step": 3859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008687976385060517,
+      "loss": 2.2155,
+      "step": 3860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000868724277842801,
+      "loss": 2.0595,
+      "step": 3861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008686508997749167,
+      "loss": 2.1135,
+      "step": 3862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008685775043058621,
+      "loss": 2.1397,
+      "step": 3863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008685040914391018,
+      "loss": 2.0084,
+      "step": 3864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008684306611781009,
+      "loss": 2.2676,
+      "step": 3865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008683572135263255,
+      "loss": 1.9959,
+      "step": 3866
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008682837484872424,
+      "loss": 2.2395,
+      "step": 3867
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008682102660643197,
+      "loss": 2.2185,
+      "step": 3868
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008681367662610255,
+      "loss": 2.23,
+      "step": 3869
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008680632490808294,
+      "loss": 2.2122,
+      "step": 3870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008679897145272013,
+      "loss": 2.1705,
+      "step": 3871
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008679161626036127,
+      "loss": 2.0706,
+      "step": 3872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000867842593313535,
+      "loss": 2.2801,
+      "step": 3873
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008677690066604411,
+      "loss": 2.1001,
+      "step": 3874
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008676954026478042,
+      "loss": 2.178,
+      "step": 3875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008676217812790988,
+      "loss": 1.9005,
+      "step": 3876
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008675481425577998,
+      "loss": 2.1723,
+      "step": 3877
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008674744864873834,
+      "loss": 2.2385,
+      "step": 3878
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000867400813071326,
+      "loss": 2.1822,
+      "step": 3879
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008673271223131054,
+      "loss": 2.1726,
+      "step": 3880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008672534142162,
+      "loss": 2.02,
+      "step": 3881
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008671796887840887,
+      "loss": 2.1262,
+      "step": 3882
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008671059460202517,
+      "loss": 2.2154,
+      "step": 3883
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00086703218592817,
+      "loss": 2.0404,
+      "step": 3884
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000866958408511325,
+      "loss": 1.9807,
+      "step": 3885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008668846137731992,
+      "loss": 2.16,
+      "step": 3886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008668108017172759,
+      "loss": 2.026,
+      "step": 3887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008667369723470393,
+      "loss": 2.088,
+      "step": 3888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008666631256659741,
+      "loss": 2.1427,
+      "step": 3889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008665892616775661,
+      "loss": 2.1673,
+      "step": 3890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008665153803853021,
+      "loss": 2.0812,
+      "step": 3891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008664414817926694,
+      "loss": 2.0696,
+      "step": 3892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008663675659031558,
+      "loss": 2.1119,
+      "step": 3893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008662936327202506,
+      "loss": 1.9932,
+      "step": 3894
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008662196822474436,
+      "loss": 2.1748,
+      "step": 3895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008661457144882254,
+      "loss": 2.1024,
+      "step": 3896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008660717294460876,
+      "loss": 2.2139,
+      "step": 3897
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008659977271245223,
+      "loss": 2.1132,
+      "step": 3898
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008659237075270225,
+      "loss": 2.2653,
+      "step": 3899
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008658496706570822,
+      "loss": 2.1023,
+      "step": 3900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008657756165181961,
+      "loss": 2.1838,
+      "step": 3901
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00086570154511386,
+      "loss": 2.2367,
+      "step": 3902
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008656274564475696,
+      "loss": 2.1909,
+      "step": 3903
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008655533505228226,
+      "loss": 2.1286,
+      "step": 3904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008654792273431167,
+      "loss": 2.1831,
+      "step": 3905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008654050869119509,
+      "loss": 2.1923,
+      "step": 3906
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008653309292328247,
+      "loss": 2.127,
+      "step": 3907
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008652567543092385,
+      "loss": 2.2012,
+      "step": 3908
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008651825621446934,
+      "loss": 2.0962,
+      "step": 3909
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008651083527426917,
+      "loss": 2.0908,
+      "step": 3910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000865034126106736,
+      "loss": 2.096,
+      "step": 3911
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008649598822403302,
+      "loss": 2.0624,
+      "step": 3912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008648856211469787,
+      "loss": 2.0749,
+      "step": 3913
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008648113428301867,
+      "loss": 2.1455,
+      "step": 3914
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008647370472934604,
+      "loss": 1.9786,
+      "step": 3915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008646627345403066,
+      "loss": 2.1031,
+      "step": 3916
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008645884045742332,
+      "loss": 2.2165,
+      "step": 3917
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008645140573987487,
+      "loss": 2.1119,
+      "step": 3918
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008644396930173624,
+      "loss": 2.0461,
+      "step": 3919
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008643653114335843,
+      "loss": 2.1168,
+      "step": 3920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008642909126509258,
+      "loss": 2.1174,
+      "step": 3921
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008642164966728985,
+      "loss": 2.1903,
+      "step": 3922
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008641420635030148,
+      "loss": 2.1673,
+      "step": 3923
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008640676131447886,
+      "loss": 2.315,
+      "step": 3924
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008639931456017335,
+      "loss": 2.4866,
+      "step": 3925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008639186608773651,
+      "loss": 2.0937,
+      "step": 3926
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008638441589751989,
+      "loss": 2.1154,
+      "step": 3927
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008637696398987515,
+      "loss": 2.11,
+      "step": 3928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008636951036515408,
+      "loss": 2.1832,
+      "step": 3929
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008636205502370849,
+      "loss": 2.2508,
+      "step": 3930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008635459796589026,
+      "loss": 2.1627,
+      "step": 3931
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008634713919205141,
+      "loss": 2.261,
+      "step": 3932
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00086339678702544,
+      "loss": 2.1305,
+      "step": 3933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008633221649772018,
+      "loss": 2.0577,
+      "step": 3934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008632475257793221,
+      "loss": 2.1699,
+      "step": 3935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008631728694353237,
+      "loss": 2.2254,
+      "step": 3936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008630981959487307,
+      "loss": 2.085,
+      "step": 3937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008630235053230679,
+      "loss": 2.0712,
+      "step": 3938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008629487975618607,
+      "loss": 2.1378,
+      "step": 3939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008628740726686357,
+      "loss": 2.1181,
+      "step": 3940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008627993306469199,
+      "loss": 2.0008,
+      "step": 3941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008627245715002414,
+      "loss": 2.1159,
+      "step": 3942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000862649795232129,
+      "loss": 2.1101,
+      "step": 3943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008625750018461122,
+      "loss": 2.0782,
+      "step": 3944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008625001913457216,
+      "loss": 2.0851,
+      "step": 3945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008624253637344883,
+      "loss": 2.1591,
+      "step": 3946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008623505190159444,
+      "loss": 2.0489,
+      "step": 3947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008622756571936228,
+      "loss": 1.9804,
+      "step": 3948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008622007782710568,
+      "loss": 2.0712,
+      "step": 3949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0008621258822517813,
+      "loss": 1.9776,
+      "step": 3950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008620509691393315,
+      "loss": 2.0745,
+      "step": 3951
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008619760389372432,
+      "loss": 2.2769,
+      "step": 3952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008619010916490535,
+      "loss": 2.1386,
+      "step": 3953
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008618261272782999,
+      "loss": 2.1756,
+      "step": 3954
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008617511458285213,
+      "loss": 2.1746,
+      "step": 3955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008616761473032564,
+      "loss": 2.165,
+      "step": 3956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008616011317060457,
+      "loss": 2.0985,
+      "step": 3957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00086152609904043,
+      "loss": 2.1728,
+      "step": 3958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000861451049309951,
+      "loss": 2.1529,
+      "step": 3959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008613759825181514,
+      "loss": 2.1432,
+      "step": 3960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008613008986685743,
+      "loss": 2.0252,
+      "step": 3961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008612257977647639,
+      "loss": 2.1492,
+      "step": 3962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008611506798102651,
+      "loss": 1.9331,
+      "step": 3963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008610755448086238,
+      "loss": 2.3996,
+      "step": 3964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008610003927633864,
+      "loss": 2.0799,
+      "step": 3965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008609252236781003,
+      "loss": 2.0053,
+      "step": 3966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008608500375563136,
+      "loss": 2.0981,
+      "step": 3967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008607748344015751,
+      "loss": 2.1717,
+      "step": 3968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008606996142174351,
+      "loss": 2.1489,
+      "step": 3969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008606243770074435,
+      "loss": 2.0408,
+      "step": 3970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008605491227751521,
+      "loss": 2.0278,
+      "step": 3971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008604738515241131,
+      "loss": 2.1461,
+      "step": 3972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008603985632578792,
+      "loss": 2.0479,
+      "step": 3973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008603232579800043,
+      "loss": 2.3391,
+      "step": 3974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000860247935694043,
+      "loss": 2.0211,
+      "step": 3975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008601725964035508,
+      "loss": 2.1614,
+      "step": 3976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008600972401120836,
+      "loss": 2.0837,
+      "step": 3977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008600218668231987,
+      "loss": 2.1225,
+      "step": 3978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008599464765404537,
+      "loss": 2.0632,
+      "step": 3979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008598710692674072,
+      "loss": 1.9372,
+      "step": 3980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008597956450076187,
+      "loss": 2.0911,
+      "step": 3981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008597202037646483,
+      "loss": 2.2102,
+      "step": 3982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008596447455420571,
+      "loss": 2.1116,
+      "step": 3983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000859569270343407,
+      "loss": 2.192,
+      "step": 3984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008594937781722603,
+      "loss": 1.9537,
+      "step": 3985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008594182690321805,
+      "loss": 2.2934,
+      "step": 3986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008593427429267318,
+      "loss": 2.0463,
+      "step": 3987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008592671998594795,
+      "loss": 2.1187,
+      "step": 3988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000859191639833989,
+      "loss": 1.9445,
+      "step": 3989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000859116062853827,
+      "loss": 2.1426,
+      "step": 3990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008590404689225611,
+      "loss": 2.1027,
+      "step": 3991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008589648580437593,
+      "loss": 2.0196,
+      "step": 3992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008588892302209908,
+      "loss": 2.0926,
+      "step": 3993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008588135854578253,
+      "loss": 2.2104,
+      "step": 3994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008587379237578334,
+      "loss": 2.2081,
+      "step": 3995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008586622451245865,
+      "loss": 2.2055,
+      "step": 3996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008585865495616569,
+      "loss": 2.1004,
+      "step": 3997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008585108370726173,
+      "loss": 2.1412,
+      "step": 3998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008584351076610419,
+      "loss": 1.9965,
+      "step": 3999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008583593613305052,
+      "loss": 2.0826,
+      "step": 4000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008582835980845825,
+      "loss": 2.2406,
+      "step": 4001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00085820781792685,
+      "loss": 2.2412,
+      "step": 4002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008581320208608847,
+      "loss": 2.1459,
+      "step": 4003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008580562068902644,
+      "loss": 2.1868,
+      "step": 4004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008579803760185678,
+      "loss": 2.1936,
+      "step": 4005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008579045282493743,
+      "loss": 2.1429,
+      "step": 4006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008578286635862638,
+      "loss": 2.0035,
+      "step": 4007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008577527820328176,
+      "loss": 2.0458,
+      "step": 4008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008576768835926174,
+      "loss": 2.0359,
+      "step": 4009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008576009682692456,
+      "loss": 2.4481,
+      "step": 4010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000857525036066286,
+      "loss": 2.0852,
+      "step": 4011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008574490869873222,
+      "loss": 2.0248,
+      "step": 4012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008573731210359397,
+      "loss": 2.2338,
+      "step": 4013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008572971382157239,
+      "loss": 2.0863,
+      "step": 4014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008572211385302617,
+      "loss": 2.0851,
+      "step": 4015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008571451219831401,
+      "loss": 2.0966,
+      "step": 4016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008570690885779475,
+      "loss": 2.111,
+      "step": 4017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008569930383182728,
+      "loss": 2.1434,
+      "step": 4018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008569169712077056,
+      "loss": 2.0924,
+      "step": 4019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008568408872498367,
+      "loss": 2.3417,
+      "step": 4020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008567647864482573,
+      "loss": 2.0884,
+      "step": 4021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008566886688065595,
+      "loss": 2.194,
+      "step": 4022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008566125343283362,
+      "loss": 2.014,
+      "step": 4023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008565363830171813,
+      "loss": 2.0998,
+      "step": 4024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008564602148766891,
+      "loss": 2.1853,
+      "step": 4025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008563840299104552,
+      "loss": 2.2855,
+      "step": 4026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008563078281220753,
+      "loss": 2.0138,
+      "step": 4027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008562316095151466,
+      "loss": 2.1275,
+      "step": 4028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008561553740932668,
+      "loss": 2.0571,
+      "step": 4029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008560791218600343,
+      "loss": 2.1379,
+      "step": 4030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008560028528190483,
+      "loss": 2.1994,
+      "step": 4031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008559265669739089,
+      "loss": 2.1439,
+      "step": 4032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008558502643282171,
+      "loss": 2.1379,
+      "step": 4033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008557739448855745,
+      "loss": 2.1409,
+      "step": 4034
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008556976086495834,
+      "loss": 2.1383,
+      "step": 4035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008556212556238474,
+      "loss": 2.1193,
+      "step": 4036
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008555448858119702,
+      "loss": 2.1402,
+      "step": 4037
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008554684992175568,
+      "loss": 2.1354,
+      "step": 4038
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008553920958442127,
+      "loss": 2.1206,
+      "step": 4039
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008553156756955445,
+      "loss": 2.2038,
+      "step": 4040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008552392387751593,
+      "loss": 2.2341,
+      "step": 4041
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000855162785086665,
+      "loss": 2.093,
+      "step": 4042
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008550863146336705,
+      "loss": 2.0757,
+      "step": 4043
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008550098274197852,
+      "loss": 2.1049,
+      "step": 4044
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008549333234486201,
+      "loss": 2.1732,
+      "step": 4045
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008548568027237855,
+      "loss": 2.1017,
+      "step": 4046
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008547802652488939,
+      "loss": 2.276,
+      "step": 4047
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000854703711027558,
+      "loss": 2.2693,
+      "step": 4048
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008546271400633911,
+      "loss": 2.169,
+      "step": 4049
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008545505523600077,
+      "loss": 2.0965,
+      "step": 4050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008544739479210228,
+      "loss": 2.1433,
+      "step": 4051
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008543973267500525,
+      "loss": 2.2287,
+      "step": 4052
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008543206888507132,
+      "loss": 1.9972,
+      "step": 4053
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008542440342266227,
+      "loss": 2.1602,
+      "step": 4054
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008541673628813992,
+      "loss": 2.2496,
+      "step": 4055
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008540906748186616,
+      "loss": 1.9333,
+      "step": 4056
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008540139700420299,
+      "loss": 2.2582,
+      "step": 4057
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008539372485551246,
+      "loss": 2.0973,
+      "step": 4058
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008538605103615675,
+      "loss": 2.1992,
+      "step": 4059
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008537837554649803,
+      "loss": 2.1474,
+      "step": 4060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008537069838689864,
+      "loss": 2.2093,
+      "step": 4061
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008536301955772092,
+      "loss": 2.1743,
+      "step": 4062
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008535533905932737,
+      "loss": 2.2627,
+      "step": 4063
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008534765689208053,
+      "loss": 2.092,
+      "step": 4064
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008533997305634296,
+      "loss": 2.0215,
+      "step": 4065
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008533228755247742,
+      "loss": 2.1062,
+      "step": 4066
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008532460038084665,
+      "loss": 1.9399,
+      "step": 4067
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008531691154181349,
+      "loss": 2.1367,
+      "step": 4068
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008530922103574091,
+      "loss": 2.0884,
+      "step": 4069
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008530152886299187,
+      "loss": 2.1921,
+      "step": 4070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008529383502392952,
+      "loss": 2.2036,
+      "step": 4071
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008528613951891697,
+      "loss": 2.2469,
+      "step": 4072
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008527844234831749,
+      "loss": 1.9923,
+      "step": 4073
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000852707435124944,
+      "loss": 2.0615,
+      "step": 4074
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008526304301181111,
+      "loss": 2.2065,
+      "step": 4075
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000852553408466311,
+      "loss": 2.1843,
+      "step": 4076
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008524763701731792,
+      "loss": 2.0929,
+      "step": 4077
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008523993152423522,
+      "loss": 2.1317,
+      "step": 4078
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008523222436774672,
+      "loss": 2.1366,
+      "step": 4079
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008522451554821619,
+      "loss": 2.1819,
+      "step": 4080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008521680506600752,
+      "loss": 2.1592,
+      "step": 4081
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008520909292148468,
+      "loss": 2.0225,
+      "step": 4082
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008520137911501167,
+      "loss": 2.1624,
+      "step": 4083
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008519366364695262,
+      "loss": 2.0348,
+      "step": 4084
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008518594651767172,
+      "loss": 2.2003,
+      "step": 4085
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008517822772753322,
+      "loss": 2.1333,
+      "step": 4086
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008517050727690148,
+      "loss": 1.9757,
+      "step": 4087
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008516278516614091,
+      "loss": 2.0574,
+      "step": 4088
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008515506139561601,
+      "loss": 2.1332,
+      "step": 4089
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008514733596569138,
+      "loss": 2.0783,
+      "step": 4090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008513960887673164,
+      "loss": 2.1663,
+      "step": 4091
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008513188012910158,
+      "loss": 2.124,
+      "step": 4092
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008512414972316598,
+      "loss": 2.2511,
+      "step": 4093
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008511641765928974,
+      "loss": 2.3277,
+      "step": 4094
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008510868393783781,
+      "loss": 2.0552,
+      "step": 4095
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008510094855917528,
+      "loss": 2.1297,
+      "step": 4096
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008509321152366726,
+      "loss": 2.1309,
+      "step": 4097
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008508547283167894,
+      "loss": 2.1761,
+      "step": 4098
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0008507773248357562,
+      "loss": 2.0513,
+      "step": 4099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008506999047972267,
+      "loss": 2.1639,
+      "step": 4100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008506224682048551,
+      "loss": 2.0939,
+      "step": 4101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008505450150622966,
+      "loss": 2.1426,
+      "step": 4102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008504675453732075,
+      "loss": 2.2415,
+      "step": 4103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008503900591412441,
+      "loss": 2.1483,
+      "step": 4104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008503125563700643,
+      "loss": 2.2125,
+      "step": 4105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000850235037063326,
+      "loss": 2.1059,
+      "step": 4106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008501575012246887,
+      "loss": 2.0789,
+      "step": 4107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008500799488578119,
+      "loss": 2.101,
+      "step": 4108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008500023799663565,
+      "loss": 2.1796,
+      "step": 4109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008499247945539841,
+      "loss": 2.1836,
+      "step": 4110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008498471926243565,
+      "loss": 2.1729,
+      "step": 4111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008497695741811369,
+      "loss": 2.045,
+      "step": 4112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008496919392279892,
+      "loss": 2.12,
+      "step": 4113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008496142877685775,
+      "loss": 2.1325,
+      "step": 4114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008495366198065676,
+      "loss": 2.0385,
+      "step": 4115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008494589353456255,
+      "loss": 2.1916,
+      "step": 4116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008493812343894179,
+      "loss": 2.1392,
+      "step": 4117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008493035169416125,
+      "loss": 2.0342,
+      "step": 4118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008492257830058778,
+      "loss": 2.1078,
+      "step": 4119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008491480325858831,
+      "loss": 2.1252,
+      "step": 4120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008490702656852985,
+      "loss": 2.0261,
+      "step": 4121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008489924823077943,
+      "loss": 2.1105,
+      "step": 4122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008489146824570426,
+      "loss": 2.1791,
+      "step": 4123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008488368661367155,
+      "loss": 2.228,
+      "step": 4124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000848759033350486,
+      "loss": 1.9646,
+      "step": 4125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008486811841020282,
+      "loss": 2.0832,
+      "step": 4126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008486033183950166,
+      "loss": 1.9339,
+      "step": 4127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008485254362331267,
+      "loss": 2.1419,
+      "step": 4128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008484475376200346,
+      "loss": 2.0635,
+      "step": 4129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008483696225594176,
+      "loss": 2.1046,
+      "step": 4130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008482916910549532,
+      "loss": 2.1325,
+      "step": 4131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00084821374311032,
+      "loss": 2.2081,
+      "step": 4132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008481357787291974,
+      "loss": 2.1166,
+      "step": 4133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008480577979152655,
+      "loss": 2.0474,
+      "step": 4134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008479798006722051,
+      "loss": 2.1008,
+      "step": 4135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008479017870036978,
+      "loss": 2.1473,
+      "step": 4136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000847823756913426,
+      "loss": 2.0618,
+      "step": 4137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008477457104050731,
+      "loss": 2.1021,
+      "step": 4138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000847667647482323,
+      "loss": 2.0317,
+      "step": 4139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008475895681488604,
+      "loss": 2.1461,
+      "step": 4140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008475114724083708,
+      "loss": 2.1345,
+      "step": 4141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008474333602645405,
+      "loss": 2.29,
+      "step": 4142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008473552317210567,
+      "loss": 2.274,
+      "step": 4143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008472770867816071,
+      "loss": 2.2723,
+      "step": 4144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008471989254498804,
+      "loss": 2.199,
+      "step": 4145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000847120747729566,
+      "loss": 2.0236,
+      "step": 4146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000847042553624354,
+      "loss": 2.1488,
+      "step": 4147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008469643431379355,
+      "loss": 2.1409,
+      "step": 4148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008468861162740021,
+      "loss": 2.2387,
+      "step": 4149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000846807873036246,
+      "loss": 2.0215,
+      "step": 4150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008467296134283611,
+      "loss": 2.1038,
+      "step": 4151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008466513374540409,
+      "loss": 2.1818,
+      "step": 4152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008465730451169805,
+      "loss": 2.0732,
+      "step": 4153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008464947364208753,
+      "loss": 2.2705,
+      "step": 4154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008464164113694216,
+      "loss": 2.1276,
+      "step": 4155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008463380699663168,
+      "loss": 2.2503,
+      "step": 4156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008462597122152585,
+      "loss": 1.9185,
+      "step": 4157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008461813381199456,
+      "loss": 2.2436,
+      "step": 4158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008461029476840773,
+      "loss": 2.251,
+      "step": 4159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008460245409113541,
+      "loss": 2.0066,
+      "step": 4160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008459461178054768,
+      "loss": 2.0962,
+      "step": 4161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008458676783701471,
+      "loss": 2.1841,
+      "step": 4162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008457892226090678,
+      "loss": 2.1449,
+      "step": 4163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008457107505259418,
+      "loss": 2.1499,
+      "step": 4164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008456322621244734,
+      "loss": 2.0979,
+      "step": 4165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008455537574083673,
+      "loss": 2.1079,
+      "step": 4166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008454752363813294,
+      "loss": 2.1775,
+      "step": 4167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008453966990470657,
+      "loss": 2.261,
+      "step": 4168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008453181454092835,
+      "loss": 2.0978,
+      "step": 4169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008452395754716908,
+      "loss": 2.1266,
+      "step": 4170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008451609892379961,
+      "loss": 1.8156,
+      "step": 4171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008450823867119091,
+      "loss": 2.1727,
+      "step": 4172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008450037678971399,
+      "loss": 2.1527,
+      "step": 4173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008449251327973992,
+      "loss": 2.1514,
+      "step": 4174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008448464814163992,
+      "loss": 1.912,
+      "step": 4175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008447678137578521,
+      "loss": 2.084,
+      "step": 4176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008446891298254714,
+      "loss": 2.1512,
+      "step": 4177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008446104296229712,
+      "loss": 2.0466,
+      "step": 4178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008445317131540662,
+      "loss": 2.2251,
+      "step": 4179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008444529804224719,
+      "loss": 2.1645,
+      "step": 4180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000844374231431905,
+      "loss": 2.2208,
+      "step": 4181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008442954661860823,
+      "loss": 1.8484,
+      "step": 4182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008442166846887218,
+      "loss": 2.1673,
+      "step": 4183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008441378869435423,
+      "loss": 2.3022,
+      "step": 4184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008440590729542629,
+      "loss": 2.0315,
+      "step": 4185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008439802427246043,
+      "loss": 2.1447,
+      "step": 4186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008439013962582872,
+      "loss": 2.2063,
+      "step": 4187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008438225335590333,
+      "loss": 2.0801,
+      "step": 4188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000843743654630565,
+      "loss": 2.2966,
+      "step": 4189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008436647594766058,
+      "loss": 2.1526,
+      "step": 4190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008435858481008796,
+      "loss": 2.2728,
+      "step": 4191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008435069205071113,
+      "loss": 2.1381,
+      "step": 4192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008434279766990263,
+      "loss": 2.1308,
+      "step": 4193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008433490166803513,
+      "loss": 1.9926,
+      "step": 4194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008432700404548128,
+      "loss": 1.9654,
+      "step": 4195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008431910480261391,
+      "loss": 2.0353,
+      "step": 4196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008431120393980587,
+      "loss": 2.1425,
+      "step": 4197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000843033014574301,
+      "loss": 2.1929,
+      "step": 4198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008429539735585961,
+      "loss": 1.9753,
+      "step": 4199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008428749163546749,
+      "loss": 2.2406,
+      "step": 4200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008427958429662694,
+      "loss": 2.1937,
+      "step": 4201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008427167533971116,
+      "loss": 2.0604,
+      "step": 4202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000842637647650935,
+      "loss": 2.1888,
+      "step": 4203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008425585257314734,
+      "loss": 2.2339,
+      "step": 4204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008424793876424617,
+      "loss": 2.0744,
+      "step": 4205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008424002333876353,
+      "loss": 2.0604,
+      "step": 4206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008423210629707303,
+      "loss": 2.138,
+      "step": 4207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000842241876395484,
+      "loss": 2.1399,
+      "step": 4208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008421626736656343,
+      "loss": 2.1604,
+      "step": 4209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008420834547849192,
+      "loss": 2.125,
+      "step": 4210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008420042197570786,
+      "loss": 2.0594,
+      "step": 4211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008419249685858522,
+      "loss": 2.0937,
+      "step": 4212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000841845701274981,
+      "loss": 2.3407,
+      "step": 4213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008417664178282066,
+      "loss": 2.1587,
+      "step": 4214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008416871182492713,
+      "loss": 2.2502,
+      "step": 4215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008416078025419183,
+      "loss": 2.217,
+      "step": 4216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008415284707098913,
+      "loss": 2.1792,
+      "step": 4217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008414491227569352,
+      "loss": 2.1626,
+      "step": 4218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008413697586867954,
+      "loss": 2.0462,
+      "step": 4219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008412903785032178,
+      "loss": 2.186,
+      "step": 4220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008412109822099494,
+      "loss": 2.1823,
+      "step": 4221
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000841131569810738,
+      "loss": 2.124,
+      "step": 4222
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008410521413093322,
+      "loss": 2.1761,
+      "step": 4223
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008409726967094808,
+      "loss": 2.2046,
+      "step": 4224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008408932360149342,
+      "loss": 2.2504,
+      "step": 4225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008408137592294427,
+      "loss": 2.1709,
+      "step": 4226
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008407342663567581,
+      "loss": 2.0532,
+      "step": 4227
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008406547574006325,
+      "loss": 2.1186,
+      "step": 4228
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000840575232364819,
+      "loss": 2.115,
+      "step": 4229
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008404956912530713,
+      "loss": 2.143,
+      "step": 4230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008404161340691441,
+      "loss": 1.8944,
+      "step": 4231
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008403365608167923,
+      "loss": 2.0759,
+      "step": 4232
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008402569714997723,
+      "loss": 2.1455,
+      "step": 4233
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008401773661218407,
+      "loss": 2.0639,
+      "step": 4234
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008400977446867552,
+      "loss": 2.1043,
+      "step": 4235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000840018107198274,
+      "loss": 2.1687,
+      "step": 4236
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008399384536601562,
+      "loss": 2.0167,
+      "step": 4237
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008398587840761616,
+      "loss": 2.091,
+      "step": 4238
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008397790984500509,
+      "loss": 2.2195,
+      "step": 4239
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008396993967855852,
+      "loss": 2.0557,
+      "step": 4240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008396196790865269,
+      "loss": 2.0797,
+      "step": 4241
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008395399453566388,
+      "loss": 2.2004,
+      "step": 4242
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008394601955996844,
+      "loss": 2.2225,
+      "step": 4243
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008393804298194281,
+      "loss": 2.2556,
+      "step": 4244
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008393006480196351,
+      "loss": 2.0626,
+      "step": 4245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008392208502040712,
+      "loss": 2.236,
+      "step": 4246
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008391410363765031,
+      "loss": 1.9574,
+      "step": 4247
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0008390612065406982,
+      "loss": 2.0057,
+      "step": 4248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008389813607004244,
+      "loss": 2.2255,
+      "step": 4249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000838901498859451,
+      "loss": 2.4481,
+      "step": 4250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008388216210215476,
+      "loss": 2.0785,
+      "step": 4251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008387417271904842,
+      "loss": 1.953,
+      "step": 4252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008386618173700326,
+      "loss": 2.1104,
+      "step": 4253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008385818915639643,
+      "loss": 2.1106,
+      "step": 4254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000838501949776052,
+      "loss": 2.2563,
+      "step": 4255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008384219920100693,
+      "loss": 1.981,
+      "step": 4256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008383420182697903,
+      "loss": 1.9463,
+      "step": 4257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00083826202855899,
+      "loss": 1.9859,
+      "step": 4258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008381820228814441,
+      "loss": 2.0685,
+      "step": 4259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000838102001240929,
+      "loss": 2.1389,
+      "step": 4260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008380219636412218,
+      "loss": 2.2298,
+      "step": 4261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008379419100861007,
+      "loss": 2.2073,
+      "step": 4262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008378618405793442,
+      "loss": 2.0698,
+      "step": 4263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008377817551247318,
+      "loss": 2.1266,
+      "step": 4264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008377016537260439,
+      "loss": 2.0453,
+      "step": 4265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008376215363870613,
+      "loss": 1.9755,
+      "step": 4266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008375414031115657,
+      "loss": 2.209,
+      "step": 4267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008374612539033397,
+      "loss": 2.0566,
+      "step": 4268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008373810887661666,
+      "loss": 2.087,
+      "step": 4269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008373009077038301,
+      "loss": 2.1943,
+      "step": 4270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008372207107201151,
+      "loss": 2.1673,
+      "step": 4271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008371404978188069,
+      "loss": 2.1722,
+      "step": 4272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008370602690036921,
+      "loss": 2.1668,
+      "step": 4273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008369800242785573,
+      "loss": 2.196,
+      "step": 4274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008368997636471905,
+      "loss": 2.1754,
+      "step": 4275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008368194871133802,
+      "loss": 2.1591,
+      "step": 4276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008367391946809155,
+      "loss": 2.0294,
+      "step": 4277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008366588863535862,
+      "loss": 2.2879,
+      "step": 4278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008365785621351836,
+      "loss": 2.1689,
+      "step": 4279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008364982220294985,
+      "loss": 2.0311,
+      "step": 4280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008364178660403237,
+      "loss": 2.096,
+      "step": 4281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008363374941714518,
+      "loss": 2.1095,
+      "step": 4282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008362571064266771,
+      "loss": 2.213,
+      "step": 4283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008361767028097933,
+      "loss": 2.1825,
+      "step": 4284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008360962833245962,
+      "loss": 2.1835,
+      "step": 4285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008360158479748816,
+      "loss": 2.0298,
+      "step": 4286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008359353967644462,
+      "loss": 2.1218,
+      "step": 4287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008358549296970876,
+      "loss": 2.1082,
+      "step": 4288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008357744467766038,
+      "loss": 2.1967,
+      "step": 4289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008356939480067941,
+      "loss": 1.9933,
+      "step": 4290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000835613433391458,
+      "loss": 2.0807,
+      "step": 4291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008355329029343961,
+      "loss": 2.2091,
+      "step": 4292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008354523566394095,
+      "loss": 2.0445,
+      "step": 4293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008353717945103003,
+      "loss": 2.1881,
+      "step": 4294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008352912165508711,
+      "loss": 2.0177,
+      "step": 4295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008352106227649254,
+      "loss": 2.2261,
+      "step": 4296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008351300131562675,
+      "loss": 2.1169,
+      "step": 4297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008350493877287023,
+      "loss": 2.1976,
+      "step": 4298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008349687464860355,
+      "loss": 2.0482,
+      "step": 4299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008348880894320735,
+      "loss": 2.0517,
+      "step": 4300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008348074165706235,
+      "loss": 2.05,
+      "step": 4301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008347267279054935,
+      "loss": 2.0913,
+      "step": 4302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008346460234404923,
+      "loss": 2.2088,
+      "step": 4303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008345653031794292,
+      "loss": 2.0155,
+      "step": 4304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008344845671261143,
+      "loss": 2.0888,
+      "step": 4305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008344038152843587,
+      "loss": 2.1182,
+      "step": 4306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008343230476579739,
+      "loss": 2.0064,
+      "step": 4307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008342422642507725,
+      "loss": 2.2204,
+      "step": 4308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008341614650665677,
+      "loss": 2.2851,
+      "step": 4309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000834080650109173,
+      "loss": 1.9181,
+      "step": 4310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008339998193824035,
+      "loss": 2.1489,
+      "step": 4311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008339189728900745,
+      "loss": 2.123,
+      "step": 4312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008338381106360019,
+      "loss": 2.2755,
+      "step": 4313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008337572326240028,
+      "loss": 1.9899,
+      "step": 4314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000833676338857895,
+      "loss": 2.0625,
+      "step": 4315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008335954293414965,
+      "loss": 2.1202,
+      "step": 4316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008335145040786265,
+      "loss": 2.0814,
+      "step": 4317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000833433563073105,
+      "loss": 2.0449,
+      "step": 4318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008333526063287526,
+      "loss": 2.1138,
+      "step": 4319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008332716338493904,
+      "loss": 2.1349,
+      "step": 4320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008331906456388409,
+      "loss": 2.0833,
+      "step": 4321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008331096417009265,
+      "loss": 2.1224,
+      "step": 4322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008330286220394711,
+      "loss": 2.1455,
+      "step": 4323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000832947586658299,
+      "loss": 2.0799,
+      "step": 4324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008328665355612348,
+      "loss": 2.2203,
+      "step": 4325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008327854687521048,
+      "loss": 2.1325,
+      "step": 4326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008327043862347355,
+      "loss": 2.3386,
+      "step": 4327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000832623288012954,
+      "loss": 2.2593,
+      "step": 4328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008325421740905883,
+      "loss": 2.1035,
+      "step": 4329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008324610444714673,
+      "loss": 2.0751,
+      "step": 4330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008323798991594203,
+      "loss": 2.0328,
+      "step": 4331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008322987381582779,
+      "loss": 2.107,
+      "step": 4332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008322175614718707,
+      "loss": 2.1474,
+      "step": 4333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008321363691040308,
+      "loss": 2.0153,
+      "step": 4334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008320551610585903,
+      "loss": 2.1006,
+      "step": 4335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008319739373393826,
+      "loss": 2.0913,
+      "step": 4336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008318926979502419,
+      "loss": 2.1745,
+      "step": 4337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008318114428950024,
+      "loss": 2.0587,
+      "step": 4338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008317301721774997,
+      "loss": 2.1349,
+      "step": 4339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00083164888580157,
+      "loss": 2.1884,
+      "step": 4340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008315675837710502,
+      "loss": 2.141,
+      "step": 4341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008314862660897779,
+      "loss": 2.0727,
+      "step": 4342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008314049327615918,
+      "loss": 2.0926,
+      "step": 4343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008313235837903306,
+      "loss": 2.1281,
+      "step": 4344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008312422191798343,
+      "loss": 2.1646,
+      "step": 4345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008311608389339433,
+      "loss": 2.0158,
+      "step": 4346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008310794430564995,
+      "loss": 2.2196,
+      "step": 4347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008309980315513443,
+      "loss": 2.1558,
+      "step": 4348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008309166044223211,
+      "loss": 2.1615,
+      "step": 4349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000830835161673273,
+      "loss": 2.0525,
+      "step": 4350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008307537033080446,
+      "loss": 2.0734,
+      "step": 4351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008306722293304806,
+      "loss": 2.1514,
+      "step": 4352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000830590739744427,
+      "loss": 2.0676,
+      "step": 4353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008305092345537303,
+      "loss": 2.2166,
+      "step": 4354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008304277137622376,
+      "loss": 2.1999,
+      "step": 4355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008303461773737969,
+      "loss": 2.2168,
+      "step": 4356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000830264625392257,
+      "loss": 2.1145,
+      "step": 4357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008301830578214673,
+      "loss": 2.2077,
+      "step": 4358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000830101474665278,
+      "loss": 2.029,
+      "step": 4359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008300198759275398,
+      "loss": 2.263,
+      "step": 4360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008299382616121046,
+      "loss": 2.3056,
+      "step": 4361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008298566317228248,
+      "loss": 2.1088,
+      "step": 4362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008297749862635533,
+      "loss": 2.2198,
+      "step": 4363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008296933252381441,
+      "loss": 2.2341,
+      "step": 4364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008296116486504518,
+      "loss": 1.9859,
+      "step": 4365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008295299565043318,
+      "loss": 2.1224,
+      "step": 4366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008294482488036397,
+      "loss": 1.9822,
+      "step": 4367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008293665255522329,
+      "loss": 2.207,
+      "step": 4368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008292847867539685,
+      "loss": 2.2508,
+      "step": 4369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008292030324127052,
+      "loss": 2.2238,
+      "step": 4370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008291212625323015,
+      "loss": 2.2783,
+      "step": 4371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008290394771166172,
+      "loss": 2.1371,
+      "step": 4372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008289576761695131,
+      "loss": 2.1512,
+      "step": 4373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008288758596948502,
+      "loss": 2.0292,
+      "step": 4374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008287940276964903,
+      "loss": 2.1758,
+      "step": 4375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008287121801782962,
+      "loss": 2.2294,
+      "step": 4376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008286303171441314,
+      "loss": 2.2307,
+      "step": 4377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008285484385978598,
+      "loss": 2.2367,
+      "step": 4378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008284665445433463,
+      "loss": 2.0743,
+      "step": 4379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008283846349844566,
+      "loss": 2.0541,
+      "step": 4380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008283027099250568,
+      "loss": 2.0869,
+      "step": 4381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008282207693690142,
+      "loss": 2.157,
+      "step": 4382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008281388133201965,
+      "loss": 2.3068,
+      "step": 4383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008280568417824723,
+      "loss": 2.1555,
+      "step": 4384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008279748547597107,
+      "loss": 2.2094,
+      "step": 4385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008278928522557816,
+      "loss": 2.1591,
+      "step": 4386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008278108342745559,
+      "loss": 2.2056,
+      "step": 4387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000827728800819905,
+      "loss": 2.2347,
+      "step": 4388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008276467518957009,
+      "loss": 2.1039,
+      "step": 4389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008275646875058169,
+      "loss": 2.11,
+      "step": 4390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008274826076541261,
+      "loss": 2.2137,
+      "step": 4391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008274005123445032,
+      "loss": 2.2663,
+      "step": 4392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008273184015808231,
+      "loss": 2.2122,
+      "step": 4393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008272362753669618,
+      "loss": 2.4023,
+      "step": 4394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008271541337067959,
+      "loss": 2.1032,
+      "step": 4395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008270719766042024,
+      "loss": 2.1613,
+      "step": 4396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0008269898040630595,
+      "loss": 2.054,
+      "step": 4397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008269076160872459,
+      "loss": 2.2179,
+      "step": 4398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000826825412680641,
+      "loss": 1.9864,
+      "step": 4399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000826743193847125,
+      "loss": 2.0908,
+      "step": 4400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008266609595905789,
+      "loss": 2.0185,
+      "step": 4401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008265787099148842,
+      "loss": 2.0788,
+      "step": 4402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008264964448239234,
+      "loss": 1.9708,
+      "step": 4403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008264141643215796,
+      "loss": 2.1564,
+      "step": 4404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008263318684117365,
+      "loss": 2.0643,
+      "step": 4405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008262495570982789,
+      "loss": 2.1964,
+      "step": 4406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000826167230385092,
+      "loss": 2.1194,
+      "step": 4407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008260848882760616,
+      "loss": 2.066,
+      "step": 4408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008260025307750746,
+      "loss": 2.173,
+      "step": 4409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008259201578860184,
+      "loss": 2.2883,
+      "step": 4410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008258377696127814,
+      "loss": 2.0243,
+      "step": 4411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008257553659592522,
+      "loss": 2.1282,
+      "step": 4412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008256729469293207,
+      "loss": 2.1955,
+      "step": 4413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008255905125268771,
+      "loss": 2.1067,
+      "step": 4414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008255080627558127,
+      "loss": 2.2405,
+      "step": 4415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000825425597620019,
+      "loss": 2.0146,
+      "step": 4416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008253431171233888,
+      "loss": 2.188,
+      "step": 4417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008252606212698153,
+      "loss": 2.0631,
+      "step": 4418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008251781100631924,
+      "loss": 2.0982,
+      "step": 4419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000825095583507415,
+      "loss": 2.1239,
+      "step": 4420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008250130416063784,
+      "loss": 2.2574,
+      "step": 4421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008249304843639788,
+      "loss": 2.1065,
+      "step": 4422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000824847911784113,
+      "loss": 2.0152,
+      "step": 4423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000824765323870679,
+      "loss": 2.0158,
+      "step": 4424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008246827206275746,
+      "loss": 2.1975,
+      "step": 4425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008246001020586993,
+      "loss": 2.1342,
+      "step": 4426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008245174681679526,
+      "loss": 2.0768,
+      "step": 4427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000824434818959235,
+      "loss": 1.9896,
+      "step": 4428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008243521544364481,
+      "loss": 2.1174,
+      "step": 4429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008242694746034935,
+      "loss": 2.1164,
+      "step": 4430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008241867794642739,
+      "loss": 2.0347,
+      "step": 4431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008241040690226928,
+      "loss": 2.2627,
+      "step": 4432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008240213432826544,
+      "loss": 2.0274,
+      "step": 4433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008239386022480635,
+      "loss": 2.1158,
+      "step": 4434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008238558459228256,
+      "loss": 2.1853,
+      "step": 4435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000823773074310847,
+      "loss": 2.2001,
+      "step": 4436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008236902874160347,
+      "loss": 2.1045,
+      "step": 4437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008236074852422963,
+      "loss": 2.0934,
+      "step": 4438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008235246677935406,
+      "loss": 2.0746,
+      "step": 4439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008234418350736765,
+      "loss": 2.367,
+      "step": 4440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008233589870866141,
+      "loss": 2.163,
+      "step": 4441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008232761238362638,
+      "loss": 2.1736,
+      "step": 4442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008231932453265369,
+      "loss": 2.1131,
+      "step": 4443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008231103515613458,
+      "loss": 2.0873,
+      "step": 4444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008230274425446029,
+      "loss": 2.1809,
+      "step": 4445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000822944518280222,
+      "loss": 1.9784,
+      "step": 4446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000822861578772117,
+      "loss": 2.2497,
+      "step": 4447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000822778624024203,
+      "loss": 2.2694,
+      "step": 4448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008226956540403959,
+      "loss": 2.1173,
+      "step": 4449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008226126688246116,
+      "loss": 2.1782,
+      "step": 4450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008225296683807675,
+      "loss": 2.0411,
+      "step": 4451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008224466527127814,
+      "loss": 2.0012,
+      "step": 4452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008223636218245717,
+      "loss": 2.0558,
+      "step": 4453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008222805757200578,
+      "loss": 2.0547,
+      "step": 4454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008221975144031595,
+      "loss": 2.0686,
+      "step": 4455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008221144378777976,
+      "loss": 2.2004,
+      "step": 4456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008220313461478934,
+      "loss": 2.0445,
+      "step": 4457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008219482392173692,
+      "loss": 1.9471,
+      "step": 4458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008218651170901477,
+      "loss": 2.1932,
+      "step": 4459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008217819797701524,
+      "loss": 2.1987,
+      "step": 4460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008216988272613078,
+      "loss": 2.1496,
+      "step": 4461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008216156595675385,
+      "loss": 2.0522,
+      "step": 4462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008215324766927708,
+      "loss": 2.2334,
+      "step": 4463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008214492786409307,
+      "loss": 2.0312,
+      "step": 4464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008213660654159454,
+      "loss": 2.2355,
+      "step": 4465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008212828370217427,
+      "loss": 2.2475,
+      "step": 4466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008211995934622511,
+      "loss": 2.2081,
+      "step": 4467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008211163347414004,
+      "loss": 2.2583,
+      "step": 4468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00082103306086312,
+      "loss": 2.0317,
+      "step": 4469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008209497718313409,
+      "loss": 2.1551,
+      "step": 4470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008208664676499945,
+      "loss": 2.121,
+      "step": 4471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008207831483230129,
+      "loss": 2.1135,
+      "step": 4472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008206998138543291,
+      "loss": 2.1,
+      "step": 4473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008206164642478765,
+      "loss": 2.0516,
+      "step": 4474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008205330995075895,
+      "loss": 2.299,
+      "step": 4475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000820449719637403,
+      "loss": 2.1056,
+      "step": 4476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008203663246412529,
+      "loss": 2.2648,
+      "step": 4477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008202829145230756,
+      "loss": 2.187,
+      "step": 4478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008201994892868081,
+      "loss": 2.1084,
+      "step": 4479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008201160489363883,
+      "loss": 2.2169,
+      "step": 4480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008200325934757549,
+      "loss": 2.1502,
+      "step": 4481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008199491229088472,
+      "loss": 2.0392,
+      "step": 4482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000819865637239605,
+      "loss": 2.1612,
+      "step": 4483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000819782136471969,
+      "loss": 2.0413,
+      "step": 4484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008196986206098811,
+      "loss": 2.1457,
+      "step": 4485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008196150896572829,
+      "loss": 2.0996,
+      "step": 4486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008195315436181174,
+      "loss": 2.0748,
+      "step": 4487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008194479824963284,
+      "loss": 1.9935,
+      "step": 4488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008193644062958599,
+      "loss": 2.1608,
+      "step": 4489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000819280815020657,
+      "loss": 2.0358,
+      "step": 4490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008191972086746654,
+      "loss": 2.1216,
+      "step": 4491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008191135872618315,
+      "loss": 2.1273,
+      "step": 4492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008190299507861024,
+      "loss": 2.1892,
+      "step": 4493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000818946299251426,
+      "loss": 2.16,
+      "step": 4494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008188626326617507,
+      "loss": 2.0704,
+      "step": 4495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000818778951021026,
+      "loss": 2.1398,
+      "step": 4496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008186952543332017,
+      "loss": 2.1945,
+      "step": 4497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008186115426022285,
+      "loss": 2.0086,
+      "step": 4498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008185278158320577,
+      "loss": 2.1959,
+      "step": 4499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008184440740266415,
+      "loss": 2.1596,
+      "step": 4500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008183603171899327,
+      "loss": 2.1924,
+      "step": 4501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008182765453258849,
+      "loss": 2.0078,
+      "step": 4502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000818192758438452,
+      "loss": 2.0356,
+      "step": 4503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008181089565315894,
+      "loss": 2.0904,
+      "step": 4504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008180251396092524,
+      "loss": 2.1664,
+      "step": 4505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008179413076753975,
+      "loss": 2.0395,
+      "step": 4506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008178574607339817,
+      "loss": 2.186,
+      "step": 4507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008177735987889627,
+      "loss": 2.0313,
+      "step": 4508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008176897218442992,
+      "loss": 2.273,
+      "step": 4509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008176058299039503,
+      "loss": 2.143,
+      "step": 4510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008175219229718757,
+      "loss": 2.0858,
+      "step": 4511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008174380010520361,
+      "loss": 2.3383,
+      "step": 4512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008173540641483929,
+      "loss": 2.0707,
+      "step": 4513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008172701122649082,
+      "loss": 2.0761,
+      "step": 4514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008171861454055445,
+      "loss": 2.2407,
+      "step": 4515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008171021635742652,
+      "loss": 2.0669,
+      "step": 4516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008170181667750346,
+      "loss": 2.0724,
+      "step": 4517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008169341550118175,
+      "loss": 2.149,
+      "step": 4518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008168501282885795,
+      "loss": 2.1293,
+      "step": 4519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008167660866092867,
+      "loss": 2.1499,
+      "step": 4520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008166820299779062,
+      "loss": 2.1154,
+      "step": 4521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008165979583984055,
+      "loss": 2.0245,
+      "step": 4522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008165138718747533,
+      "loss": 2.0663,
+      "step": 4523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008164297704109184,
+      "loss": 2.0077,
+      "step": 4524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008163456540108706,
+      "loss": 2.1729,
+      "step": 4525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008162615226785805,
+      "loss": 2.1995,
+      "step": 4526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008161773764180191,
+      "loss": 2.1266,
+      "step": 4527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008160932152331586,
+      "loss": 2.1518,
+      "step": 4528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008160090391279713,
+      "loss": 2.1076,
+      "step": 4529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008159248481064308,
+      "loss": 2.1543,
+      "step": 4530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008158406421725109,
+      "loss": 2.1044,
+      "step": 4531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008157564213301863,
+      "loss": 2.0493,
+      "step": 4532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008156721855834323,
+      "loss": 2.0356,
+      "step": 4533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008155879349362254,
+      "loss": 2.0542,
+      "step": 4534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008155036693925422,
+      "loss": 2.1492,
+      "step": 4535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008154193889563603,
+      "loss": 2.2582,
+      "step": 4536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008153350936316578,
+      "loss": 1.7925,
+      "step": 4537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008152507834224137,
+      "loss": 2.2558,
+      "step": 4538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008151664583326077,
+      "loss": 2.3055,
+      "step": 4539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008150821183662201,
+      "loss": 1.9996,
+      "step": 4540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008149977635272319,
+      "loss": 1.9731,
+      "step": 4541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008149133938196249,
+      "loss": 2.1097,
+      "step": 4542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008148290092473816,
+      "loss": 2.2253,
+      "step": 4543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008147446098144849,
+      "loss": 1.9554,
+      "step": 4544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008146601955249188,
+      "loss": 2.185,
+      "step": 4545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0008145757663826678,
+      "loss": 2.2304,
+      "step": 4546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008144913223917175,
+      "loss": 2.0708,
+      "step": 4547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008144068635560533,
+      "loss": 2.2031,
+      "step": 4548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008143223898796621,
+      "loss": 2.2675,
+      "step": 4549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008142379013665313,
+      "loss": 2.0584,
+      "step": 4550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008141533980206491,
+      "loss": 2.2022,
+      "step": 4551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008140688798460038,
+      "loss": 2.2472,
+      "step": 4552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008139843468465853,
+      "loss": 2.2565,
+      "step": 4553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008138997990263835,
+      "loss": 2.2601,
+      "step": 4554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008138152363893893,
+      "loss": 2.2877,
+      "step": 4555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008137306589395942,
+      "loss": 2.0375,
+      "step": 4556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008136460666809906,
+      "loss": 2.1257,
+      "step": 4557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008135614596175713,
+      "loss": 2.0153,
+      "step": 4558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008134768377533302,
+      "loss": 2.1827,
+      "step": 4559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008133922010922614,
+      "loss": 2.0678,
+      "step": 4560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00081330754963836,
+      "loss": 2.1133,
+      "step": 4561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008132228833956217,
+      "loss": 2.0461,
+      "step": 4562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008131382023680429,
+      "loss": 2.158,
+      "step": 4563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000813053506559621,
+      "loss": 2.2127,
+      "step": 4564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008129687959743536,
+      "loss": 2.2989,
+      "step": 4565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008128840706162392,
+      "loss": 2.0827,
+      "step": 4566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008127993304892773,
+      "loss": 1.9947,
+      "step": 4567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008127145755974677,
+      "loss": 2.1192,
+      "step": 4568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008126298059448108,
+      "loss": 2.0311,
+      "step": 4569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008125450215353083,
+      "loss": 2.0244,
+      "step": 4570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008124602223729619,
+      "loss": 2.1941,
+      "step": 4571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008123754084617744,
+      "loss": 2.1642,
+      "step": 4572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008122905798057495,
+      "loss": 2.1271,
+      "step": 4573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008122057364088909,
+      "loss": 2.274,
+      "step": 4574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008121208782752036,
+      "loss": 2.1487,
+      "step": 4575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000812036005408693,
+      "loss": 2.1275,
+      "step": 4576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008119511178133655,
+      "loss": 2.0622,
+      "step": 4577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008118662154932277,
+      "loss": 2.1604,
+      "step": 4578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008117812984522874,
+      "loss": 2.1171,
+      "step": 4579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000811696366694553,
+      "loss": 2.2027,
+      "step": 4580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000811611420224033,
+      "loss": 2.2256,
+      "step": 4581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008115264590447376,
+      "loss": 2.2073,
+      "step": 4582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008114414831606768,
+      "loss": 2.1078,
+      "step": 4583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008113564925758618,
+      "loss": 1.9583,
+      "step": 4584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008112714872943043,
+      "loss": 2.0655,
+      "step": 4585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008111864673200169,
+      "loss": 2.11,
+      "step": 4586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008111014326570124,
+      "loss": 2.0915,
+      "step": 4587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008110163833093049,
+      "loss": 1.9474,
+      "step": 4588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000810931319280909,
+      "loss": 2.2369,
+      "step": 4589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008108462405758398,
+      "loss": 2.0961,
+      "step": 4590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008107611471981131,
+      "loss": 2.1466,
+      "step": 4591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008106760391517456,
+      "loss": 2.1129,
+      "step": 4592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008105909164407547,
+      "loss": 2.1321,
+      "step": 4593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008105057790691581,
+      "loss": 2.2563,
+      "step": 4594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008104206270409749,
+      "loss": 2.197,
+      "step": 4595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008103354603602241,
+      "loss": 2.1562,
+      "step": 4596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000810250279030926,
+      "loss": 2.0389,
+      "step": 4597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008101650830571011,
+      "loss": 2.2921,
+      "step": 4598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008100798724427711,
+      "loss": 2.225,
+      "step": 4599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000809994647191958,
+      "loss": 2.1634,
+      "step": 4600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008099094073086848,
+      "loss": 2.1283,
+      "step": 4601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008098241527969747,
+      "loss": 2.0138,
+      "step": 4602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008097388836608521,
+      "loss": 2.1901,
+      "step": 4603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008096535999043419,
+      "loss": 2.0844,
+      "step": 4604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008095683015314699,
+      "loss": 2.0371,
+      "step": 4605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000809482988546262,
+      "loss": 2.1278,
+      "step": 4606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008093976609527453,
+      "loss": 2.1882,
+      "step": 4607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008093123187549475,
+      "loss": 2.1676,
+      "step": 4608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008092269619568971,
+      "loss": 2.0871,
+      "step": 4609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008091415905626227,
+      "loss": 2.0527,
+      "step": 4610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008090562045761545,
+      "loss": 2.2032,
+      "step": 4611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008089708040015227,
+      "loss": 2.1653,
+      "step": 4612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008088853888427583,
+      "loss": 2.123,
+      "step": 4613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008087999591038932,
+      "loss": 2.2291,
+      "step": 4614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00080871451478896,
+      "loss": 2.0729,
+      "step": 4615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008086290559019917,
+      "loss": 2.0065,
+      "step": 4616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008085435824470223,
+      "loss": 2.2738,
+      "step": 4617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008084580944280862,
+      "loss": 2.1196,
+      "step": 4618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008083725918492187,
+      "loss": 2.1491,
+      "step": 4619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008082870747144557,
+      "loss": 2.131,
+      "step": 4620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008082015430278339,
+      "loss": 2.1521,
+      "step": 4621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008081159967933904,
+      "loss": 2.0976,
+      "step": 4622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008080304360151634,
+      "loss": 2.1544,
+      "step": 4623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008079448606971913,
+      "loss": 2.1555,
+      "step": 4624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008078592708435136,
+      "loss": 2.0987,
+      "step": 4625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008077736664581706,
+      "loss": 2.1246,
+      "step": 4626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008076880475452025,
+      "loss": 2.1502,
+      "step": 4627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008076024141086508,
+      "loss": 2.0623,
+      "step": 4628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000807516766152558,
+      "loss": 2.043,
+      "step": 4629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008074311036809668,
+      "loss": 2.0617,
+      "step": 4630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008073454266979203,
+      "loss": 2.1274,
+      "step": 4631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000807259735207463,
+      "loss": 2.2166,
+      "step": 4632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008071740292136396,
+      "loss": 2.1474,
+      "step": 4633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008070883087204954,
+      "loss": 2.1277,
+      "step": 4634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008070025737320771,
+      "loss": 2.0933,
+      "step": 4635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008069168242524311,
+      "loss": 2.1125,
+      "step": 4636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008068310602856055,
+      "loss": 2.161,
+      "step": 4637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000806745281835648,
+      "loss": 2.0185,
+      "step": 4638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008066594889066077,
+      "loss": 2.1443,
+      "step": 4639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008065736815025345,
+      "loss": 1.9453,
+      "step": 4640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008064878596274785,
+      "loss": 1.9506,
+      "step": 4641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008064020232854906,
+      "loss": 2.1306,
+      "step": 4642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008063161724806226,
+      "loss": 2.1478,
+      "step": 4643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008062303072169268,
+      "loss": 1.9559,
+      "step": 4644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008061444274984564,
+      "loss": 2.1106,
+      "step": 4645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008060585333292649,
+      "loss": 2.1697,
+      "step": 4646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008059726247134068,
+      "loss": 2.1368,
+      "step": 4647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008058867016549372,
+      "loss": 2.1144,
+      "step": 4648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008058007641579119,
+      "loss": 2.0475,
+      "step": 4649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008057148122263871,
+      "loss": 2.1784,
+      "step": 4650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008056288458644204,
+      "loss": 2.0963,
+      "step": 4651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000805542865076069,
+      "loss": 2.0727,
+      "step": 4652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008054568698653919,
+      "loss": 2.1092,
+      "step": 4653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008053708602364481,
+      "loss": 2.1173,
+      "step": 4654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008052848361932974,
+      "loss": 2.0378,
+      "step": 4655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008051987977400005,
+      "loss": 2.2673,
+      "step": 4656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008051127448806183,
+      "loss": 2.0613,
+      "step": 4657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000805026677619213,
+      "loss": 2.1533,
+      "step": 4658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000804940595959847,
+      "loss": 2.1294,
+      "step": 4659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008048544999065836,
+      "loss": 2.1269,
+      "step": 4660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008047683894634868,
+      "loss": 1.9889,
+      "step": 4661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008046822646346212,
+      "loss": 2.0882,
+      "step": 4662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008045961254240519,
+      "loss": 2.2862,
+      "step": 4663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008045099718358451,
+      "loss": 1.9858,
+      "step": 4664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008044238038740673,
+      "loss": 2.2739,
+      "step": 4665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000804337621542786,
+      "loss": 2.2003,
+      "step": 4666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000804251424846069,
+      "loss": 1.9955,
+      "step": 4667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000804165213787985,
+      "loss": 2.0588,
+      "step": 4668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008040789883726034,
+      "loss": 2.1266,
+      "step": 4669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008039927486039945,
+      "loss": 2.0685,
+      "step": 4670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008039064944862285,
+      "loss": 2.1745,
+      "step": 4671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008038202260233773,
+      "loss": 2.1111,
+      "step": 4672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008037339432195126,
+      "loss": 2.224,
+      "step": 4673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008036476460787075,
+      "loss": 1.9804,
+      "step": 4674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000803561334605035,
+      "loss": 2.234,
+      "step": 4675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008034750088025695,
+      "loss": 2.0654,
+      "step": 4676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008033886686753856,
+      "loss": 1.9928,
+      "step": 4677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008033023142275589,
+      "loss": 1.9248,
+      "step": 4678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008032159454631655,
+      "loss": 1.9349,
+      "step": 4679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000803129562386282,
+      "loss": 2.0988,
+      "step": 4680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008030431650009864,
+      "loss": 2.161,
+      "step": 4681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008029567533113562,
+      "loss": 2.2631,
+      "step": 4682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008028703273214708,
+      "loss": 2.0359,
+      "step": 4683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008027838870354093,
+      "loss": 2.0947,
+      "step": 4684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000802697432457252,
+      "loss": 2.2062,
+      "step": 4685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008026109635910799,
+      "loss": 2.144,
+      "step": 4686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008025244804409743,
+      "loss": 2.002,
+      "step": 4687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008024379830110176,
+      "loss": 2.1027,
+      "step": 4688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008023514713052925,
+      "loss": 2.1272,
+      "step": 4689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008022649453278827,
+      "loss": 2.0895,
+      "step": 4690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008021784050828725,
+      "loss": 2.1592,
+      "step": 4691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008020918505743464,
+      "loss": 2.1153,
+      "step": 4692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008020052818063903,
+      "loss": 2.1242,
+      "step": 4693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008019186987830905,
+      "loss": 2.1521,
+      "step": 4694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008018321015085338,
+      "loss": 2.1094,
+      "step": 4695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0008017454899868077,
+      "loss": 2.1495,
+      "step": 4696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008016588642220006,
+      "loss": 2.084,
+      "step": 4697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008015722242182015,
+      "loss": 2.2547,
+      "step": 4698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008014855699794998,
+      "loss": 2.0626,
+      "step": 4699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008013989015099858,
+      "loss": 1.9427,
+      "step": 4700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008013122188137507,
+      "loss": 2.0002,
+      "step": 4701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008012255218948858,
+      "loss": 2.146,
+      "step": 4702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008011388107574838,
+      "loss": 2.1299,
+      "step": 4703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008010520854056371,
+      "loss": 2.0423,
+      "step": 4704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008009653458434399,
+      "loss": 2.1193,
+      "step": 4705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000800878592074986,
+      "loss": 2.3024,
+      "step": 4706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008007918241043708,
+      "loss": 2.2348,
+      "step": 4707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008007050419356898,
+      "loss": 2.16,
+      "step": 4708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008006182455730392,
+      "loss": 2.096,
+      "step": 4709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008005314350205162,
+      "loss": 2.2421,
+      "step": 4710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008004446102822183,
+      "loss": 2.0756,
+      "step": 4711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008003577713622437,
+      "loss": 2.0066,
+      "step": 4712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008002709182646917,
+      "loss": 2.0642,
+      "step": 4713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008001840509936619,
+      "loss": 2.2201,
+      "step": 4714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008000971695532544,
+      "loss": 2.0387,
+      "step": 4715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0008000102739475706,
+      "loss": 2.0907,
+      "step": 4716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007999233641807118,
+      "loss": 2.0893,
+      "step": 4717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007998364402567805,
+      "loss": 2.0037,
+      "step": 4718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007997495021798797,
+      "loss": 2.03,
+      "step": 4719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000799662549954113,
+      "loss": 2.1276,
+      "step": 4720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000799575583583585,
+      "loss": 2.2202,
+      "step": 4721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007994886030724004,
+      "loss": 2.1626,
+      "step": 4722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000799401608424665,
+      "loss": 2.1218,
+      "step": 4723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007993145996444852,
+      "loss": 2.0201,
+      "step": 4724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007992275767359679,
+      "loss": 2.0155,
+      "step": 4725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000799140539703221,
+      "loss": 2.1225,
+      "step": 4726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007990534885503525,
+      "loss": 2.0961,
+      "step": 4727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007989664232814717,
+      "loss": 2.1107,
+      "step": 4728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007988793439006883,
+      "loss": 2.1959,
+      "step": 4729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007987922504121125,
+      "loss": 2.2338,
+      "step": 4730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007987051428198553,
+      "loss": 2.1968,
+      "step": 4731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007986180211280284,
+      "loss": 2.2627,
+      "step": 4732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007985308853407443,
+      "loss": 2.1752,
+      "step": 4733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007984437354621157,
+      "loss": 2.0831,
+      "step": 4734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007983565714962566,
+      "loss": 2.2459,
+      "step": 4735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007982693934472813,
+      "loss": 2.1067,
+      "step": 4736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007981822013193045,
+      "loss": 2.2073,
+      "step": 4737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007980949951164422,
+      "loss": 2.1388,
+      "step": 4738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007980077748428105,
+      "loss": 2.2003,
+      "step": 4739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007979205405025267,
+      "loss": 1.9711,
+      "step": 4740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007978332920997081,
+      "loss": 2.2193,
+      "step": 4741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007977460296384733,
+      "loss": 2.1989,
+      "step": 4742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007976587531229411,
+      "loss": 2.3291,
+      "step": 4743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007975714625572312,
+      "loss": 2.2243,
+      "step": 4744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007974841579454641,
+      "loss": 2.1071,
+      "step": 4745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007973968392917606,
+      "loss": 2.1307,
+      "step": 4746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007973095066002423,
+      "loss": 2.0905,
+      "step": 4747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007972221598750315,
+      "loss": 2.0518,
+      "step": 4748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007971347991202513,
+      "loss": 2.1438,
+      "step": 4749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007970474243400254,
+      "loss": 2.1805,
+      "step": 4750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007969600355384779,
+      "loss": 1.9565,
+      "step": 4751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007968726327197337,
+      "loss": 2.2517,
+      "step": 4752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007967852158879185,
+      "loss": 2.0597,
+      "step": 4753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007966977850471589,
+      "loss": 2.0256,
+      "step": 4754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007966103402015812,
+      "loss": 2.2839,
+      "step": 4755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007965228813553135,
+      "loss": 2.0372,
+      "step": 4756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007964354085124839,
+      "loss": 2.1053,
+      "step": 4757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007963479216772213,
+      "loss": 2.1217,
+      "step": 4758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007962604208536552,
+      "loss": 2.143,
+      "step": 4759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000796172906045916,
+      "loss": 1.8792,
+      "step": 4760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007960853772581345,
+      "loss": 2.18,
+      "step": 4761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007959978344944424,
+      "loss": 2.1663,
+      "step": 4762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007959102777589717,
+      "loss": 1.9608,
+      "step": 4763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007958227070558554,
+      "loss": 2.2249,
+      "step": 4764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007957351223892271,
+      "loss": 2.1466,
+      "step": 4765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007956475237632211,
+      "loss": 2.1399,
+      "step": 4766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007955599111819718,
+      "loss": 2.1504,
+      "step": 4767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007954722846496149,
+      "loss": 2.1027,
+      "step": 4768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000795384644170287,
+      "loss": 2.1651,
+      "step": 4769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007952969897481244,
+      "loss": 2.0484,
+      "step": 4770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007952093213872648,
+      "loss": 2.1136,
+      "step": 4771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007951216390918466,
+      "loss": 2.1678,
+      "step": 4772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000795033942866008,
+      "loss": 2.067,
+      "step": 4773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000794946232713889,
+      "loss": 1.9643,
+      "step": 4774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007948585086396296,
+      "loss": 2.2462,
+      "step": 4775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007947707706473703,
+      "loss": 2.1341,
+      "step": 4776
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007946830187412529,
+      "loss": 2.0482,
+      "step": 4777
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007945952529254195,
+      "loss": 2.1663,
+      "step": 4778
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007945074732040124,
+      "loss": 2.2866,
+      "step": 4779
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007944196795811754,
+      "loss": 2.0983,
+      "step": 4780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007943318720610524,
+      "loss": 2.0477,
+      "step": 4781
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007942440506477883,
+      "loss": 2.1479,
+      "step": 4782
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007941562153455282,
+      "loss": 2.2507,
+      "step": 4783
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007940683661584184,
+      "loss": 2.1129,
+      "step": 4784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007939805030906053,
+      "loss": 2.0399,
+      "step": 4785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007938926261462366,
+      "loss": 2.1579,
+      "step": 4786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007938047353294599,
+      "loss": 2.3134,
+      "step": 4787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007937168306444242,
+      "loss": 2.2325,
+      "step": 4788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007936289120952786,
+      "loss": 2.1916,
+      "step": 4789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000793540979686173,
+      "loss": 2.0174,
+      "step": 4790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007934530334212582,
+      "loss": 1.957,
+      "step": 4791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007933650733046853,
+      "loss": 2.1957,
+      "step": 4792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007932770993406064,
+      "loss": 2.0727,
+      "step": 4793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000793189111533174,
+      "loss": 2.0872,
+      "step": 4794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007931011098865412,
+      "loss": 2.1619,
+      "step": 4795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007930130944048619,
+      "loss": 2.1528,
+      "step": 4796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007929250650922908,
+      "loss": 2.024,
+      "step": 4797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007928370219529831,
+      "loss": 2.108,
+      "step": 4798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007927489649910944,
+      "loss": 2.2636,
+      "step": 4799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007926608942107813,
+      "loss": 2.2243,
+      "step": 4800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007925728096162011,
+      "loss": 2.1213,
+      "step": 4801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007924847112115113,
+      "loss": 2.0307,
+      "step": 4802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007923965990008706,
+      "loss": 2.2027,
+      "step": 4803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000792308472988438,
+      "loss": 2.0526,
+      "step": 4804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007922203331783733,
+      "loss": 2.0604,
+      "step": 4805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007921321795748369,
+      "loss": 2.0884,
+      "step": 4806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007920440121819898,
+      "loss": 2.0717,
+      "step": 4807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007919558310039936,
+      "loss": 2.0751,
+      "step": 4808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007918676360450107,
+      "loss": 2.1111,
+      "step": 4809
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007917794273092045,
+      "loss": 2.1311,
+      "step": 4810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007916912048007381,
+      "loss": 2.1446,
+      "step": 4811
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007916029685237761,
+      "loss": 2.0485,
+      "step": 4812
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007915147184824834,
+      "loss": 2.0991,
+      "step": 4813
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007914264546810256,
+      "loss": 1.8325,
+      "step": 4814
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007913381771235689,
+      "loss": 2.181,
+      "step": 4815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007912498858142803,
+      "loss": 2.1365,
+      "step": 4816
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007911615807573274,
+      "loss": 2.0215,
+      "step": 4817
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007910732619568782,
+      "loss": 2.091,
+      "step": 4818
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007909849294171017,
+      "loss": 2.0677,
+      "step": 4819
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007908965831421673,
+      "loss": 2.0149,
+      "step": 4820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007908082231362452,
+      "loss": 2.1098,
+      "step": 4821
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007907198494035064,
+      "loss": 2.1851,
+      "step": 4822
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007906314619481219,
+      "loss": 2.1695,
+      "step": 4823
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007905430607742642,
+      "loss": 2.1586,
+      "step": 4824
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007904546458861057,
+      "loss": 2.0879,
+      "step": 4825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007903662172878201,
+      "loss": 2.1961,
+      "step": 4826
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007902777749835814,
+      "loss": 2.1132,
+      "step": 4827
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007901893189775639,
+      "loss": 2.2399,
+      "step": 4828
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007901008492739434,
+      "loss": 2.1203,
+      "step": 4829
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007900123658768957,
+      "loss": 2.0267,
+      "step": 4830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007899238687905974,
+      "loss": 2.0602,
+      "step": 4831
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007898353580192257,
+      "loss": 2.1184,
+      "step": 4832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007897468335669585,
+      "loss": 2.1679,
+      "step": 4833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007896582954379745,
+      "loss": 2.1386,
+      "step": 4834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007895697436364528,
+      "loss": 2.2335,
+      "step": 4835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007894811781665734,
+      "loss": 2.1835,
+      "step": 4836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007893925990325166,
+      "loss": 2.263,
+      "step": 4837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007893040062384636,
+      "loss": 2.026,
+      "step": 4838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007892153997885962,
+      "loss": 2.0944,
+      "step": 4839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007891267796870969,
+      "loss": 2.1519,
+      "step": 4840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007890381459381488,
+      "loss": 2.1405,
+      "step": 4841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007889494985459354,
+      "loss": 2.2998,
+      "step": 4842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007888608375146411,
+      "loss": 1.9734,
+      "step": 4843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000788772162848451,
+      "loss": 2.1897,
+      "step": 4844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0007886834745515507,
+      "loss": 2.127,
+      "step": 4845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007885947726281266,
+      "loss": 2.0532,
+      "step": 4846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007885060570823657,
+      "loss": 2.0936,
+      "step": 4847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007884173279184552,
+      "loss": 2.1827,
+      "step": 4848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007883285851405837,
+      "loss": 1.9923,
+      "step": 4849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007882398287529398,
+      "loss": 2.0637,
+      "step": 4850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007881510587597132,
+      "loss": 2.1494,
+      "step": 4851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007880622751650939,
+      "loss": 2.1729,
+      "step": 4852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007879734779732729,
+      "loss": 2.1651,
+      "step": 4853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007878846671884413,
+      "loss": 2.1293,
+      "step": 4854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007877958428147916,
+      "loss": 2.0694,
+      "step": 4855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000787707004856516,
+      "loss": 1.9398,
+      "step": 4856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007876181533178084,
+      "loss": 2.0801,
+      "step": 4857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007875292882028624,
+      "loss": 2.1416,
+      "step": 4858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000787440409515873,
+      "loss": 2.118,
+      "step": 4859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007873515172610349,
+      "loss": 2.2463,
+      "step": 4860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007872626114425445,
+      "loss": 2.1836,
+      "step": 4861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007871736920645984,
+      "loss": 1.9873,
+      "step": 4862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007870847591313934,
+      "loss": 2.4055,
+      "step": 4863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007869958126471277,
+      "loss": 2.0119,
+      "step": 4864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007869068526159997,
+      "loss": 2.0808,
+      "step": 4865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007868178790422084,
+      "loss": 2.151,
+      "step": 4866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007867288919299536,
+      "loss": 1.9865,
+      "step": 4867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007866398912834356,
+      "loss": 2.2316,
+      "step": 4868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007865508771068557,
+      "loss": 2.025,
+      "step": 4869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007864618494044154,
+      "loss": 2.2202,
+      "step": 4870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007863728081803171,
+      "loss": 2.1093,
+      "step": 4871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007862837534387637,
+      "loss": 2.0622,
+      "step": 4872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007861946851839587,
+      "loss": 1.9866,
+      "step": 4873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007861056034201064,
+      "loss": 2.0309,
+      "step": 4874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007860165081514116,
+      "loss": 2.0678,
+      "step": 4875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00078592739938208,
+      "loss": 2.099,
+      "step": 4876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007858382771163175,
+      "loss": 2.184,
+      "step": 4877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000785749141358331,
+      "loss": 2.1806,
+      "step": 4878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007856599921123281,
+      "loss": 2.1982,
+      "step": 4879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007855708293825165,
+      "loss": 2.0998,
+      "step": 4880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000785481653173105,
+      "loss": 2.0855,
+      "step": 4881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000785392463488303,
+      "loss": 2.2843,
+      "step": 4882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007853032603323205,
+      "loss": 1.9639,
+      "step": 4883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007852140437093677,
+      "loss": 2.1724,
+      "step": 4884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007851248136236564,
+      "loss": 2.0066,
+      "step": 4885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007850355700793983,
+      "loss": 2.1016,
+      "step": 4886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007849463130808057,
+      "loss": 2.2359,
+      "step": 4887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007848570426320917,
+      "loss": 2.096,
+      "step": 4888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007847677587374703,
+      "loss": 2.1084,
+      "step": 4889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007846784614011557,
+      "loss": 1.9759,
+      "step": 4890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007845891506273633,
+      "loss": 2.0729,
+      "step": 4891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007844998264203084,
+      "loss": 2.1145,
+      "step": 4892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007844104887842074,
+      "loss": 2.1512,
+      "step": 4893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007843211377232773,
+      "loss": 2.1811,
+      "step": 4894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007842317732417356,
+      "loss": 2.1177,
+      "step": 4895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007841423953438007,
+      "loss": 2.1046,
+      "step": 4896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007840530040336913,
+      "loss": 2.209,
+      "step": 4897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007839635993156269,
+      "loss": 2.1408,
+      "step": 4898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007838741811938274,
+      "loss": 2.0715,
+      "step": 4899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007837847496725139,
+      "loss": 2.041,
+      "step": 4900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007836953047559074,
+      "loss": 2.068,
+      "step": 4901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007836058464482303,
+      "loss": 2.3064,
+      "step": 4902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007835163747537051,
+      "loss": 2.0534,
+      "step": 4903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007834268896765548,
+      "loss": 2.0374,
+      "step": 4904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007833373912210037,
+      "loss": 2.056,
+      "step": 4905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007832478793912761,
+      "loss": 2.069,
+      "step": 4906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007831583541915972,
+      "loss": 2.0986,
+      "step": 4907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007830688156261927,
+      "loss": 2.0822,
+      "step": 4908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007829792636992892,
+      "loss": 2.1003,
+      "step": 4909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007828896984151139,
+      "loss": 2.212,
+      "step": 4910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000782800119777894,
+      "loss": 2.1191,
+      "step": 4911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007827105277918582,
+      "loss": 2.0988,
+      "step": 4912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007826209224612355,
+      "loss": 2.1603,
+      "step": 4913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007825313037902552,
+      "loss": 2.1885,
+      "step": 4914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007824416717831475,
+      "loss": 2.1524,
+      "step": 4915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007823520264441435,
+      "loss": 2.0136,
+      "step": 4916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007822623677774746,
+      "loss": 2.0536,
+      "step": 4917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007821726957873727,
+      "loss": 2.2185,
+      "step": 4918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007820830104780707,
+      "loss": 2.0871,
+      "step": 4919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007819933118538019,
+      "loss": 2.1249,
+      "step": 4920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007819035999188004,
+      "loss": 2.1649,
+      "step": 4921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007818138746773005,
+      "loss": 2.0575,
+      "step": 4922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007817241361335378,
+      "loss": 2.1972,
+      "step": 4923
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000781634384291748,
+      "loss": 2.0695,
+      "step": 4924
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007815446191561674,
+      "loss": 2.0569,
+      "step": 4925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007814548407310335,
+      "loss": 2.02,
+      "step": 4926
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007813650490205839,
+      "loss": 2.0476,
+      "step": 4927
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007812752440290568,
+      "loss": 2.0467,
+      "step": 4928
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007811854257606915,
+      "loss": 2.2343,
+      "step": 4929
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007810955942197273,
+      "loss": 2.2444,
+      "step": 4930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007810057494104046,
+      "loss": 2.05,
+      "step": 4931
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007809158913369645,
+      "loss": 2.1609,
+      "step": 4932
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007808260200036481,
+      "loss": 2.1675,
+      "step": 4933
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007807361354146979,
+      "loss": 2.2792,
+      "step": 4934
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007806462375743565,
+      "loss": 2.2146,
+      "step": 4935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007805563264868672,
+      "loss": 2.157,
+      "step": 4936
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007804664021564742,
+      "loss": 1.9424,
+      "step": 4937
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000780376464587422,
+      "loss": 2.0595,
+      "step": 4938
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007802865137839559,
+      "loss": 2.0491,
+      "step": 4939
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007801965497503217,
+      "loss": 2.2298,
+      "step": 4940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007801065724907661,
+      "loss": 2.1653,
+      "step": 4941
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007800165820095362,
+      "loss": 2.1234,
+      "step": 4942
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007799265783108796,
+      "loss": 1.8776,
+      "step": 4943
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007798365613990446,
+      "loss": 2.0987,
+      "step": 4944
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007797465312782806,
+      "loss": 2.1733,
+      "step": 4945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007796564879528369,
+      "loss": 2.1983,
+      "step": 4946
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007795664314269637,
+      "loss": 1.9631,
+      "step": 4947
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007794763617049123,
+      "loss": 2.0145,
+      "step": 4948
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007793862787909338,
+      "loss": 2.1573,
+      "step": 4949
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007792961826892803,
+      "loss": 2.0372,
+      "step": 4950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000779206073404205,
+      "loss": 1.9956,
+      "step": 4951
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007791159509399608,
+      "loss": 1.9929,
+      "step": 4952
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007790258153008017,
+      "loss": 2.0988,
+      "step": 4953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007789356664909825,
+      "loss": 2.0586,
+      "step": 4954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007788455045147587,
+      "loss": 2.2591,
+      "step": 4955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007787553293763854,
+      "loss": 2.203,
+      "step": 4956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007786651410801198,
+      "loss": 2.1336,
+      "step": 4957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007785749396302187,
+      "loss": 2.1408,
+      "step": 4958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007784847250309396,
+      "loss": 2.1249,
+      "step": 4959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007783944972865411,
+      "loss": 2.1388,
+      "step": 4960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007783042564012823,
+      "loss": 2.0823,
+      "step": 4961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007782140023794225,
+      "loss": 2.2347,
+      "step": 4962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000778123735225222,
+      "loss": 2.2254,
+      "step": 4963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007780334549429415,
+      "loss": 2.0203,
+      "step": 4964
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007779431615368426,
+      "loss": 2.1322,
+      "step": 4965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007778528550111873,
+      "loss": 2.0526,
+      "step": 4966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007777625353702383,
+      "loss": 2.0741,
+      "step": 4967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007776722026182587,
+      "loss": 2.1114,
+      "step": 4968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007775818567595128,
+      "loss": 2.0208,
+      "step": 4969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007774914977982648,
+      "loss": 2.069,
+      "step": 4970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007774011257387801,
+      "loss": 2.1533,
+      "step": 4971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007773107405853241,
+      "loss": 1.9597,
+      "step": 4972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007772203423421636,
+      "loss": 2.1669,
+      "step": 4973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007771299310135654,
+      "loss": 2.069,
+      "step": 4974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007770395066037972,
+      "loss": 2.1042,
+      "step": 4975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000776949069117127,
+      "loss": 2.1925,
+      "step": 4976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007768586185578241,
+      "loss": 1.9192,
+      "step": 4977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007767681549301575,
+      "loss": 2.1013,
+      "step": 4978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007766776782383977,
+      "loss": 2.008,
+      "step": 4979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007765871884868149,
+      "loss": 2.2088,
+      "step": 4980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007764966856796811,
+      "loss": 2.3118,
+      "step": 4981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007764061698212676,
+      "loss": 2.0394,
+      "step": 4982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007763156409158475,
+      "loss": 2.1019,
+      "step": 4983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007762250989676936,
+      "loss": 2.0346,
+      "step": 4984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007761345439810798,
+      "loss": 2.1772,
+      "step": 4985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007760439759602805,
+      "loss": 2.0643,
+      "step": 4986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007759533949095708,
+      "loss": 2.2393,
+      "step": 4987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007758628008332261,
+      "loss": 2.0078,
+      "step": 4988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007757721937355228,
+      "loss": 2.0328,
+      "step": 4989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007756815736207379,
+      "loss": 2.152,
+      "step": 4990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007755909404931487,
+      "loss": 2.2402,
+      "step": 4991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007755002943570334,
+      "loss": 2.0865,
+      "step": 4992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007754096352166706,
+      "loss": 1.9925,
+      "step": 4993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0007753189630763397,
+      "loss": 2.0817,
+      "step": 4994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007752282779403205,
+      "loss": 2.04,
+      "step": 4995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007751375798128937,
+      "loss": 2.138,
+      "step": 4996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007750468686983404,
+      "loss": 2.1385,
+      "step": 4997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007749561446009424,
+      "loss": 1.8653,
+      "step": 4998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000774865407524982,
+      "loss": 2.1116,
+      "step": 4999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007747746574747423,
+      "loss": 1.9477,
+      "step": 5000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000774683894454507,
+      "loss": 2.0501,
+      "step": 5001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007745931184685601,
+      "loss": 2.1034,
+      "step": 5002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007745023295211865,
+      "loss": 2.0464,
+      "step": 5003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007744115276166716,
+      "loss": 2.0843,
+      "step": 5004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007743207127593017,
+      "loss": 2.2061,
+      "step": 5005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007742298849533633,
+      "loss": 2.2641,
+      "step": 5006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007741390442031436,
+      "loss": 1.9581,
+      "step": 5007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007740481905129307,
+      "loss": 2.0745,
+      "step": 5008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007739573238870129,
+      "loss": 2.1758,
+      "step": 5009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007738664443296795,
+      "loss": 2.0892,
+      "step": 5010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00077377555184522,
+      "loss": 2.1271,
+      "step": 5011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007736846464379249,
+      "loss": 2.137,
+      "step": 5012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007735937281120851,
+      "loss": 2.1047,
+      "step": 5013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007735027968719921,
+      "loss": 2.0692,
+      "step": 5014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007734118527219383,
+      "loss": 2.0321,
+      "step": 5015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007733208956662161,
+      "loss": 2.1428,
+      "step": 5016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007732299257091191,
+      "loss": 2.1806,
+      "step": 5017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007731389428549414,
+      "loss": 2.1863,
+      "step": 5018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007730479471079773,
+      "loss": 2.1076,
+      "step": 5019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007729569384725223,
+      "loss": 2.1797,
+      "step": 5020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000772865916952872,
+      "loss": 2.2268,
+      "step": 5021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007727748825533231,
+      "loss": 2.0694,
+      "step": 5022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007726838352781722,
+      "loss": 2.2521,
+      "step": 5023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007725927751317174,
+      "loss": 2.1207,
+      "step": 5024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007725017021182566,
+      "loss": 2.2523,
+      "step": 5025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007724106162420889,
+      "loss": 2.2418,
+      "step": 5026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007723195175075137,
+      "loss": 1.9026,
+      "step": 5027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007722284059188308,
+      "loss": 2.0316,
+      "step": 5028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007721372814803413,
+      "loss": 2.1977,
+      "step": 5029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007720461441963462,
+      "loss": 2.0021,
+      "step": 5030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007719549940711475,
+      "loss": 2.1406,
+      "step": 5031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007718638311090477,
+      "loss": 2.0014,
+      "step": 5032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007717726553143498,
+      "loss": 2.0936,
+      "step": 5033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007716814666913576,
+      "loss": 2.0786,
+      "step": 5034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007715902652443755,
+      "loss": 2.0561,
+      "step": 5035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007714990509777082,
+      "loss": 2.1183,
+      "step": 5036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007714078238956614,
+      "loss": 2.1651,
+      "step": 5037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007713165840025411,
+      "loss": 2.1039,
+      "step": 5038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007712253313026542,
+      "loss": 2.2209,
+      "step": 5039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007711340658003078,
+      "loss": 2.0966,
+      "step": 5040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00077104278749981,
+      "loss": 2.1817,
+      "step": 5041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007709514964054693,
+      "loss": 2.1296,
+      "step": 5042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000770860192521595,
+      "loss": 2.1928,
+      "step": 5043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007707688758524967,
+      "loss": 2.0724,
+      "step": 5044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007706775464024847,
+      "loss": 2.1596,
+      "step": 5045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007705862041758701,
+      "loss": 2.1356,
+      "step": 5046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007704948491769644,
+      "loss": 2.1009,
+      "step": 5047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007704034814100798,
+      "loss": 2.0298,
+      "step": 5048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007703121008795289,
+      "loss": 1.9877,
+      "step": 5049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007702207075896253,
+      "loss": 2.0917,
+      "step": 5050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007701293015446828,
+      "loss": 2.0747,
+      "step": 5051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007700378827490162,
+      "loss": 2.2648,
+      "step": 5052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007699464512069404,
+      "loss": 2.0167,
+      "step": 5053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007698550069227714,
+      "loss": 2.1266,
+      "step": 5054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007697635499008255,
+      "loss": 2.1129,
+      "step": 5055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007696720801454197,
+      "loss": 1.9857,
+      "step": 5056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007695805976608714,
+      "loss": 2.1383,
+      "step": 5057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007694891024514992,
+      "loss": 2.1837,
+      "step": 5058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007693975945216214,
+      "loss": 2.1078,
+      "step": 5059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007693060738755578,
+      "loss": 2.1045,
+      "step": 5060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000769214540517628,
+      "loss": 2.1156,
+      "step": 5061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007691229944521529,
+      "loss": 2.2018,
+      "step": 5062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007690314356834535,
+      "loss": 2.0369,
+      "step": 5063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007689398642158517,
+      "loss": 2.0449,
+      "step": 5064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007688482800536699,
+      "loss": 2.1708,
+      "step": 5065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007687566832012309,
+      "loss": 2.1312,
+      "step": 5066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007686650736628584,
+      "loss": 2.0987,
+      "step": 5067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007685734514428766,
+      "loss": 2.0511,
+      "step": 5068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007684818165456102,
+      "loss": 2.1273,
+      "step": 5069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000768390168975385,
+      "loss": 2.104,
+      "step": 5070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007682985087365262,
+      "loss": 2.2994,
+      "step": 5071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000768206835833361,
+      "loss": 2.1994,
+      "step": 5072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007681151502702162,
+      "loss": 1.9767,
+      "step": 5073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007680234520514199,
+      "loss": 2.1126,
+      "step": 5074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007679317411813002,
+      "loss": 2.1471,
+      "step": 5075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007678400176641862,
+      "loss": 2.1262,
+      "step": 5076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007677482815044073,
+      "loss": 2.0311,
+      "step": 5077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007676565327062939,
+      "loss": 2.0423,
+      "step": 5078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007675647712741767,
+      "loss": 2.2291,
+      "step": 5079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007674729972123868,
+      "loss": 2.1684,
+      "step": 5080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007673812105252564,
+      "loss": 2.0254,
+      "step": 5081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007672894112171181,
+      "loss": 2.1897,
+      "step": 5082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007671975992923048,
+      "loss": 2.1539,
+      "step": 5083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007671057747551504,
+      "loss": 2.1088,
+      "step": 5084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000767013937609989,
+      "loss": 2.1976,
+      "step": 5085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007669220878611558,
+      "loss": 2.3134,
+      "step": 5086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007668302255129864,
+      "loss": 2.1081,
+      "step": 5087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007667383505698165,
+      "loss": 1.9908,
+      "step": 5088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007666464630359831,
+      "loss": 2.0447,
+      "step": 5089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007665545629158235,
+      "loss": 2.1366,
+      "step": 5090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007664626502136755,
+      "loss": 2.2221,
+      "step": 5091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007663707249338777,
+      "loss": 2.0494,
+      "step": 5092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007662787870807691,
+      "loss": 2.091,
+      "step": 5093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007661868366586894,
+      "loss": 1.9789,
+      "step": 5094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007660948736719789,
+      "loss": 2.3201,
+      "step": 5095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007660028981249783,
+      "loss": 2.0886,
+      "step": 5096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007659109100220292,
+      "loss": 2.0095,
+      "step": 5097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007658189093674737,
+      "loss": 2.0121,
+      "step": 5098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007657268961656543,
+      "loss": 1.9642,
+      "step": 5099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007656348704209146,
+      "loss": 2.3446,
+      "step": 5100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007655428321375978,
+      "loss": 2.0942,
+      "step": 5101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007654507813200489,
+      "loss": 2.0583,
+      "step": 5102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007653587179726125,
+      "loss": 2.1093,
+      "step": 5103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007652666420996343,
+      "loss": 2.0867,
+      "step": 5104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007651745537054608,
+      "loss": 2.1634,
+      "step": 5105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007650824527944385,
+      "loss": 2.003,
+      "step": 5106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007649903393709147,
+      "loss": 2.1088,
+      "step": 5107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007648982134392377,
+      "loss": 2.0002,
+      "step": 5108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007648060750037556,
+      "loss": 2.0092,
+      "step": 5109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000764713924068818,
+      "loss": 2.1449,
+      "step": 5110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007646217606387743,
+      "loss": 2.0195,
+      "step": 5111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000764529584717975,
+      "loss": 1.9699,
+      "step": 5112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000764437396310771,
+      "loss": 2.1484,
+      "step": 5113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007643451954215138,
+      "loss": 2.0734,
+      "step": 5114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007642529820545553,
+      "loss": 2.2126,
+      "step": 5115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007641607562142485,
+      "loss": 2.1145,
+      "step": 5116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007640685179049465,
+      "loss": 2.0723,
+      "step": 5117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007639762671310031,
+      "loss": 2.0573,
+      "step": 5118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007638840038967729,
+      "loss": 1.8903,
+      "step": 5119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007637917282066109,
+      "loss": 2.1428,
+      "step": 5120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007636994400648727,
+      "loss": 2.2918,
+      "step": 5121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007636071394759146,
+      "loss": 2.0427,
+      "step": 5122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000763514826444093,
+      "loss": 2.0288,
+      "step": 5123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007634225009737659,
+      "loss": 2.0956,
+      "step": 5124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007633301630692907,
+      "loss": 2.1504,
+      "step": 5125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007632378127350264,
+      "loss": 2.1841,
+      "step": 5126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007631454499753319,
+      "loss": 2.181,
+      "step": 5127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007630530747945672,
+      "loss": 2.178,
+      "step": 5128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007629606871970922,
+      "loss": 2.1063,
+      "step": 5129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007628682871872681,
+      "loss": 1.9581,
+      "step": 5130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007627758747694565,
+      "loss": 2.0487,
+      "step": 5131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007626834499480191,
+      "loss": 2.2116,
+      "step": 5132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007625910127273188,
+      "loss": 2.1575,
+      "step": 5133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000762498563111719,
+      "loss": 2.2248,
+      "step": 5134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007624061011055832,
+      "loss": 2.0203,
+      "step": 5135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007623136267132759,
+      "loss": 2.0278,
+      "step": 5136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007622211399391624,
+      "loss": 1.9607,
+      "step": 5137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000762128640787608,
+      "loss": 2.0525,
+      "step": 5138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007620361292629789,
+      "loss": 1.9852,
+      "step": 5139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007619436053696421,
+      "loss": 2.0984,
+      "step": 5140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007618510691119644,
+      "loss": 2.0575,
+      "step": 5141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007617585204943143,
+      "loss": 2.0485,
+      "step": 5142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0007616659595210598,
+      "loss": 2.1256,
+      "step": 5143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007615733861965706,
+      "loss": 2.1306,
+      "step": 5144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007614808005252158,
+      "loss": 2.0501,
+      "step": 5145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007613882025113661,
+      "loss": 2.0359,
+      "step": 5146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000761295592159392,
+      "loss": 2.0737,
+      "step": 5147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007612029694736652,
+      "loss": 2.2385,
+      "step": 5148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007611103344585574,
+      "loss": 2.0633,
+      "step": 5149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007610176871184415,
+      "loss": 2.1331,
+      "step": 5150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007609250274576902,
+      "loss": 2.13,
+      "step": 5151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000760832355480678,
+      "loss": 1.9996,
+      "step": 5152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007607396711917784,
+      "loss": 2.1164,
+      "step": 5153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007606469745953669,
+      "loss": 2.0779,
+      "step": 5154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007605542656958187,
+      "loss": 2.1256,
+      "step": 5155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007604615444975102,
+      "loss": 1.9831,
+      "step": 5156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007603688110048174,
+      "loss": 2.1514,
+      "step": 5157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007602760652221183,
+      "loss": 2.1217,
+      "step": 5158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007601833071537904,
+      "loss": 2.2433,
+      "step": 5159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007600905368042119,
+      "loss": 2.1131,
+      "step": 5160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007599977541777621,
+      "loss": 1.9626,
+      "step": 5161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007599049592788203,
+      "loss": 2.0391,
+      "step": 5162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007598121521117666,
+      "loss": 2.0136,
+      "step": 5163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007597193326809819,
+      "loss": 2.0267,
+      "step": 5164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007596265009908476,
+      "loss": 2.1194,
+      "step": 5165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007595336570457451,
+      "loss": 2.0485,
+      "step": 5166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007594408008500574,
+      "loss": 2.1789,
+      "step": 5167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007593479324081671,
+      "loss": 2.0983,
+      "step": 5168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007592550517244582,
+      "loss": 2.2151,
+      "step": 5169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007591621588033144,
+      "loss": 2.012,
+      "step": 5170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007590692536491209,
+      "loss": 2.1842,
+      "step": 5171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007589763362662627,
+      "loss": 2.155,
+      "step": 5172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007588834066591261,
+      "loss": 2.2589,
+      "step": 5173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007587904648320971,
+      "loss": 2.1731,
+      "step": 5174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007586975107895633,
+      "loss": 2.1988,
+      "step": 5175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007586045445359118,
+      "loss": 2.0654,
+      "step": 5176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007585115660755313,
+      "loss": 2.1517,
+      "step": 5177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007584185754128102,
+      "loss": 2.0962,
+      "step": 5178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007583255725521383,
+      "loss": 1.963,
+      "step": 5179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000758232557497905,
+      "loss": 2.1044,
+      "step": 5180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007581395302545013,
+      "loss": 2.0431,
+      "step": 5181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007580464908263182,
+      "loss": 2.1296,
+      "step": 5182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007579534392177473,
+      "loss": 2.1959,
+      "step": 5183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007578603754331808,
+      "loss": 2.1267,
+      "step": 5184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007577672994770117,
+      "loss": 2.0669,
+      "step": 5185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000757674211353633,
+      "loss": 1.9506,
+      "step": 5186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007575811110674392,
+      "loss": 2.06,
+      "step": 5187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007574879986228245,
+      "loss": 2.2046,
+      "step": 5188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007573948740241843,
+      "loss": 2.0759,
+      "step": 5189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007573017372759139,
+      "loss": 2.0925,
+      "step": 5190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00075720858838241,
+      "loss": 1.9931,
+      "step": 5191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000757115427348069,
+      "loss": 2.0968,
+      "step": 5192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007570222541772887,
+      "loss": 2.1992,
+      "step": 5193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007569290688744669,
+      "loss": 2.3182,
+      "step": 5194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007568358714440022,
+      "loss": 2.0707,
+      "step": 5195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007567426618902937,
+      "loss": 1.9855,
+      "step": 5196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007566494402177411,
+      "loss": 2.0148,
+      "step": 5197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007565562064307447,
+      "loss": 2.1542,
+      "step": 5198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007564629605337057,
+      "loss": 2.0298,
+      "step": 5199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007563697025310248,
+      "loss": 1.9669,
+      "step": 5200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007562764324271045,
+      "loss": 2.3057,
+      "step": 5201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007561831502263472,
+      "loss": 2.2403,
+      "step": 5202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007560898559331561,
+      "loss": 2.0504,
+      "step": 5203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007559965495519349,
+      "loss": 2.1427,
+      "step": 5204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007559032310870879,
+      "loss": 2.0965,
+      "step": 5205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007558099005430197,
+      "loss": 2.0256,
+      "step": 5206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007557165579241361,
+      "loss": 2.1693,
+      "step": 5207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007556232032348428,
+      "loss": 2.1345,
+      "step": 5208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007555298364795466,
+      "loss": 2.2145,
+      "step": 5209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007554364576626543,
+      "loss": 1.8937,
+      "step": 5210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007553430667885739,
+      "loss": 2.1534,
+      "step": 5211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007552496638617135,
+      "loss": 2.1297,
+      "step": 5212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000755156248886482,
+      "loss": 2.0789,
+      "step": 5213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007550628218672887,
+      "loss": 2.3036,
+      "step": 5214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007549693828085437,
+      "loss": 2.1343,
+      "step": 5215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007548759317146575,
+      "loss": 2.1388,
+      "step": 5216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007547824685900413,
+      "loss": 2.2784,
+      "step": 5217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007546889934391065,
+      "loss": 2.3303,
+      "step": 5218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007545955062662655,
+      "loss": 2.1054,
+      "step": 5219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007545020070759312,
+      "loss": 2.2659,
+      "step": 5220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007544084958725168,
+      "loss": 2.188,
+      "step": 5221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007543149726604364,
+      "loss": 2.0835,
+      "step": 5222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007542214374441045,
+      "loss": 2.1831,
+      "step": 5223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000754127890227936,
+      "loss": 2.0593,
+      "step": 5224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007540343310163469,
+      "loss": 2.0251,
+      "step": 5225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000753940759813753,
+      "loss": 2.194,
+      "step": 5226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007538471766245714,
+      "loss": 2.1496,
+      "step": 5227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000753753581453219,
+      "loss": 1.971,
+      "step": 5228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007536599743041143,
+      "loss": 2.1322,
+      "step": 5229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007535663551816754,
+      "loss": 2.136,
+      "step": 5230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007534727240903214,
+      "loss": 2.0665,
+      "step": 5231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007533790810344719,
+      "loss": 2.068,
+      "step": 5232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007532854260185471,
+      "loss": 2.1232,
+      "step": 5233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007531917590469677,
+      "loss": 1.9775,
+      "step": 5234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007530980801241549,
+      "loss": 2.0548,
+      "step": 5235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007530043892545309,
+      "loss": 2.1741,
+      "step": 5236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007529106864425177,
+      "loss": 2.0457,
+      "step": 5237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007528169716925386,
+      "loss": 2.3843,
+      "step": 5238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007527232450090172,
+      "loss": 2.1315,
+      "step": 5239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007526295063963771,
+      "loss": 2.1244,
+      "step": 5240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007525357558590436,
+      "loss": 2.24,
+      "step": 5241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007524419934014415,
+      "loss": 2.2213,
+      "step": 5242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007523482190279969,
+      "loss": 2.0685,
+      "step": 5243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000752254432743136,
+      "loss": 2.2014,
+      "step": 5244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000752160634551286,
+      "loss": 2.1482,
+      "step": 5245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007520668244568738,
+      "loss": 2.101,
+      "step": 5246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007519730024643279,
+      "loss": 2.1688,
+      "step": 5247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007518791685780768,
+      "loss": 2.146,
+      "step": 5248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007517853228025499,
+      "loss": 2.0038,
+      "step": 5249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007516914651421767,
+      "loss": 2.199,
+      "step": 5250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007515975956013874,
+      "loss": 2.1668,
+      "step": 5251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007515037141846131,
+      "loss": 2.191,
+      "step": 5252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007514098208962851,
+      "loss": 2.1833,
+      "step": 5253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007513159157408354,
+      "loss": 2.1082,
+      "step": 5254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007512219987226965,
+      "loss": 2.1835,
+      "step": 5255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007511280698463015,
+      "loss": 2.0874,
+      "step": 5256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007510341291160841,
+      "loss": 1.9968,
+      "step": 5257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007509401765364785,
+      "loss": 1.8312,
+      "step": 5258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007508462121119195,
+      "loss": 2.0485,
+      "step": 5259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007507522358468424,
+      "loss": 1.9889,
+      "step": 5260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007506582477456831,
+      "loss": 2.0222,
+      "step": 5261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007505642478128783,
+      "loss": 2.1229,
+      "step": 5262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007504702360528646,
+      "loss": 2.0203,
+      "step": 5263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007503762124700797,
+      "loss": 2.1037,
+      "step": 5264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007502821770689618,
+      "loss": 1.9066,
+      "step": 5265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007501881298539498,
+      "loss": 2.1124,
+      "step": 5266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007500940708294826,
+      "loss": 2.1319,
+      "step": 5267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00075,
+      "loss": 2.0245,
+      "step": 5268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007499059173699426,
+      "loss": 2.1175,
+      "step": 5269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007498118229437512,
+      "loss": 2.1058,
+      "step": 5270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007497177167258672,
+      "loss": 2.0222,
+      "step": 5271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007496235987207329,
+      "loss": 1.8759,
+      "step": 5272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007495294689327904,
+      "loss": 2.139,
+      "step": 5273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007494353273664833,
+      "loss": 2.179,
+      "step": 5274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000749341174026255,
+      "loss": 2.151,
+      "step": 5275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007492470089165501,
+      "loss": 2.1104,
+      "step": 5276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007491528320418129,
+      "loss": 2.2071,
+      "step": 5277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007490586434064894,
+      "loss": 2.1615,
+      "step": 5278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007489644430150249,
+      "loss": 2.0519,
+      "step": 5279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007488702308718663,
+      "loss": 2.2646,
+      "step": 5280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007487760069814604,
+      "loss": 2.2033,
+      "step": 5281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000748681771348255,
+      "loss": 2.0934,
+      "step": 5282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000748587523976698,
+      "loss": 2.1435,
+      "step": 5283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007484932648712381,
+      "loss": 2.0788,
+      "step": 5284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007483989940363249,
+      "loss": 2.0936,
+      "step": 5285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007483047114764077,
+      "loss": 2.0885,
+      "step": 5286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007482104171959372,
+      "loss": 2.1479,
+      "step": 5287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007481161111993642,
+      "loss": 2.0294,
+      "step": 5288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007480217934911402,
+      "loss": 2.0157,
+      "step": 5289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007479274640757172,
+      "loss": 2.0143,
+      "step": 5290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007478331229575476,
+      "loss": 2.0702,
+      "step": 5291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0007477387701410849,
+      "loss": 2.0899,
+      "step": 5292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007476444056307823,
+      "loss": 2.0572,
+      "step": 5293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007475500294310944,
+      "loss": 2.1575,
+      "step": 5294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007474556415464757,
+      "loss": 2.2235,
+      "step": 5295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007473612419813818,
+      "loss": 2.055,
+      "step": 5296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007472668307402683,
+      "loss": 2.2036,
+      "step": 5297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007471724078275917,
+      "loss": 2.1954,
+      "step": 5298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007470779732478091,
+      "loss": 2.0327,
+      "step": 5299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000746983527005378,
+      "loss": 1.9278,
+      "step": 5300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007468890691047565,
+      "loss": 2.0941,
+      "step": 5301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007467945995504032,
+      "loss": 2.1651,
+      "step": 5302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007467001183467771,
+      "loss": 2.0835,
+      "step": 5303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007466056254983382,
+      "loss": 1.9143,
+      "step": 5304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007465111210095465,
+      "loss": 2.0971,
+      "step": 5305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007464166048848631,
+      "loss": 2.0991,
+      "step": 5306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007463220771287494,
+      "loss": 2.0128,
+      "step": 5307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000746227537745667,
+      "loss": 2.0026,
+      "step": 5308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007461329867400786,
+      "loss": 1.989,
+      "step": 5309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007460384241164473,
+      "loss": 2.0015,
+      "step": 5310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007459438498792365,
+      "loss": 2.3235,
+      "step": 5311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007458492640329106,
+      "loss": 2.1128,
+      "step": 5312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007457546665819337,
+      "loss": 2.1212,
+      "step": 5313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007456600575307717,
+      "loss": 2.0309,
+      "step": 5314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007455654368838898,
+      "loss": 2.1387,
+      "step": 5315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007454708046457548,
+      "loss": 2.0907,
+      "step": 5316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007453761608208332,
+      "loss": 2.1255,
+      "step": 5317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007452815054135925,
+      "loss": 2.0336,
+      "step": 5318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007451868384285007,
+      "loss": 2.2024,
+      "step": 5319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007450921598700263,
+      "loss": 2.1211,
+      "step": 5320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007449974697426383,
+      "loss": 2.0814,
+      "step": 5321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007449027680508063,
+      "loss": 1.9823,
+      "step": 5322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007448080547990004,
+      "loss": 2.0525,
+      "step": 5323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007447133299916915,
+      "loss": 2.2208,
+      "step": 5324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007446185936333505,
+      "loss": 2.2306,
+      "step": 5325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007445238457284495,
+      "loss": 2.2185,
+      "step": 5326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007444290862814604,
+      "loss": 2.0936,
+      "step": 5327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007443343152968566,
+      "loss": 2.0807,
+      "step": 5328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000744239532779111,
+      "loss": 2.0898,
+      "step": 5329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000744144738732698,
+      "loss": 2.1845,
+      "step": 5330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007440499331620918,
+      "loss": 2.1326,
+      "step": 5331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007439551160717675,
+      "loss": 2.0286,
+      "step": 5332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007438602874662006,
+      "loss": 1.9325,
+      "step": 5333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007437654473498675,
+      "loss": 1.9766,
+      "step": 5334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007436705957272446,
+      "loss": 2.1753,
+      "step": 5335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007435757326028093,
+      "loss": 2.2334,
+      "step": 5336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007434808579810391,
+      "loss": 2.0597,
+      "step": 5337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007433859718664127,
+      "loss": 1.9561,
+      "step": 5338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007432910742634086,
+      "loss": 2.251,
+      "step": 5339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007431961651765064,
+      "loss": 1.9764,
+      "step": 5340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007431012446101858,
+      "loss": 1.9694,
+      "step": 5341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007430063125689275,
+      "loss": 2.0577,
+      "step": 5342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007429113690572124,
+      "loss": 2.0479,
+      "step": 5343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007428164140795221,
+      "loss": 2.1696,
+      "step": 5344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007427214476403384,
+      "loss": 2.1075,
+      "step": 5345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007426264697441445,
+      "loss": 1.947,
+      "step": 5346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000742531480395423,
+      "loss": 2.0242,
+      "step": 5347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007424364795986583,
+      "loss": 2.0927,
+      "step": 5348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007423414673583339,
+      "loss": 2.0049,
+      "step": 5349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000742246443678935,
+      "loss": 2.182,
+      "step": 5350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007421514085649468,
+      "loss": 2.2099,
+      "step": 5351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007420563620208554,
+      "loss": 2.0502,
+      "step": 5352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000741961304051147,
+      "loss": 2.0102,
+      "step": 5353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007418662346603085,
+      "loss": 2.0086,
+      "step": 5354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007417711538528276,
+      "loss": 2.1316,
+      "step": 5355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007416760616331924,
+      "loss": 2.0819,
+      "step": 5356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007415809580058911,
+      "loss": 2.2083,
+      "step": 5357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007414858429754131,
+      "loss": 2.2351,
+      "step": 5358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007413907165462479,
+      "loss": 2.2387,
+      "step": 5359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007412955787228859,
+      "loss": 2.0815,
+      "step": 5360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007412004295098176,
+      "loss": 1.9715,
+      "step": 5361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007411052689115344,
+      "loss": 2.2353,
+      "step": 5362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007410100969325279,
+      "loss": 2.0577,
+      "step": 5363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007409149135772908,
+      "loss": 2.1421,
+      "step": 5364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007408197188503156,
+      "loss": 2.0205,
+      "step": 5365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007407245127560959,
+      "loss": 1.9803,
+      "step": 5366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007406292952991256,
+      "loss": 2.1536,
+      "step": 5367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007405340664838993,
+      "loss": 2.0635,
+      "step": 5368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007404388263149117,
+      "loss": 2.0255,
+      "step": 5369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007403435747966588,
+      "loss": 1.9326,
+      "step": 5370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007402483119336362,
+      "loss": 1.9614,
+      "step": 5371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000740153037730341,
+      "loss": 2.0263,
+      "step": 5372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007400577521912701,
+      "loss": 2.0376,
+      "step": 5373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007399624553209212,
+      "loss": 2.0752,
+      "step": 5374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007398671471237926,
+      "loss": 2.0113,
+      "step": 5375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007397718276043831,
+      "loss": 2.1265,
+      "step": 5376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007396764967671918,
+      "loss": 2.1831,
+      "step": 5377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007395811546167188,
+      "loss": 2.2888,
+      "step": 5378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007394858011574643,
+      "loss": 2.2064,
+      "step": 5379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007393904363939293,
+      "loss": 2.02,
+      "step": 5380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000739295060330615,
+      "loss": 2.1016,
+      "step": 5381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007391996729720238,
+      "loss": 2.0908,
+      "step": 5382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007391042743226577,
+      "loss": 2.1269,
+      "step": 5383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007390088643870202,
+      "loss": 2.1447,
+      "step": 5384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007389134431696145,
+      "loss": 2.1217,
+      "step": 5385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007388180106749449,
+      "loss": 2.0453,
+      "step": 5386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007387225669075161,
+      "loss": 2.2958,
+      "step": 5387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007386271118718329,
+      "loss": 2.0293,
+      "step": 5388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007385316455724016,
+      "loss": 2.0829,
+      "step": 5389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007384361680137278,
+      "loss": 2.2013,
+      "step": 5390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007383406792003188,
+      "loss": 2.1072,
+      "step": 5391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007382451791366814,
+      "loss": 2.0575,
+      "step": 5392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007381496678273238,
+      "loss": 2.0924,
+      "step": 5393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007380541452767542,
+      "loss": 2.0829,
+      "step": 5394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007379586114894815,
+      "loss": 2.1847,
+      "step": 5395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000737863066470015,
+      "loss": 2.1868,
+      "step": 5396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007377675102228649,
+      "loss": 2.0454,
+      "step": 5397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007376719427525415,
+      "loss": 2.099,
+      "step": 5398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007375763640635558,
+      "loss": 1.9172,
+      "step": 5399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007374807741604192,
+      "loss": 2.2859,
+      "step": 5400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007373851730476443,
+      "loss": 2.0496,
+      "step": 5401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000737289560729743,
+      "loss": 2.1199,
+      "step": 5402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000737193937211229,
+      "loss": 2.0452,
+      "step": 5403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007370983024966155,
+      "loss": 2.1106,
+      "step": 5404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000737002656590417,
+      "loss": 1.9773,
+      "step": 5405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000736906999497148,
+      "loss": 2.124,
+      "step": 5406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007368113312213238,
+      "loss": 2.1111,
+      "step": 5407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007367156517674603,
+      "loss": 2.1776,
+      "step": 5408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007366199611400736,
+      "loss": 2.0979,
+      "step": 5409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007365242593436805,
+      "loss": 2.1757,
+      "step": 5410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007364285463827984,
+      "loss": 2.1584,
+      "step": 5411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007363328222619453,
+      "loss": 2.0758,
+      "step": 5412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007362370869856397,
+      "loss": 1.9651,
+      "step": 5413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007361413405584,
+      "loss": 2.1334,
+      "step": 5414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007360455829847461,
+      "loss": 2.2138,
+      "step": 5415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007359498142691978,
+      "loss": 1.9363,
+      "step": 5416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007358540344162759,
+      "loss": 2.0538,
+      "step": 5417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007357582434305009,
+      "loss": 2.0612,
+      "step": 5418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007356624413163948,
+      "loss": 2.162,
+      "step": 5419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007355666280784794,
+      "loss": 2.0878,
+      "step": 5420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007354708037212776,
+      "loss": 2.0703,
+      "step": 5421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007353749682493121,
+      "loss": 2.1153,
+      "step": 5422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007352791216671071,
+      "loss": 2.0487,
+      "step": 5423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007351832639791863,
+      "loss": 2.0837,
+      "step": 5424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007350873951900747,
+      "loss": 2.072,
+      "step": 5425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007349915153042974,
+      "loss": 2.1257,
+      "step": 5426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007348956243263803,
+      "loss": 2.0259,
+      "step": 5427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007347997222608493,
+      "loss": 2.0477,
+      "step": 5428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007347038091122316,
+      "loss": 2.1352,
+      "step": 5429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007346078848850543,
+      "loss": 1.9877,
+      "step": 5430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007345119495838452,
+      "loss": 2.14,
+      "step": 5431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000734416003213133,
+      "loss": 2.121,
+      "step": 5432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007343200457774464,
+      "loss": 2.0816,
+      "step": 5433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007342240772813146,
+      "loss": 2.2759,
+      "step": 5434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007341280977292677,
+      "loss": 2.1025,
+      "step": 5435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007340321071258362,
+      "loss": 2.0758,
+      "step": 5436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007339361054755513,
+      "loss": 2.0476,
+      "step": 5437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007338400927829438,
+      "loss": 2.1384,
+      "step": 5438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007337440690525465,
+      "loss": 1.9742,
+      "step": 5439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0007336480342888914,
+      "loss": 2.2906,
+      "step": 5440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000733551988496512,
+      "loss": 2.0501,
+      "step": 5441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007334559316799415,
+      "loss": 2.0977,
+      "step": 5442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007333598638437143,
+      "loss": 2.0196,
+      "step": 5443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007332637849923647,
+      "loss": 2.2575,
+      "step": 5444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000733167695130428,
+      "loss": 2.1035,
+      "step": 5445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00073307159426244,
+      "loss": 1.9663,
+      "step": 5446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007329754823929368,
+      "loss": 2.182,
+      "step": 5447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007328793595264549,
+      "loss": 2.0977,
+      "step": 5448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007327832256675317,
+      "loss": 2.147,
+      "step": 5449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000732687080820705,
+      "loss": 2.1318,
+      "step": 5450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007325909249905131,
+      "loss": 2.3076,
+      "step": 5451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007324947581814945,
+      "loss": 2.2267,
+      "step": 5452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007323985803981886,
+      "loss": 2.1246,
+      "step": 5453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007323023916451354,
+      "loss": 2.0181,
+      "step": 5454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000732206191926875,
+      "loss": 2.147,
+      "step": 5455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007321099812479483,
+      "loss": 2.2069,
+      "step": 5456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007320137596128969,
+      "loss": 2.0527,
+      "step": 5457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007319175270262624,
+      "loss": 2.0648,
+      "step": 5458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007318212834925873,
+      "loss": 2.1845,
+      "step": 5459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007317250290164144,
+      "loss": 2.2679,
+      "step": 5460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007316287636022873,
+      "loss": 2.0366,
+      "step": 5461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007315324872547499,
+      "loss": 2.2395,
+      "step": 5462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007314361999783466,
+      "loss": 2.2608,
+      "step": 5463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007313399017776225,
+      "loss": 2.1436,
+      "step": 5464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000731243592657123,
+      "loss": 2.062,
+      "step": 5465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007311472726213941,
+      "loss": 2.0257,
+      "step": 5466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007310509416749826,
+      "loss": 2.089,
+      "step": 5467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000730954599822435,
+      "loss": 1.9657,
+      "step": 5468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007308582470682994,
+      "loss": 2.1745,
+      "step": 5469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007307618834171234,
+      "loss": 2.0796,
+      "step": 5470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007306655088734561,
+      "loss": 2.0974,
+      "step": 5471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007305691234418461,
+      "loss": 2.1437,
+      "step": 5472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007304727271268435,
+      "loss": 1.9362,
+      "step": 5473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007303763199329979,
+      "loss": 2.1206,
+      "step": 5474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007302799018648604,
+      "loss": 2.0858,
+      "step": 5475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007301834729269819,
+      "loss": 2.1447,
+      "step": 5476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000730087033123914,
+      "loss": 2.1041,
+      "step": 5477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007299905824602091,
+      "loss": 2.2301,
+      "step": 5478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007298941209404197,
+      "loss": 2.2442,
+      "step": 5479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007297976485690994,
+      "loss": 2.1532,
+      "step": 5480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007297011653508014,
+      "loss": 2.0448,
+      "step": 5481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007296046712900803,
+      "loss": 2.081,
+      "step": 5482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007295081663914906,
+      "loss": 2.0932,
+      "step": 5483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007294116506595877,
+      "loss": 1.9635,
+      "step": 5484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007293151240989272,
+      "loss": 2.1631,
+      "step": 5485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007292185867140656,
+      "loss": 2.1118,
+      "step": 5486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007291220385095596,
+      "loss": 2.036,
+      "step": 5487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007290254794899664,
+      "loss": 2.0812,
+      "step": 5488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000728928909659844,
+      "loss": 2.0453,
+      "step": 5489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007288323290237505,
+      "loss": 2.0092,
+      "step": 5490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007287357375862448,
+      "loss": 2.1932,
+      "step": 5491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007286391353518864,
+      "loss": 2.242,
+      "step": 5492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007285425223252349,
+      "loss": 2.151,
+      "step": 5493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007284458985108509,
+      "loss": 2.0549,
+      "step": 5494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000728349263913295,
+      "loss": 1.9347,
+      "step": 5495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007282526185371287,
+      "loss": 2.1217,
+      "step": 5496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007281559623869139,
+      "loss": 2.1302,
+      "step": 5497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007280592954672132,
+      "loss": 2.1371,
+      "step": 5498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000727962617782589,
+      "loss": 2.0053,
+      "step": 5499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007278659293376053,
+      "loss": 2.1234,
+      "step": 5500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007277692301368255,
+      "loss": 2.0023,
+      "step": 5501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007276725201848144,
+      "loss": 2.0775,
+      "step": 5502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007275757994861366,
+      "loss": 2.137,
+      "step": 5503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007274790680453579,
+      "loss": 2.1075,
+      "step": 5504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007273823258670439,
+      "loss": 2.1891,
+      "step": 5505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007272855729557613,
+      "loss": 2.2927,
+      "step": 5506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007271888093160769,
+      "loss": 2.23,
+      "step": 5507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007270920349525584,
+      "loss": 2.204,
+      "step": 5508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007269952498697733,
+      "loss": 2.1564,
+      "step": 5509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007268984540722907,
+      "loss": 2.1164,
+      "step": 5510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007268016475646793,
+      "loss": 2.1247,
+      "step": 5511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007267048303515084,
+      "loss": 2.1637,
+      "step": 5512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007266080024373482,
+      "loss": 2.1013,
+      "step": 5513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007265111638267692,
+      "loss": 1.9612,
+      "step": 5514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007264143145243422,
+      "loss": 2.1899,
+      "step": 5515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007263174545346391,
+      "loss": 2.142,
+      "step": 5516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007262205838622315,
+      "loss": 2.0646,
+      "step": 5517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007261237025116922,
+      "loss": 2.1052,
+      "step": 5518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007260268104875941,
+      "loss": 2.1186,
+      "step": 5519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007259299077945108,
+      "loss": 2.1171,
+      "step": 5520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007258329944370164,
+      "loss": 2.0882,
+      "step": 5521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007257360704196851,
+      "loss": 2.1617,
+      "step": 5522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007256391357470923,
+      "loss": 2.0269,
+      "step": 5523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007255421904238134,
+      "loss": 2.1937,
+      "step": 5524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007254452344544242,
+      "loss": 2.0508,
+      "step": 5525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007253482678435018,
+      "loss": 2.0687,
+      "step": 5526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007252512905956228,
+      "loss": 2.0607,
+      "step": 5527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007251543027153649,
+      "loss": 2.1232,
+      "step": 5528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007250573042073061,
+      "loss": 2.0782,
+      "step": 5529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000724960295076025,
+      "loss": 2.1492,
+      "step": 5530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007248632753261006,
+      "loss": 2.1099,
+      "step": 5531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007247662449621124,
+      "loss": 2.005,
+      "step": 5532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007246692039886407,
+      "loss": 2.0423,
+      "step": 5533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000724572152410266,
+      "loss": 2.2351,
+      "step": 5534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000724475090231569,
+      "loss": 1.9527,
+      "step": 5535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007243780174571317,
+      "loss": 2.0854,
+      "step": 5536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000724280934091536,
+      "loss": 2.0181,
+      "step": 5537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007241838401393643,
+      "loss": 2.2186,
+      "step": 5538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007240867356051999,
+      "loss": 2.039,
+      "step": 5539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007239896204936263,
+      "loss": 2.083,
+      "step": 5540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007238924948092274,
+      "loss": 1.9506,
+      "step": 5541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007237953585565879,
+      "loss": 2.0059,
+      "step": 5542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007236982117402928,
+      "loss": 2.059,
+      "step": 5543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007236010543649277,
+      "loss": 2.0676,
+      "step": 5544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007235038864350787,
+      "loss": 2.1023,
+      "step": 5545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007234067079553322,
+      "loss": 1.9876,
+      "step": 5546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007233095189302752,
+      "loss": 2.2067,
+      "step": 5547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007232123193644956,
+      "loss": 2.2359,
+      "step": 5548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007231151092625811,
+      "loss": 2.1988,
+      "step": 5549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007230178886291203,
+      "loss": 2.1987,
+      "step": 5550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007229206574687024,
+      "loss": 2.1366,
+      "step": 5551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007228234157859168,
+      "loss": 2.1554,
+      "step": 5552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007227261635853535,
+      "loss": 1.9179,
+      "step": 5553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000722628900871603,
+      "loss": 2.0835,
+      "step": 5554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007225316276492565,
+      "loss": 2.0458,
+      "step": 5555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007224343439229056,
+      "loss": 2.0938,
+      "step": 5556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007223370496971418,
+      "loss": 2.1404,
+      "step": 5557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007222397449765582,
+      "loss": 2.1418,
+      "step": 5558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007221424297657473,
+      "loss": 2.0535,
+      "step": 5559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007220451040693032,
+      "loss": 2.156,
+      "step": 5560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007219477678918194,
+      "loss": 1.985,
+      "step": 5561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007218504212378906,
+      "loss": 2.1716,
+      "step": 5562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007217530641121116,
+      "loss": 2.0007,
+      "step": 5563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007216556965190782,
+      "loss": 2.1278,
+      "step": 5564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007215583184633859,
+      "loss": 2.0815,
+      "step": 5565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007214609299496318,
+      "loss": 2.3198,
+      "step": 5566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007213635309824123,
+      "loss": 2.2657,
+      "step": 5567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007212661215663252,
+      "loss": 2.151,
+      "step": 5568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007211687017059682,
+      "loss": 2.074,
+      "step": 5569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007210712714059399,
+      "loss": 2.1677,
+      "step": 5570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007209738306708393,
+      "loss": 2.1451,
+      "step": 5571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007208763795052655,
+      "loss": 1.9225,
+      "step": 5572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007207789179138188,
+      "loss": 2.1267,
+      "step": 5573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007206814459010995,
+      "loss": 2.133,
+      "step": 5574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007205839634717085,
+      "loss": 2.1161,
+      "step": 5575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000720486470630247,
+      "loss": 2.1137,
+      "step": 5576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000720388967381317,
+      "loss": 2.0719,
+      "step": 5577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000720291453729521,
+      "loss": 2.1302,
+      "step": 5578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007201939296794617,
+      "loss": 2.0321,
+      "step": 5579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007200963952357428,
+      "loss": 2.1939,
+      "step": 5580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007199988504029677,
+      "loss": 2.0833,
+      "step": 5581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000719901295185741,
+      "loss": 1.9802,
+      "step": 5582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007198037295886673,
+      "loss": 2.0174,
+      "step": 5583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007197061536163524,
+      "loss": 2.1267,
+      "step": 5584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007196085672734015,
+      "loss": 2.1489,
+      "step": 5585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007195109705644214,
+      "loss": 1.8952,
+      "step": 5586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007194133634940185,
+      "loss": 2.1309,
+      "step": 5587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007193157460668005,
+      "loss": 2.0558,
+      "step": 5588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007192181182873747,
+      "loss": 2.0264,
+      "step": 5589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0007191204801603497,
+      "loss": 2.1593,
+      "step": 5590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007190228316903342,
+      "loss": 2.011,
+      "step": 5591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007189251728819373,
+      "loss": 2.1644,
+      "step": 5592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007188275037397687,
+      "loss": 2.0839,
+      "step": 5593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007187298242684388,
+      "loss": 1.977,
+      "step": 5594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007186321344725581,
+      "loss": 2.0551,
+      "step": 5595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007185344343567379,
+      "loss": 2.1696,
+      "step": 5596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007184367239255898,
+      "loss": 2.2194,
+      "step": 5597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007183390031837261,
+      "loss": 2.1847,
+      "step": 5598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007182412721357593,
+      "loss": 2.1609,
+      "step": 5599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007181435307863026,
+      "loss": 2.198,
+      "step": 5600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007180457791399695,
+      "loss": 2.0695,
+      "step": 5601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007179480172013743,
+      "loss": 2.1444,
+      "step": 5602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007178502449751315,
+      "loss": 2.0508,
+      "step": 5603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000717752462465856,
+      "loss": 2.0479,
+      "step": 5604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007176546696781635,
+      "loss": 2.1675,
+      "step": 5605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007175568666166701,
+      "loss": 2.0508,
+      "step": 5606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007174590532859924,
+      "loss": 2.1361,
+      "step": 5607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007173612296907472,
+      "loss": 2.0015,
+      "step": 5608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007172633958355522,
+      "loss": 2.1665,
+      "step": 5609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007171655517250251,
+      "loss": 2.1276,
+      "step": 5610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007170676973637847,
+      "loss": 1.9964,
+      "step": 5611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007169698327564498,
+      "loss": 2.0737,
+      "step": 5612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007168719579076398,
+      "loss": 2.0984,
+      "step": 5613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007167740728219746,
+      "loss": 2.2195,
+      "step": 5614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007166761775040747,
+      "loss": 2.2677,
+      "step": 5615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000716578271958561,
+      "loss": 2.1762,
+      "step": 5616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007164803561900547,
+      "loss": 2.1638,
+      "step": 5617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007163824302031778,
+      "loss": 2.1091,
+      "step": 5618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007162844940025526,
+      "loss": 2.0695,
+      "step": 5619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007161865475928018,
+      "loss": 2.165,
+      "step": 5620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007160885909785489,
+      "loss": 2.0949,
+      "step": 5621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007159906241644175,
+      "loss": 2.1213,
+      "step": 5622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007158926471550321,
+      "loss": 1.9736,
+      "step": 5623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007157946599550171,
+      "loss": 2.0445,
+      "step": 5624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000715696662568998,
+      "loss": 2.1646,
+      "step": 5625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007155986550016003,
+      "loss": 2.0987,
+      "step": 5626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007155006372574502,
+      "loss": 1.9312,
+      "step": 5627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007154026093411746,
+      "loss": 2.0914,
+      "step": 5628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007153045712574006,
+      "loss": 2.167,
+      "step": 5629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007152065230107554,
+      "loss": 2.05,
+      "step": 5630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007151084646058677,
+      "loss": 2.055,
+      "step": 5631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007150103960473656,
+      "loss": 1.9882,
+      "step": 5632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007149123173398784,
+      "loss": 2.0921,
+      "step": 5633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007148142284880356,
+      "loss": 2.2228,
+      "step": 5634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007147161294964673,
+      "loss": 2.1465,
+      "step": 5635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007146180203698036,
+      "loss": 2.1106,
+      "step": 5636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007145199011126762,
+      "loss": 2.1338,
+      "step": 5637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007144217717297157,
+      "loss": 2.0662,
+      "step": 5638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007143236322255549,
+      "loss": 2.0353,
+      "step": 5639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007142254826048255,
+      "loss": 2.0681,
+      "step": 5640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007141273228721609,
+      "loss": 2.0898,
+      "step": 5641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000714029153032194,
+      "loss": 2.2254,
+      "step": 5642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007139309730895591,
+      "loss": 1.9785,
+      "step": 5643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007138327830488901,
+      "loss": 2.1836,
+      "step": 5644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007137345829148221,
+      "loss": 2.2234,
+      "step": 5645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007136363726919904,
+      "loss": 2.0535,
+      "step": 5646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007135381523850304,
+      "loss": 2.0953,
+      "step": 5647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007134399219985786,
+      "loss": 2.1669,
+      "step": 5648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007133416815372718,
+      "loss": 2.0683,
+      "step": 5649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007132434310057467,
+      "loss": 2.1237,
+      "step": 5650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007131451704086415,
+      "loss": 2.1172,
+      "step": 5651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007130468997505939,
+      "loss": 2.0154,
+      "step": 5652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007129486190362429,
+      "loss": 2.1601,
+      "step": 5653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007128503282702272,
+      "loss": 2.2942,
+      "step": 5654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007127520274571867,
+      "loss": 2.1443,
+      "step": 5655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007126537166017612,
+      "loss": 2.1305,
+      "step": 5656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007125553957085911,
+      "loss": 2.0325,
+      "step": 5657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007124570647823176,
+      "loss": 2.0299,
+      "step": 5658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007123587238275821,
+      "loss": 1.9233,
+      "step": 5659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007122603728490264,
+      "loss": 2.2623,
+      "step": 5660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007121620118512932,
+      "loss": 2.0691,
+      "step": 5661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007120636408390248,
+      "loss": 2.1164,
+      "step": 5662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007119652598168651,
+      "loss": 1.9809,
+      "step": 5663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007118668687894576,
+      "loss": 2.1554,
+      "step": 5664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007117684677614468,
+      "loss": 2.208,
+      "step": 5665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007116700567374771,
+      "loss": 2.1451,
+      "step": 5666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007115716357221942,
+      "loss": 2.1417,
+      "step": 5667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007114732047202433,
+      "loss": 2.1942,
+      "step": 5668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007113747637362709,
+      "loss": 2.1781,
+      "step": 5669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007112763127749235,
+      "loss": 2.218,
+      "step": 5670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007111778518408485,
+      "loss": 2.0066,
+      "step": 5671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007110793809386931,
+      "loss": 2.1964,
+      "step": 5672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007109809000731055,
+      "loss": 2.0785,
+      "step": 5673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007108824092487342,
+      "loss": 2.0782,
+      "step": 5674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007107839084702283,
+      "loss": 2.1275,
+      "step": 5675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007106853977422372,
+      "loss": 2.1277,
+      "step": 5676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007105868770694108,
+      "loss": 1.9563,
+      "step": 5677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007104883464563995,
+      "loss": 2.0774,
+      "step": 5678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007103898059078542,
+      "loss": 2.0743,
+      "step": 5679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007102912554284262,
+      "loss": 2.2466,
+      "step": 5680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007101926950227674,
+      "loss": 1.8873,
+      "step": 5681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007100941246955299,
+      "loss": 2.1404,
+      "step": 5682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007099955444513667,
+      "loss": 2.0759,
+      "step": 5683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007098969542949308,
+      "loss": 2.0305,
+      "step": 5684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007097983542308762,
+      "loss": 2.0371,
+      "step": 5685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007096997442638564,
+      "loss": 2.1009,
+      "step": 5686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007096011243985268,
+      "loss": 2.0866,
+      "step": 5687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007095024946395419,
+      "loss": 2.1289,
+      "step": 5688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007094038549915574,
+      "loss": 2.1388,
+      "step": 5689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007093052054592295,
+      "loss": 2.2,
+      "step": 5690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007092065460472147,
+      "loss": 2.0229,
+      "step": 5691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007091078767601696,
+      "loss": 2.1338,
+      "step": 5692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000709009197602752,
+      "loss": 2.0251,
+      "step": 5693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007089105085796195,
+      "loss": 2.1094,
+      "step": 5694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007088118096954307,
+      "loss": 2.1537,
+      "step": 5695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007087131009548442,
+      "loss": 2.0844,
+      "step": 5696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007086143823625193,
+      "loss": 2.0571,
+      "step": 5697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007085156539231159,
+      "loss": 2.0337,
+      "step": 5698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000708416915641294,
+      "loss": 1.9223,
+      "step": 5699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007083181675217145,
+      "loss": 2.2752,
+      "step": 5700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007082194095690384,
+      "loss": 2.0263,
+      "step": 5701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007081206417879274,
+      "loss": 2.065,
+      "step": 5702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007080218641830434,
+      "loss": 2.0882,
+      "step": 5703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007079230767590492,
+      "loss": 2.1749,
+      "step": 5704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007078242795206073,
+      "loss": 2.1834,
+      "step": 5705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007077254724723818,
+      "loss": 2.1012,
+      "step": 5706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007076266556190361,
+      "loss": 1.9043,
+      "step": 5707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007075278289652349,
+      "loss": 2.0289,
+      "step": 5708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007074289925156429,
+      "loss": 2.1229,
+      "step": 5709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007073301462749254,
+      "loss": 2.0048,
+      "step": 5710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007072312902477483,
+      "loss": 2.0023,
+      "step": 5711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007071324244387776,
+      "loss": 2.0205,
+      "step": 5712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007070335488526801,
+      "loss": 2.0194,
+      "step": 5713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007069346634941231,
+      "loss": 2.1291,
+      "step": 5714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007068357683677741,
+      "loss": 2.0747,
+      "step": 5715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007067368634783011,
+      "loss": 2.1931,
+      "step": 5716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007066379488303728,
+      "loss": 2.0434,
+      "step": 5717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000706539024428658,
+      "loss": 2.0798,
+      "step": 5718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007064400902778264,
+      "loss": 2.0316,
+      "step": 5719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007063411463825478,
+      "loss": 2.0568,
+      "step": 5720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007062421927474925,
+      "loss": 2.0168,
+      "step": 5721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007061432293773315,
+      "loss": 1.9805,
+      "step": 5722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007060442562767358,
+      "loss": 2.1112,
+      "step": 5723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007059452734503776,
+      "loss": 2.1557,
+      "step": 5724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007058462809029288,
+      "loss": 2.0635,
+      "step": 5725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007057472786390622,
+      "loss": 2.124,
+      "step": 5726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007056482666634509,
+      "loss": 2.0521,
+      "step": 5727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007055492449807683,
+      "loss": 2.1778,
+      "step": 5728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007054502135956888,
+      "loss": 1.9675,
+      "step": 5729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007053511725128868,
+      "loss": 2.0813,
+      "step": 5730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007052521217370371,
+      "loss": 2.2281,
+      "step": 5731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007051530612728154,
+      "loss": 2.1402,
+      "step": 5732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007050539911248972,
+      "loss": 2.1966,
+      "step": 5733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007049549112979593,
+      "loss": 2.1816,
+      "step": 5734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000704855821796678,
+      "loss": 2.0987,
+      "step": 5735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007047567226257311,
+      "loss": 2.1725,
+      "step": 5736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007046576137897957,
+      "loss": 1.939,
+      "step": 5737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007045584952935506,
+      "loss": 2.1242,
+      "step": 5738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0007044593671416739,
+      "loss": 2.0899,
+      "step": 5739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000704360229338845,
+      "loss": 2.1218,
+      "step": 5740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007042610818897432,
+      "loss": 2.1508,
+      "step": 5741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007041619247990488,
+      "loss": 2.1533,
+      "step": 5742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007040627580714419,
+      "loss": 1.8524,
+      "step": 5743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007039635817116034,
+      "loss": 2.2026,
+      "step": 5744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000703864395724215,
+      "loss": 1.9534,
+      "step": 5745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007037652001139582,
+      "loss": 1.9929,
+      "step": 5746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007036659948855155,
+      "loss": 2.1959,
+      "step": 5747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007035667800435693,
+      "loss": 2.1226,
+      "step": 5748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007034675555928029,
+      "loss": 2.1678,
+      "step": 5749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007033683215379002,
+      "loss": 2.0785,
+      "step": 5750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007032690778835448,
+      "loss": 2.1385,
+      "step": 5751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007031698246344217,
+      "loss": 2.088,
+      "step": 5752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007030705617952156,
+      "loss": 2.0879,
+      "step": 5753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007029712893706119,
+      "loss": 2.184,
+      "step": 5754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007028720073652965,
+      "loss": 2.0247,
+      "step": 5755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000702772715783956,
+      "loss": 2.2189,
+      "step": 5756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000702673414631277,
+      "loss": 2.198,
+      "step": 5757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007025741039119465,
+      "loss": 2.0932,
+      "step": 5758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007024747836306527,
+      "loss": 1.9853,
+      "step": 5759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007023754537920834,
+      "loss": 2.1334,
+      "step": 5760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007022761144009271,
+      "loss": 2.1365,
+      "step": 5761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007021767654618734,
+      "loss": 2.0979,
+      "step": 5762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007020774069796112,
+      "loss": 2.1339,
+      "step": 5763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007019780389588309,
+      "loss": 2.0357,
+      "step": 5764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007018786614042223,
+      "loss": 2.2459,
+      "step": 5765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000701779274320477,
+      "loss": 2.1185,
+      "step": 5766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007016798777122857,
+      "loss": 2.1797,
+      "step": 5767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007015804715843403,
+      "loss": 2.1065,
+      "step": 5768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007014810559413333,
+      "loss": 2.2026,
+      "step": 5769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007013816307879571,
+      "loss": 2.0653,
+      "step": 5770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007012821961289047,
+      "loss": 2.1239,
+      "step": 5771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007011827519688698,
+      "loss": 2.0921,
+      "step": 5772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007010832983125463,
+      "loss": 2.2162,
+      "step": 5773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007009838351646288,
+      "loss": 2.2533,
+      "step": 5774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000700884362529812,
+      "loss": 2.1612,
+      "step": 5775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007007848804127915,
+      "loss": 2.1541,
+      "step": 5776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007006853888182626,
+      "loss": 1.9562,
+      "step": 5777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000700585887750922,
+      "loss": 1.929,
+      "step": 5778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007004863772154662,
+      "loss": 2.1179,
+      "step": 5779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007003868572165923,
+      "loss": 2.1197,
+      "step": 5780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000700287327758998,
+      "loss": 2.0496,
+      "step": 5781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007001877888473812,
+      "loss": 1.9493,
+      "step": 5782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0007000882404864404,
+      "loss": 2.1741,
+      "step": 5783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006999886826808745,
+      "loss": 2.009,
+      "step": 5784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006998891154353828,
+      "loss": 2.1215,
+      "step": 5785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006997895387546654,
+      "loss": 2.1172,
+      "step": 5786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000699689952643422,
+      "loss": 2.031,
+      "step": 5787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000699590357106354,
+      "loss": 2.1257,
+      "step": 5788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006994907521481618,
+      "loss": 2.1592,
+      "step": 5789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006993911377735476,
+      "loss": 1.9502,
+      "step": 5790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006992915139872129,
+      "loss": 2.1314,
+      "step": 5791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006991918807938606,
+      "loss": 2.1356,
+      "step": 5792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006990922381981935,
+      "loss": 2.1143,
+      "step": 5793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006989925862049147,
+      "loss": 2.1226,
+      "step": 5794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006988929248187283,
+      "loss": 1.9231,
+      "step": 5795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006987932540443387,
+      "loss": 2.077,
+      "step": 5796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006986935738864501,
+      "loss": 2.1622,
+      "step": 5797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006985938843497679,
+      "loss": 2.1057,
+      "step": 5798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006984941854389976,
+      "loss": 2.0589,
+      "step": 5799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006983944771588455,
+      "loss": 2.1711,
+      "step": 5800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006982947595140177,
+      "loss": 2.1988,
+      "step": 5801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006981950325092212,
+      "loss": 2.0648,
+      "step": 5802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006980952961491635,
+      "loss": 2.1411,
+      "step": 5803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006979955504385522,
+      "loss": 1.9315,
+      "step": 5804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006978957953820955,
+      "loss": 2.1649,
+      "step": 5805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006977960309845023,
+      "loss": 2.064,
+      "step": 5806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006976962572504815,
+      "loss": 2.1332,
+      "step": 5807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006975964741847427,
+      "loss": 2.0685,
+      "step": 5808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006974966817919959,
+      "loss": 2.1198,
+      "step": 5809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006973968800769517,
+      "loss": 2.1025,
+      "step": 5810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006972970690443207,
+      "loss": 2.0615,
+      "step": 5811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006971972486988142,
+      "loss": 2.1633,
+      "step": 5812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006970974190451442,
+      "loss": 2.0141,
+      "step": 5813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000696997580088023,
+      "loss": 1.9567,
+      "step": 5814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006968977318321625,
+      "loss": 2.1137,
+      "step": 5815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006967978742822767,
+      "loss": 1.986,
+      "step": 5816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006966980074430784,
+      "loss": 2.1013,
+      "step": 5817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006965981313192819,
+      "loss": 2.0633,
+      "step": 5818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006964982459156014,
+      "loss": 1.9969,
+      "step": 5819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000696398351236752,
+      "loss": 1.9612,
+      "step": 5820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006962984472874485,
+      "loss": 2.0801,
+      "step": 5821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000696198534072407,
+      "loss": 2.2096,
+      "step": 5822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006960986115963436,
+      "loss": 2.1791,
+      "step": 5823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006959986798639748,
+      "loss": 2.2297,
+      "step": 5824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006958987388800173,
+      "loss": 2.0897,
+      "step": 5825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000695798788649189,
+      "loss": 2.189,
+      "step": 5826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006956988291762076,
+      "loss": 1.9052,
+      "step": 5827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006955988604657914,
+      "loss": 2.1216,
+      "step": 5828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006954988825226592,
+      "loss": 2.1021,
+      "step": 5829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006953988953515301,
+      "loss": 2.0255,
+      "step": 5830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006952988989571238,
+      "loss": 2.1594,
+      "step": 5831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006951988933441604,
+      "loss": 2.2053,
+      "step": 5832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006950988785173604,
+      "loss": 2.1986,
+      "step": 5833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006949988544814445,
+      "loss": 2.0546,
+      "step": 5834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006948988212411343,
+      "loss": 2.0464,
+      "step": 5835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006947987788011517,
+      "loss": 2.112,
+      "step": 5836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006946987271662185,
+      "loss": 2.076,
+      "step": 5837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006945986663410577,
+      "loss": 2.3363,
+      "step": 5838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006944985963303925,
+      "loss": 2.0527,
+      "step": 5839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006943985171389461,
+      "loss": 2.1284,
+      "step": 5840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006942984287714428,
+      "loss": 2.2745,
+      "step": 5841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006941983312326067,
+      "loss": 2.1515,
+      "step": 5842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006940982245271628,
+      "loss": 2.0996,
+      "step": 5843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006939981086598364,
+      "loss": 2.1789,
+      "step": 5844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000693897983635353,
+      "loss": 2.1203,
+      "step": 5845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006937978494584389,
+      "loss": 2.0409,
+      "step": 5846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006936977061338208,
+      "loss": 2.048,
+      "step": 5847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006935975536662254,
+      "loss": 2.1943,
+      "step": 5848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006934973920603803,
+      "loss": 2.1049,
+      "step": 5849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006933972213210133,
+      "loss": 2.0672,
+      "step": 5850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006932970414528527,
+      "loss": 2.144,
+      "step": 5851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006931968524606271,
+      "loss": 2.2134,
+      "step": 5852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000693096654349066,
+      "loss": 2.0778,
+      "step": 5853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006929964471228987,
+      "loss": 2.2048,
+      "step": 5854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006928962307868551,
+      "loss": 2.2216,
+      "step": 5855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006927960053456659,
+      "loss": 2.0506,
+      "step": 5856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000692695770804062,
+      "loss": 2.1511,
+      "step": 5857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006925955271667744,
+      "loss": 2.0394,
+      "step": 5858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006924952744385353,
+      "loss": 2.1367,
+      "step": 5859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006923950126240762,
+      "loss": 2.0117,
+      "step": 5860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006922947417281302,
+      "loss": 2.1226,
+      "step": 5861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006921944617554303,
+      "loss": 2.1172,
+      "step": 5862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006920941727107098,
+      "loss": 1.9956,
+      "step": 5863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006919938745987024,
+      "loss": 2.1441,
+      "step": 5864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006918935674241428,
+      "loss": 2.1799,
+      "step": 5865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006917932511917653,
+      "loss": 2.1316,
+      "step": 5866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006916929259063056,
+      "loss": 2.0988,
+      "step": 5867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006915925915724987,
+      "loss": 2.1695,
+      "step": 5868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000691492248195081,
+      "loss": 2.1818,
+      "step": 5869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006913918957787887,
+      "loss": 2.1136,
+      "step": 5870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006912915343283589,
+      "loss": 2.108,
+      "step": 5871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006911911638485287,
+      "loss": 1.9666,
+      "step": 5872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006910907843440359,
+      "loss": 2.0049,
+      "step": 5873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006909903958196187,
+      "loss": 2.1097,
+      "step": 5874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006908899982800155,
+      "loss": 2.2053,
+      "step": 5875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006907895917299655,
+      "loss": 2.0812,
+      "step": 5876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000690689176174208,
+      "loss": 2.1213,
+      "step": 5877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006905887516174827,
+      "loss": 1.9941,
+      "step": 5878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006904883180645303,
+      "loss": 2.2439,
+      "step": 5879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000690387875520091,
+      "loss": 2.098,
+      "step": 5880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006902874239889063,
+      "loss": 2.1507,
+      "step": 5881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006901869634757174,
+      "loss": 2.2645,
+      "step": 5882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006900864939852667,
+      "loss": 2.0379,
+      "step": 5883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006899860155222961,
+      "loss": 2.0477,
+      "step": 5884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006898855280915489,
+      "loss": 2.1269,
+      "step": 5885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006897850316977679,
+      "loss": 2.1438,
+      "step": 5886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006896845263456971,
+      "loss": 2.0661,
+      "step": 5887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006895840120400803,
+      "loss": 2.0109,
+      "step": 5888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006894834887856624,
+      "loss": 2.1257,
+      "step": 5889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006893829565871878,
+      "loss": 1.9698,
+      "step": 5890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006892824154494024,
+      "loss": 2.0025,
+      "step": 5891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006891818653770515,
+      "loss": 2.1728,
+      "step": 5892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006890813063748815,
+      "loss": 2.205,
+      "step": 5893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000688980738447639,
+      "loss": 2.1958,
+      "step": 5894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006888801616000713,
+      "loss": 2.1025,
+      "step": 5895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006887795758369253,
+      "loss": 2.0201,
+      "step": 5896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006886789811629495,
+      "loss": 2.0449,
+      "step": 5897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006885783775828915,
+      "loss": 2.0322,
+      "step": 5898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006884777651015008,
+      "loss": 2.1983,
+      "step": 5899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006883771437235259,
+      "loss": 1.9272,
+      "step": 5900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006882765134537166,
+      "loss": 2.2098,
+      "step": 5901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006881758742968229,
+      "loss": 2.096,
+      "step": 5902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006880752262575955,
+      "loss": 2.2012,
+      "step": 5903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006879745693407845,
+      "loss": 2.1807,
+      "step": 5904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006878739035511418,
+      "loss": 2.106,
+      "step": 5905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006877732288934187,
+      "loss": 2.0606,
+      "step": 5906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006876725453723673,
+      "loss": 2.0375,
+      "step": 5907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006875718529927403,
+      "loss": 1.941,
+      "step": 5908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006874711517592905,
+      "loss": 2.0894,
+      "step": 5909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006873704416767712,
+      "loss": 2.105,
+      "step": 5910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006872697227499362,
+      "loss": 1.9829,
+      "step": 5911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006871689949835395,
+      "loss": 2.1544,
+      "step": 5912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006870682583823359,
+      "loss": 2.1507,
+      "step": 5913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006869675129510802,
+      "loss": 1.9266,
+      "step": 5914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000686866758694528,
+      "loss": 2.3115,
+      "step": 5915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006867659956174351,
+      "loss": 2.1479,
+      "step": 5916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006866652237245576,
+      "loss": 1.9029,
+      "step": 5917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006865644430206524,
+      "loss": 2.1826,
+      "step": 5918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006864636535104762,
+      "loss": 2.0187,
+      "step": 5919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000686362855198787,
+      "loss": 2.2053,
+      "step": 5920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006862620480903425,
+      "loss": 2.2126,
+      "step": 5921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006861612321899009,
+      "loss": 2.2247,
+      "step": 5922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006860604075022209,
+      "loss": 2.1204,
+      "step": 5923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000685959574032062,
+      "loss": 2.1891,
+      "step": 5924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006858587317841833,
+      "loss": 2.1516,
+      "step": 5925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006857578807633453,
+      "loss": 2.086,
+      "step": 5926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000685657020974308,
+      "loss": 2.0241,
+      "step": 5927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006855561524218326,
+      "loss": 2.0177,
+      "step": 5928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006854552751106798,
+      "loss": 2.1396,
+      "step": 5929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006853543890456117,
+      "loss": 2.1052,
+      "step": 5930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00068525349423139,
+      "loss": 2.1676,
+      "step": 5931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006851525906727776,
+      "loss": 2.0715,
+      "step": 5932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006850516783745369,
+      "loss": 2.0665,
+      "step": 5933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006849507573414316,
+      "loss": 1.9155,
+      "step": 5934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006848498275782251,
+      "loss": 1.9431,
+      "step": 5935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006847488890896816,
+      "loss": 2.0174,
+      "step": 5936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006846479418805658,
+      "loss": 2.1813,
+      "step": 5937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006845469859556426,
+      "loss": 2.0228,
+      "step": 5938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000684446021319677,
+      "loss": 2.0812,
+      "step": 5939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006843450479774352,
+      "loss": 2.122,
+      "step": 5940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006842440659336832,
+      "loss": 2.1853,
+      "step": 5941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006841430751931875,
+      "loss": 2.042,
+      "step": 5942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006840420757607153,
+      "loss": 2.0397,
+      "step": 5943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006839410676410338,
+      "loss": 2.0643,
+      "step": 5944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006838400508389109,
+      "loss": 2.1987,
+      "step": 5945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006837390253591149,
+      "loss": 1.9271,
+      "step": 5946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006836379912064143,
+      "loss": 2.1639,
+      "step": 5947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006835369483855782,
+      "loss": 2.2377,
+      "step": 5948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006834358969013761,
+      "loss": 2.0999,
+      "step": 5949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006833348367585779,
+      "loss": 2.1779,
+      "step": 5950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006832337679619536,
+      "loss": 2.0353,
+      "step": 5951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006831326905162743,
+      "loss": 2.0501,
+      "step": 5952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006830316044263106,
+      "loss": 2.175,
+      "step": 5953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006829305096968345,
+      "loss": 2.2249,
+      "step": 5954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006828294063326174,
+      "loss": 2.1446,
+      "step": 5955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006827282943384322,
+      "loss": 1.8968,
+      "step": 5956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006826271737190511,
+      "loss": 1.9589,
+      "step": 5957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006825260444792475,
+      "loss": 1.9845,
+      "step": 5958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006824249066237948,
+      "loss": 2.0373,
+      "step": 5959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000682323760157467,
+      "loss": 2.0606,
+      "step": 5960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006822226050850384,
+      "loss": 2.077,
+      "step": 5961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006821214414112839,
+      "loss": 2.0723,
+      "step": 5962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006820202691409784,
+      "loss": 2.0581,
+      "step": 5963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006819190882788977,
+      "loss": 2.0658,
+      "step": 5964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006818178988298176,
+      "loss": 2.0869,
+      "step": 5965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006817167007985146,
+      "loss": 1.997,
+      "step": 5966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006816154941897654,
+      "loss": 2.2436,
+      "step": 5967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006815142790083473,
+      "loss": 2.17,
+      "step": 5968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006814130552590377,
+      "loss": 2.0743,
+      "step": 5969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006813118229466148,
+      "loss": 2.2505,
+      "step": 5970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006812105820758568,
+      "loss": 2.2103,
+      "step": 5971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006811093326515428,
+      "loss": 1.9998,
+      "step": 5972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006810080746784514,
+      "loss": 2.2797,
+      "step": 5973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000680906808161363,
+      "loss": 2.168,
+      "step": 5974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006808055331050571,
+      "loss": 2.1007,
+      "step": 5975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006807042495143143,
+      "loss": 2.1848,
+      "step": 5976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006806029573939153,
+      "loss": 2.1265,
+      "step": 5977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006805016567486415,
+      "loss": 2.2109,
+      "step": 5978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006804003475832742,
+      "loss": 2.2232,
+      "step": 5979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006802990299025957,
+      "loss": 2.1835,
+      "step": 5980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006801977037113884,
+      "loss": 2.0622,
+      "step": 5981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006800963690144352,
+      "loss": 1.9667,
+      "step": 5982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000679995025816519,
+      "loss": 2.0892,
+      "step": 5983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006798936741224238,
+      "loss": 2.0132,
+      "step": 5984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006797923139369333,
+      "loss": 2.2184,
+      "step": 5985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006796909452648323,
+      "loss": 2.1484,
+      "step": 5986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006795895681109054,
+      "loss": 2.2339,
+      "step": 5987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000679488182479938,
+      "loss": 2.1102,
+      "step": 5988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006793867883767154,
+      "loss": 2.1431,
+      "step": 5989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000679285385806024,
+      "loss": 2.2762,
+      "step": 5990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006791839747726501,
+      "loss": 2.1156,
+      "step": 5991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006790825552813806,
+      "loss": 2.0016,
+      "step": 5992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006789811273370024,
+      "loss": 2.0273,
+      "step": 5993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006788796909443037,
+      "loss": 2.0754,
+      "step": 5994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006787782461080721,
+      "loss": 2.0377,
+      "step": 5995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006786767928330962,
+      "loss": 1.9901,
+      "step": 5996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006785753311241645,
+      "loss": 2.1884,
+      "step": 5997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006784738609860668,
+      "loss": 2.1221,
+      "step": 5998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006783723824235924,
+      "loss": 2.1016,
+      "step": 5999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006782708954415313,
+      "loss": 2.0629,
+      "step": 6000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006781694000446739,
+      "loss": 2.1296,
+      "step": 6001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006780678962378111,
+      "loss": 2.1346,
+      "step": 6002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006779663840257342,
+      "loss": 2.1556,
+      "step": 6003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006778648634132347,
+      "loss": 2.0859,
+      "step": 6004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006777633344051045,
+      "loss": 2.1302,
+      "step": 6005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006776617970061363,
+      "loss": 2.045,
+      "step": 6006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006775602512211225,
+      "loss": 2.214,
+      "step": 6007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006774586970548567,
+      "loss": 2.0677,
+      "step": 6008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006773571345121322,
+      "loss": 2.0285,
+      "step": 6009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006772555635977431,
+      "loss": 2.0673,
+      "step": 6010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006771539843164838,
+      "loss": 2.0561,
+      "step": 6011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000677052396673149,
+      "loss": 2.0353,
+      "step": 6012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000676950800672534,
+      "loss": 2.242,
+      "step": 6013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006768491963194342,
+      "loss": 2.2642,
+      "step": 6014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006767475836186456,
+      "loss": 2.2264,
+      "step": 6015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006766459625749649,
+      "loss": 2.0649,
+      "step": 6016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000676544333193188,
+      "loss": 2.0381,
+      "step": 6017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000676442695478113,
+      "loss": 2.2254,
+      "step": 6018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006763410494345368,
+      "loss": 2.0098,
+      "step": 6019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006762393950672577,
+      "loss": 2.1628,
+      "step": 6020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006761377323810738,
+      "loss": 2.017,
+      "step": 6021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006760360613807839,
+      "loss": 2.2401,
+      "step": 6022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000675934382071187,
+      "loss": 2.0552,
+      "step": 6023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006758326944570827,
+      "loss": 1.9136,
+      "step": 6024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006757309985432708,
+      "loss": 2.232,
+      "step": 6025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006756292943345518,
+      "loss": 2.175,
+      "step": 6026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006755275818357261,
+      "loss": 2.1207,
+      "step": 6027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006754258610515948,
+      "loss": 2.0699,
+      "step": 6028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006753241319869596,
+      "loss": 2.036,
+      "step": 6029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006752223946466222,
+      "loss": 2.095,
+      "step": 6030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006751206490353848,
+      "loss": 2.0259,
+      "step": 6031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006750188951580499,
+      "loss": 2.1144,
+      "step": 6032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006749171330194208,
+      "loss": 2.219,
+      "step": 6033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006748153626243008,
+      "loss": 2.1674,
+      "step": 6034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006747135839774935,
+      "loss": 2.0587,
+      "step": 6035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006746117970838033,
+      "loss": 2.2695,
+      "step": 6036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0006745100019480349,
+      "loss": 2.1289,
+      "step": 6037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000674408198574993,
+      "loss": 2.1612,
+      "step": 6038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006743063869694829,
+      "loss": 2.0346,
+      "step": 6039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006742045671363106,
+      "loss": 2.1946,
+      "step": 6040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006741027390802821,
+      "loss": 2.0579,
+      "step": 6041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006740009028062039,
+      "loss": 2.2448,
+      "step": 6042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006738990583188831,
+      "loss": 2.0697,
+      "step": 6043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006737972056231268,
+      "loss": 2.1766,
+      "step": 6044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006736953447237425,
+      "loss": 2.271,
+      "step": 6045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006735934756255386,
+      "loss": 2.1809,
+      "step": 6046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006734915983333236,
+      "loss": 2.1342,
+      "step": 6047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000673389712851906,
+      "loss": 2.142,
+      "step": 6048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006732878191860953,
+      "loss": 2.171,
+      "step": 6049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006731859173407009,
+      "loss": 2.1761,
+      "step": 6050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006730840073205331,
+      "loss": 1.9134,
+      "step": 6051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006729820891304021,
+      "loss": 2.1366,
+      "step": 6052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006728801627751186,
+      "loss": 2.1327,
+      "step": 6053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006727782282594938,
+      "loss": 2.2586,
+      "step": 6054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006726762855883395,
+      "loss": 2.2479,
+      "step": 6055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006725743347664673,
+      "loss": 2.1807,
+      "step": 6056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006724723757986897,
+      "loss": 2.0182,
+      "step": 6057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006723704086898192,
+      "loss": 2.0242,
+      "step": 6058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006722684334446691,
+      "loss": 1.9061,
+      "step": 6059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006721664500680528,
+      "loss": 2.0934,
+      "step": 6060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006720644585647841,
+      "loss": 1.8254,
+      "step": 6061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006719624589396772,
+      "loss": 2.1616,
+      "step": 6062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006718604511975469,
+      "loss": 1.968,
+      "step": 6063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000671758435343208,
+      "loss": 2.1365,
+      "step": 6064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000671656411381476,
+      "loss": 2.1264,
+      "step": 6065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006715543793171665,
+      "loss": 1.9976,
+      "step": 6066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000671452339155096,
+      "loss": 2.1778,
+      "step": 6067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006713502909000807,
+      "loss": 2.1799,
+      "step": 6068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006712482345569377,
+      "loss": 2.1518,
+      "step": 6069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006711461701304841,
+      "loss": 2.0703,
+      "step": 6070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006710440976255377,
+      "loss": 1.9689,
+      "step": 6071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006709420170469166,
+      "loss": 2.2084,
+      "step": 6072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006708399283994391,
+      "loss": 2.1271,
+      "step": 6073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006707378316879243,
+      "loss": 2.1741,
+      "step": 6074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000670635726917191,
+      "loss": 2.1668,
+      "step": 6075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006705336140920591,
+      "loss": 2.1445,
+      "step": 6076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006704314932173483,
+      "loss": 1.9789,
+      "step": 6077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006703293642978792,
+      "loss": 2.0866,
+      "step": 6078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006702272273384724,
+      "loss": 2.2911,
+      "step": 6079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000670125082343949,
+      "loss": 2.1203,
+      "step": 6080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006700229293191305,
+      "loss": 2.0319,
+      "step": 6081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006699207682688386,
+      "loss": 2.1033,
+      "step": 6082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006698185991978958,
+      "loss": 2.0077,
+      "step": 6083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006697164221111245,
+      "loss": 2.0186,
+      "step": 6084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000669614237013348,
+      "loss": 2.03,
+      "step": 6085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006695120439093893,
+      "loss": 2.1267,
+      "step": 6086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006694098428040722,
+      "loss": 2.1577,
+      "step": 6087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006693076337022211,
+      "loss": 2.1389,
+      "step": 6088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006692054166086603,
+      "loss": 2.2324,
+      "step": 6089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006691031915282146,
+      "loss": 2.1922,
+      "step": 6090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006690009584657095,
+      "loss": 2.0231,
+      "step": 6091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006688987174259705,
+      "loss": 2.0864,
+      "step": 6092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006687964684138237,
+      "loss": 2.0318,
+      "step": 6093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006686942114340952,
+      "loss": 2.0876,
+      "step": 6094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006685919464916123,
+      "loss": 2.1207,
+      "step": 6095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006684896735912015,
+      "loss": 2.0558,
+      "step": 6096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006683873927376909,
+      "loss": 2.1148,
+      "step": 6097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000668285103935908,
+      "loss": 1.9707,
+      "step": 6098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006681828071906812,
+      "loss": 2.2367,
+      "step": 6099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000668080502506839,
+      "loss": 1.9942,
+      "step": 6100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006679781898892108,
+      "loss": 2.2627,
+      "step": 6101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006678758693426254,
+      "loss": 2.2339,
+      "step": 6102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006677735408719132,
+      "loss": 2.0747,
+      "step": 6103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000667671204481904,
+      "loss": 2.1833,
+      "step": 6104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006675688601774283,
+      "loss": 2.16,
+      "step": 6105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000667466507963317,
+      "loss": 2.0062,
+      "step": 6106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006673641478444016,
+      "loss": 2.085,
+      "step": 6107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006672617798255133,
+      "loss": 2.1304,
+      "step": 6108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006671594039114845,
+      "loss": 2.0573,
+      "step": 6109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006670570201071474,
+      "loss": 1.8961,
+      "step": 6110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006669546284173349,
+      "loss": 2.1515,
+      "step": 6111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006668522288468798,
+      "loss": 2.0035,
+      "step": 6112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006667498214006158,
+      "loss": 1.9524,
+      "step": 6113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006666474060833768,
+      "loss": 2.1322,
+      "step": 6114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006665449828999971,
+      "loss": 1.9621,
+      "step": 6115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000666442551855311,
+      "loss": 2.1444,
+      "step": 6116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000666340112954154,
+      "loss": 2.0959,
+      "step": 6117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006662376662013609,
+      "loss": 2.098,
+      "step": 6118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006661352116017677,
+      "loss": 1.9965,
+      "step": 6119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006660327491602104,
+      "loss": 2.0071,
+      "step": 6120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006659302788815255,
+      "loss": 2.2818,
+      "step": 6121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00066582780077055,
+      "loss": 2.0454,
+      "step": 6122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006657253148321209,
+      "loss": 2.1446,
+      "step": 6123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006656228210710759,
+      "loss": 2.217,
+      "step": 6124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006655203194922528,
+      "loss": 2.0055,
+      "step": 6125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006654178101004898,
+      "loss": 2.1188,
+      "step": 6126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000665315292900626,
+      "loss": 2.2177,
+      "step": 6127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006652127678975001,
+      "loss": 1.9935,
+      "step": 6128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006651102350959517,
+      "loss": 1.9796,
+      "step": 6129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006650076945008205,
+      "loss": 2.1408,
+      "step": 6130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006649051461169466,
+      "loss": 2.2155,
+      "step": 6131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006648025899491705,
+      "loss": 2.1414,
+      "step": 6132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006647000260023333,
+      "loss": 2.1856,
+      "step": 6133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006645974542812761,
+      "loss": 2.0437,
+      "step": 6134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006644948747908405,
+      "loss": 2.0582,
+      "step": 6135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006643922875358685,
+      "loss": 2.1233,
+      "step": 6136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006642896925212023,
+      "loss": 1.992,
+      "step": 6137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000664187089751685,
+      "loss": 2.1048,
+      "step": 6138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006640844792321595,
+      "loss": 2.1726,
+      "step": 6139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006639818609674694,
+      "loss": 1.9863,
+      "step": 6140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000663879234962458,
+      "loss": 2.0137,
+      "step": 6141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006637766012219701,
+      "loss": 2.0221,
+      "step": 6142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006636739597508499,
+      "loss": 2.2465,
+      "step": 6143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006635713105539425,
+      "loss": 2.1457,
+      "step": 6144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000663468653636093,
+      "loss": 2.133,
+      "step": 6145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006633659890021471,
+      "loss": 2.1894,
+      "step": 6146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006632633166569509,
+      "loss": 1.9799,
+      "step": 6147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006631606366053507,
+      "loss": 2.1579,
+      "step": 6148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006630579488521931,
+      "loss": 2.1594,
+      "step": 6149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006629552534023254,
+      "loss": 2.2065,
+      "step": 6150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000662852550260595,
+      "loss": 2.1603,
+      "step": 6151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006627498394318499,
+      "loss": 2.1486,
+      "step": 6152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006626471209209378,
+      "loss": 2.2236,
+      "step": 6153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006625443947327078,
+      "loss": 2.035,
+      "step": 6154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006624416608720084,
+      "loss": 2.0561,
+      "step": 6155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006623389193436891,
+      "loss": 2.1378,
+      "step": 6156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006622361701525996,
+      "loss": 1.9775,
+      "step": 6157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006621334133035897,
+      "loss": 2.0896,
+      "step": 6158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006620306488015098,
+      "loss": 2.0539,
+      "step": 6159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006619278766512107,
+      "loss": 2.0159,
+      "step": 6160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006618250968575434,
+      "loss": 2.0785,
+      "step": 6161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006617223094253594,
+      "loss": 2.1176,
+      "step": 6162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006616195143595104,
+      "loss": 2.1414,
+      "step": 6163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006615167116648488,
+      "loss": 2.0487,
+      "step": 6164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006614139013462268,
+      "loss": 2.1233,
+      "step": 6165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006613110834084977,
+      "loss": 2.1032,
+      "step": 6166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006612082578565143,
+      "loss": 2.1321,
+      "step": 6167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006611054246951306,
+      "loss": 2.281,
+      "step": 6168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006610025839292001,
+      "loss": 2.0446,
+      "step": 6169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006608997355635777,
+      "loss": 2.0474,
+      "step": 6170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006607968796031174,
+      "loss": 2.1121,
+      "step": 6171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006606940160526749,
+      "loss": 2.0675,
+      "step": 6172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006605911449171051,
+      "loss": 2.2074,
+      "step": 6173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006604882662012641,
+      "loss": 2.1691,
+      "step": 6174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006603853799100078,
+      "loss": 2.098,
+      "step": 6175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006602824860481928,
+      "loss": 2.1477,
+      "step": 6176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006601795846206758,
+      "loss": 2.0042,
+      "step": 6177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000660076675632314,
+      "loss": 2.0568,
+      "step": 6178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000659973759087965,
+      "loss": 2.0839,
+      "step": 6179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006598708349924868,
+      "loss": 2.2231,
+      "step": 6180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006597679033507373,
+      "loss": 2.1474,
+      "step": 6181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006596649641675756,
+      "loss": 2.1554,
+      "step": 6182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006595620174478603,
+      "loss": 1.9573,
+      "step": 6183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006594590631964508,
+      "loss": 1.9798,
+      "step": 6184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006593561014182069,
+      "loss": 2.0682,
+      "step": 6185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0006592531321179887,
+      "loss": 2.0192,
+      "step": 6186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006591501553006561,
+      "loss": 2.0909,
+      "step": 6187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006590471709710704,
+      "loss": 2.0784,
+      "step": 6188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006589441791340923,
+      "loss": 2.084,
+      "step": 6189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006588411797945836,
+      "loss": 2.1835,
+      "step": 6190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006587381729574057,
+      "loss": 2.1238,
+      "step": 6191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006586351586274213,
+      "loss": 2.0873,
+      "step": 6192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006585321368094924,
+      "loss": 2.0435,
+      "step": 6193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006584291075084821,
+      "loss": 2.0687,
+      "step": 6194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006583260707292537,
+      "loss": 1.9842,
+      "step": 6195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006582230264766706,
+      "loss": 2.098,
+      "step": 6196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006581199747555966,
+      "loss": 2.1542,
+      "step": 6197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006580169155708964,
+      "loss": 2.1395,
+      "step": 6198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006579138489274343,
+      "loss": 2.074,
+      "step": 6199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006578107748300754,
+      "loss": 2.1424,
+      "step": 6200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006577076932836851,
+      "loss": 2.1807,
+      "step": 6201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000657604604293129,
+      "loss": 1.9619,
+      "step": 6202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006575015078632729,
+      "loss": 2.129,
+      "step": 6203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006573984039989837,
+      "loss": 2.0472,
+      "step": 6204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006572952927051276,
+      "loss": 2.0267,
+      "step": 6205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006571921739865722,
+      "loss": 2.0424,
+      "step": 6206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006570890478481845,
+      "loss": 1.9573,
+      "step": 6207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006569859142948328,
+      "loss": 2.146,
+      "step": 6208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006568827733313847,
+      "loss": 1.8834,
+      "step": 6209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006567796249627088,
+      "loss": 2.0536,
+      "step": 6210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006566764691936742,
+      "loss": 2.1183,
+      "step": 6211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00065657330602915,
+      "loss": 2.0926,
+      "step": 6212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006564701354740055,
+      "loss": 2.083,
+      "step": 6213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006563669575331111,
+      "loss": 2.1711,
+      "step": 6214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006562637722113365,
+      "loss": 2.1307,
+      "step": 6215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006561605795135525,
+      "loss": 2.0876,
+      "step": 6216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00065605737944463,
+      "loss": 2.1039,
+      "step": 6217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006559541720094404,
+      "loss": 2.2018,
+      "step": 6218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006558509572128551,
+      "loss": 2.0772,
+      "step": 6219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006557477350597465,
+      "loss": 1.9912,
+      "step": 6220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006556445055549864,
+      "loss": 2.1279,
+      "step": 6221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006555412687034478,
+      "loss": 2.222,
+      "step": 6222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006554380245100035,
+      "loss": 2.2246,
+      "step": 6223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006553347729795271,
+      "loss": 2.0276,
+      "step": 6224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006552315141168921,
+      "loss": 2.1212,
+      "step": 6225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006551282479269728,
+      "loss": 2.1439,
+      "step": 6226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006550249744146433,
+      "loss": 2.2118,
+      "step": 6227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006549216935847786,
+      "loss": 1.9563,
+      "step": 6228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006548184054422537,
+      "loss": 2.0276,
+      "step": 6229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000654715109991944,
+      "loss": 2.1118,
+      "step": 6230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006546118072387252,
+      "loss": 2.0358,
+      "step": 6231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006545084971874737,
+      "loss": 2.1426,
+      "step": 6232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006544051798430658,
+      "loss": 2.0654,
+      "step": 6233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006543018552103784,
+      "loss": 2.2537,
+      "step": 6234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006541985232942883,
+      "loss": 2.1725,
+      "step": 6235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006540951840996736,
+      "loss": 2.0398,
+      "step": 6236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006539918376314116,
+      "loss": 1.9911,
+      "step": 6237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000653888483894381,
+      "loss": 2.1656,
+      "step": 6238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006537851228934598,
+      "loss": 2.0621,
+      "step": 6239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006536817546335275,
+      "loss": 2.2881,
+      "step": 6240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006535783791194625,
+      "loss": 2.1238,
+      "step": 6241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006534749963561452,
+      "loss": 2.2679,
+      "step": 6242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006533716063484549,
+      "loss": 2.0097,
+      "step": 6243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006532682091012721,
+      "loss": 2.0968,
+      "step": 6244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006531648046194775,
+      "loss": 1.9537,
+      "step": 6245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006530613929079519,
+      "loss": 2.0708,
+      "step": 6246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006529579739715763,
+      "loss": 2.1302,
+      "step": 6247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006528545478152329,
+      "loss": 2.2477,
+      "step": 6248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006527511144438031,
+      "loss": 2.083,
+      "step": 6249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006526476738621696,
+      "loss": 2.1396,
+      "step": 6250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006525442260752147,
+      "loss": 1.9542,
+      "step": 6251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006524407710878216,
+      "loss": 2.0232,
+      "step": 6252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006523373089048734,
+      "loss": 2.0342,
+      "step": 6253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006522338395312542,
+      "loss": 2.039,
+      "step": 6254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006521303629718474,
+      "loss": 2.0221,
+      "step": 6255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006520268792315378,
+      "loss": 2.0806,
+      "step": 6256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006519233883152097,
+      "loss": 2.1048,
+      "step": 6257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006518198902277486,
+      "loss": 2.171,
+      "step": 6258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006517163849740392,
+      "loss": 2.0441,
+      "step": 6259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006516128725589678,
+      "loss": 2.0584,
+      "step": 6260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006515093529874199,
+      "loss": 2.1023,
+      "step": 6261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006514058262642823,
+      "loss": 1.9779,
+      "step": 6262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006513022923944415,
+      "loss": 2.0317,
+      "step": 6263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006511987513827846,
+      "loss": 2.0374,
+      "step": 6264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006510952032341988,
+      "loss": 2.1605,
+      "step": 6265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006509916479535719,
+      "loss": 2.2567,
+      "step": 6266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006508880855457922,
+      "loss": 2.0785,
+      "step": 6267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006507845160157477,
+      "loss": 2.0667,
+      "step": 6268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006506809393683273,
+      "loss": 2.1335,
+      "step": 6269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00065057735560842,
+      "loss": 2.1923,
+      "step": 6270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006504737647409154,
+      "loss": 2.1168,
+      "step": 6271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000650370166770703,
+      "loss": 1.9747,
+      "step": 6272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006502665617026728,
+      "loss": 2.1238,
+      "step": 6273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006501629495417155,
+      "loss": 2.137,
+      "step": 6274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006500593302927216,
+      "loss": 2.0187,
+      "step": 6275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006499557039605821,
+      "loss": 2.1135,
+      "step": 6276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006498520705501888,
+      "loss": 2.0372,
+      "step": 6277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006497484300664329,
+      "loss": 2.196,
+      "step": 6278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006496447825142069,
+      "loss": 2.2975,
+      "step": 6279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006495411278984029,
+      "loss": 2.1108,
+      "step": 6280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006494374662239139,
+      "loss": 2.1159,
+      "step": 6281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006493337974956327,
+      "loss": 2.1566,
+      "step": 6282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006492301217184529,
+      "loss": 2.0289,
+      "step": 6283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006491264388972682,
+      "loss": 2.0808,
+      "step": 6284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006490227490369727,
+      "loss": 2.1473,
+      "step": 6285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006489190521424605,
+      "loss": 2.1943,
+      "step": 6286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006488153482186267,
+      "loss": 2.0107,
+      "step": 6287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006487116372703663,
+      "loss": 2.1867,
+      "step": 6288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006486079193025746,
+      "loss": 1.9582,
+      "step": 6289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006485041943201474,
+      "loss": 2.1018,
+      "step": 6290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006484004623279806,
+      "loss": 2.067,
+      "step": 6291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006482967233309709,
+      "loss": 1.9515,
+      "step": 6292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006481929773340147,
+      "loss": 2.1519,
+      "step": 6293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006480892243420091,
+      "loss": 2.1271,
+      "step": 6294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006479854643598517,
+      "loss": 2.0783,
+      "step": 6295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006478816973924399,
+      "loss": 2.1978,
+      "step": 6296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006477779234446722,
+      "loss": 2.1488,
+      "step": 6297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006476741425214462,
+      "loss": 2.0497,
+      "step": 6298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006475703546276615,
+      "loss": 2.1127,
+      "step": 6299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006474665597682166,
+      "loss": 2.0606,
+      "step": 6300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006473627579480108,
+      "loss": 2.3659,
+      "step": 6301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000647258949171944,
+      "loss": 2.1186,
+      "step": 6302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006471551334449163,
+      "loss": 2.102,
+      "step": 6303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006470513107718277,
+      "loss": 2.1194,
+      "step": 6304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006469474811575793,
+      "loss": 2.0974,
+      "step": 6305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006468436446070715,
+      "loss": 2.0163,
+      "step": 6306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006467398011252063,
+      "loss": 2.07,
+      "step": 6307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006466359507168849,
+      "loss": 2.1736,
+      "step": 6308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006465320933870095,
+      "loss": 2.2321,
+      "step": 6309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006464282291404822,
+      "loss": 2.2013,
+      "step": 6310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006463243579822058,
+      "loss": 2.0803,
+      "step": 6311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006462204799170832,
+      "loss": 2.0775,
+      "step": 6312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006461165949500176,
+      "loss": 1.9873,
+      "step": 6313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006460127030859129,
+      "loss": 2.1724,
+      "step": 6314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006459088043296727,
+      "loss": 2.1876,
+      "step": 6315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006458048986862013,
+      "loss": 2.2144,
+      "step": 6316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006457009861604036,
+      "loss": 2.1651,
+      "step": 6317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006455970667571841,
+      "loss": 2.0887,
+      "step": 6318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006454931404814483,
+      "loss": 2.2043,
+      "step": 6319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006453892073381017,
+      "loss": 2.0624,
+      "step": 6320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006452852673320502,
+      "loss": 2.0946,
+      "step": 6321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006451813204681999,
+      "loss": 1.922,
+      "step": 6322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006450773667514574,
+      "loss": 2.1392,
+      "step": 6323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006449734061867296,
+      "loss": 2.2017,
+      "step": 6324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006448694387789236,
+      "loss": 2.1249,
+      "step": 6325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006447654645329469,
+      "loss": 2.2289,
+      "step": 6326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006446614834537076,
+      "loss": 2.2294,
+      "step": 6327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006445574955461133,
+      "loss": 2.0558,
+      "step": 6328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006444535008150729,
+      "loss": 2.1109,
+      "step": 6329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006443494992654951,
+      "loss": 2.2185,
+      "step": 6330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000644245490902289,
+      "loss": 2.1677,
+      "step": 6331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006441414757303638,
+      "loss": 1.9684,
+      "step": 6332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006440374537546298,
+      "loss": 2.1305,
+      "step": 6333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006439334249799965,
+      "loss": 2.1401,
+      "step": 6334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0006438293894113747,
+      "loss": 1.9684,
+      "step": 6335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006437253470536748,
+      "loss": 1.9719,
+      "step": 6336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006436212979118081,
+      "loss": 2.2389,
+      "step": 6337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006435172419906858,
+      "loss": 2.2314,
+      "step": 6338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006434131792952197,
+      "loss": 2.1522,
+      "step": 6339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006433091098303216,
+      "loss": 2.0261,
+      "step": 6340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006432050336009041,
+      "loss": 2.1804,
+      "step": 6341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006431009506118795,
+      "loss": 2.12,
+      "step": 6342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006429968608681612,
+      "loss": 2.0914,
+      "step": 6343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000642892764374662,
+      "loss": 1.9623,
+      "step": 6344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006427886611362959,
+      "loss": 2.0005,
+      "step": 6345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006426845511579765,
+      "loss": 2.1208,
+      "step": 6346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006425804344446183,
+      "loss": 1.9886,
+      "step": 6347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006424763110011355,
+      "loss": 2.0453,
+      "step": 6348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006423721808324434,
+      "loss": 2.0434,
+      "step": 6349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006422680439434567,
+      "loss": 2.1357,
+      "step": 6350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006421639003390914,
+      "loss": 2.1542,
+      "step": 6351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006420597500242629,
+      "loss": 2.0164,
+      "step": 6352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006419555930038876,
+      "loss": 2.1493,
+      "step": 6353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006418514292828818,
+      "loss": 1.9862,
+      "step": 6354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006417472588661625,
+      "loss": 2.0116,
+      "step": 6355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006416430817586464,
+      "loss": 2.1773,
+      "step": 6356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006415388979652511,
+      "loss": 2.0338,
+      "step": 6357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006414347074908944,
+      "loss": 2.099,
+      "step": 6358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006413305103404943,
+      "loss": 2.0941,
+      "step": 6359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000641226306518969,
+      "loss": 1.9807,
+      "step": 6360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006411220960312373,
+      "loss": 2.0802,
+      "step": 6361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006410178788822182,
+      "loss": 2.0898,
+      "step": 6362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006409136550768308,
+      "loss": 2.0881,
+      "step": 6363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006408094246199949,
+      "loss": 2.0959,
+      "step": 6364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006407051875166304,
+      "loss": 2.1015,
+      "step": 6365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006406009437716574,
+      "loss": 1.9626,
+      "step": 6366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006404966933899964,
+      "loss": 2.0398,
+      "step": 6367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006403924363765685,
+      "loss": 2.068,
+      "step": 6368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006402881727362948,
+      "loss": 2.166,
+      "step": 6369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006401839024740967,
+      "loss": 1.8979,
+      "step": 6370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000640079625594896,
+      "loss": 2.0548,
+      "step": 6371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006399753421036147,
+      "loss": 2.0501,
+      "step": 6372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006398710520051756,
+      "loss": 2.0814,
+      "step": 6373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000639766755304501,
+      "loss": 2.119,
+      "step": 6374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006396624520065142,
+      "loss": 2.1167,
+      "step": 6375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006395581421161386,
+      "loss": 2.1711,
+      "step": 6376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006394538256382975,
+      "loss": 2.1354,
+      "step": 6377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006393495025779154,
+      "loss": 2.0222,
+      "step": 6378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006392451729399162,
+      "loss": 2.0343,
+      "step": 6379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006391408367292246,
+      "loss": 2.0769,
+      "step": 6380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006390364939507654,
+      "loss": 2.1096,
+      "step": 6381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006389321446094642,
+      "loss": 2.1698,
+      "step": 6382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006388277887102461,
+      "loss": 2.1315,
+      "step": 6383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006387234262580372,
+      "loss": 2.0541,
+      "step": 6384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006386190572577636,
+      "loss": 2.1985,
+      "step": 6385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006385146817143517,
+      "loss": 2.1535,
+      "step": 6386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006384102996327282,
+      "loss": 2.0743,
+      "step": 6387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006383059110178205,
+      "loss": 2.0173,
+      "step": 6388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006382015158745553,
+      "loss": 1.9913,
+      "step": 6389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006380971142078611,
+      "loss": 2.2322,
+      "step": 6390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006379927060226653,
+      "loss": 2.1771,
+      "step": 6391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006378882913238967,
+      "loss": 2.0095,
+      "step": 6392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006377838701164834,
+      "loss": 2.252,
+      "step": 6393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006376794424053547,
+      "loss": 2.104,
+      "step": 6394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006375750081954397,
+      "loss": 2.0429,
+      "step": 6395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006374705674916679,
+      "loss": 2.1654,
+      "step": 6396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006373661202989691,
+      "loss": 2.0893,
+      "step": 6397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006372616666222737,
+      "loss": 1.9779,
+      "step": 6398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006371572064665119,
+      "loss": 2.1893,
+      "step": 6399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006370527398366146,
+      "loss": 2.054,
+      "step": 6400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006369482667375126,
+      "loss": 2.1801,
+      "step": 6401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006368437871741377,
+      "loss": 2.0985,
+      "step": 6402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006367393011514212,
+      "loss": 2.1198,
+      "step": 6403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006366348086742954,
+      "loss": 2.2509,
+      "step": 6404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006365303097476922,
+      "loss": 1.9978,
+      "step": 6405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006364258043765445,
+      "loss": 2.0448,
+      "step": 6406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006363212925657851,
+      "loss": 2.1382,
+      "step": 6407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006362167743203473,
+      "loss": 2.1713,
+      "step": 6408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006361122496451643,
+      "loss": 2.1459,
+      "step": 6409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006360077185451703,
+      "loss": 2.0354,
+      "step": 6410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006359031810252991,
+      "loss": 2.1522,
+      "step": 6411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006357986370904852,
+      "loss": 2.2028,
+      "step": 6412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006356940867456633,
+      "loss": 1.9868,
+      "step": 6413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006355895299957686,
+      "loss": 1.98,
+      "step": 6414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006354849668457362,
+      "loss": 2.1992,
+      "step": 6415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006353803973005019,
+      "loss": 2.2245,
+      "step": 6416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006352758213650013,
+      "loss": 2.1202,
+      "step": 6417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000635171239044171,
+      "loss": 2.1576,
+      "step": 6418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006350666503429472,
+      "loss": 2.1238,
+      "step": 6419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006349620552662671,
+      "loss": 1.9492,
+      "step": 6420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006348574538190675,
+      "loss": 2.1063,
+      "step": 6421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000634752846006286,
+      "loss": 1.9273,
+      "step": 6422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006346482318328602,
+      "loss": 2.3033,
+      "step": 6423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006345436113037284,
+      "loss": 2.0671,
+      "step": 6424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006344389844238286,
+      "loss": 2.1434,
+      "step": 6425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006343343511980996,
+      "loss": 1.9778,
+      "step": 6426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006342297116314802,
+      "loss": 2.2224,
+      "step": 6427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00063412506572891,
+      "loss": 2.2266,
+      "step": 6428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006340204134953279,
+      "loss": 2.1679,
+      "step": 6429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006339157549356743,
+      "loss": 2.088,
+      "step": 6430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000633811090054889,
+      "loss": 2.0275,
+      "step": 6431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006337064188579127,
+      "loss": 2.0396,
+      "step": 6432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006336017413496858,
+      "loss": 2.0768,
+      "step": 6433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006334970575351495,
+      "loss": 2.0782,
+      "step": 6434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000633392367419245,
+      "loss": 2.1046,
+      "step": 6435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000633287671006914,
+      "loss": 2.0896,
+      "step": 6436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006331829683030985,
+      "loss": 2.185,
+      "step": 6437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006330782593127407,
+      "loss": 2.1209,
+      "step": 6438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006329735440407829,
+      "loss": 2.2851,
+      "step": 6439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006328688224921682,
+      "loss": 2.1108,
+      "step": 6440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006327640946718391,
+      "loss": 2.2482,
+      "step": 6441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00063265936058474,
+      "loss": 2.0223,
+      "step": 6442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006325546202358136,
+      "loss": 2.0578,
+      "step": 6443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006324498736300044,
+      "loss": 2.0644,
+      "step": 6444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006323451207722564,
+      "loss": 2.1476,
+      "step": 6445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006322403616675148,
+      "loss": 2.1425,
+      "step": 6446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006321355963207236,
+      "loss": 2.114,
+      "step": 6447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006320308247368285,
+      "loss": 2.0805,
+      "step": 6448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006319260469207748,
+      "loss": 1.987,
+      "step": 6449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006318212628775085,
+      "loss": 2.1137,
+      "step": 6450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006317164726119753,
+      "loss": 2.1036,
+      "step": 6451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006316116761291218,
+      "loss": 2.0771,
+      "step": 6452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006315068734338943,
+      "loss": 2.2168,
+      "step": 6453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006314020645312401,
+      "loss": 2.0569,
+      "step": 6454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006312972494261062,
+      "loss": 1.9968,
+      "step": 6455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006311924281234404,
+      "loss": 2.0116,
+      "step": 6456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00063108760062819,
+      "loss": 2.078,
+      "step": 6457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006309827669453037,
+      "loss": 2.1547,
+      "step": 6458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006308779270797292,
+      "loss": 2.175,
+      "step": 6459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006307730810364158,
+      "loss": 2.0657,
+      "step": 6460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006306682288203122,
+      "loss": 1.9827,
+      "step": 6461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006305633704363678,
+      "loss": 2.1352,
+      "step": 6462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000630458505889532,
+      "loss": 2.0819,
+      "step": 6463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006303536351847548,
+      "loss": 2.2205,
+      "step": 6464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000630248758326986,
+      "loss": 2.0739,
+      "step": 6465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006301438753211764,
+      "loss": 2.0656,
+      "step": 6466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006300389861722766,
+      "loss": 2.0953,
+      "step": 6467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006299340908852377,
+      "loss": 2.0846,
+      "step": 6468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006298291894650108,
+      "loss": 2.2178,
+      "step": 6469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006297242819165476,
+      "loss": 2.1504,
+      "step": 6470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006296193682447999,
+      "loss": 2.0148,
+      "step": 6471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006295144484547202,
+      "loss": 1.993,
+      "step": 6472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006294095225512603,
+      "loss": 2.0379,
+      "step": 6473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006293045905393737,
+      "loss": 2.0668,
+      "step": 6474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006291996524240129,
+      "loss": 2.1252,
+      "step": 6475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006290947082101315,
+      "loss": 2.0997,
+      "step": 6476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006289897579026827,
+      "loss": 2.0727,
+      "step": 6477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006288848015066211,
+      "loss": 2.0397,
+      "step": 6478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006287798390269001,
+      "loss": 2.1164,
+      "step": 6479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006286748704684748,
+      "loss": 2.0234,
+      "step": 6480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006285698958362995,
+      "loss": 2.113,
+      "step": 6481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006284649151353296,
+      "loss": 2.1248,
+      "step": 6482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006283599283705203,
+      "loss": 2.0575,
+      "step": 6483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0006282549355468273,
+      "loss": 2.1389,
+      "step": 6484
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000628149936669206,
+      "loss": 1.9885,
+      "step": 6485
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006280449317426132,
+      "loss": 2.0456,
+      "step": 6486
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006279399207720053,
+      "loss": 2.1322,
+      "step": 6487
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006278349037623388,
+      "loss": 2.1429,
+      "step": 6488
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006277298807185709,
+      "loss": 2.1656,
+      "step": 6489
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006276248516456589,
+      "loss": 2.1321,
+      "step": 6490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006275198165485603,
+      "loss": 2.1501,
+      "step": 6491
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006274147754322331,
+      "loss": 1.9398,
+      "step": 6492
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006273097283016357,
+      "loss": 2.1595,
+      "step": 6493
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006272046751617262,
+      "loss": 2.1357,
+      "step": 6494
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006270996160174636,
+      "loss": 2.0522,
+      "step": 6495
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006269945508738066,
+      "loss": 1.9632,
+      "step": 6496
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000626889479735715,
+      "loss": 2.1617,
+      "step": 6497
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006267844026081479,
+      "loss": 2.3341,
+      "step": 6498
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006266793194960657,
+      "loss": 2.0263,
+      "step": 6499
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006265742304044281,
+      "loss": 1.9595,
+      "step": 6500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006264691353381959,
+      "loss": 2.0284,
+      "step": 6501
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006263640343023296,
+      "loss": 2.2075,
+      "step": 6502
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006262589273017905,
+      "loss": 1.9968,
+      "step": 6503
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006261538143415392,
+      "loss": 2.0954,
+      "step": 6504
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006260486954265382,
+      "loss": 2.0074,
+      "step": 6505
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006259435705617487,
+      "loss": 1.9301,
+      "step": 6506
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000625838439752133,
+      "loss": 2.1689,
+      "step": 6507
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006257333030026537,
+      "loss": 2.094,
+      "step": 6508
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006256281603182733,
+      "loss": 2.0631,
+      "step": 6509
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006255230117039548,
+      "loss": 2.1466,
+      "step": 6510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006254178571646616,
+      "loss": 2.1522,
+      "step": 6511
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000625312696705357,
+      "loss": 2.0973,
+      "step": 6512
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006252075303310049,
+      "loss": 2.1113,
+      "step": 6513
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006251023580465695,
+      "loss": 2.0149,
+      "step": 6514
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006249971798570151,
+      "loss": 2.2288,
+      "step": 6515
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006248919957673063,
+      "loss": 1.9644,
+      "step": 6516
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006247868057824082,
+      "loss": 2.1819,
+      "step": 6517
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006246816099072859,
+      "loss": 2.1348,
+      "step": 6518
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006245764081469048,
+      "loss": 2.1114,
+      "step": 6519
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006244712005062308,
+      "loss": 2.1773,
+      "step": 6520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006243659869902299,
+      "loss": 1.9195,
+      "step": 6521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006242607676038683,
+      "loss": 1.9684,
+      "step": 6522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006241555423521128,
+      "loss": 1.9069,
+      "step": 6523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006240503112399302,
+      "loss": 1.9297,
+      "step": 6524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006239450742722876,
+      "loss": 2.131,
+      "step": 6525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006238398314541524,
+      "loss": 2.0013,
+      "step": 6526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006237345827904925,
+      "loss": 2.1382,
+      "step": 6527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006236293282862757,
+      "loss": 2.0989,
+      "step": 6528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006235240679464703,
+      "loss": 2.1388,
+      "step": 6529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006234188017760447,
+      "loss": 2.1324,
+      "step": 6530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006233135297799681,
+      "loss": 2.1245,
+      "step": 6531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000623208251963209,
+      "loss": 2.0338,
+      "step": 6532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006231029683307374,
+      "loss": 2.221,
+      "step": 6533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006229976788875223,
+      "loss": 2.0397,
+      "step": 6534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000622892383638534,
+      "loss": 1.9326,
+      "step": 6535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006227870825887426,
+      "loss": 2.2176,
+      "step": 6536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006226817757431185,
+      "loss": 2.029,
+      "step": 6537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006225764631066325,
+      "loss": 2.2453,
+      "step": 6538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006224711446842556,
+      "loss": 2.0081,
+      "step": 6539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006223658204809589,
+      "loss": 2.1111,
+      "step": 6540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006222604905017141,
+      "loss": 2.0216,
+      "step": 6541
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006221551547514931,
+      "loss": 2.0964,
+      "step": 6542
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006220498132352679,
+      "loss": 2.0181,
+      "step": 6543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006219444659580107,
+      "loss": 2.0946,
+      "step": 6544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006218391129246943,
+      "loss": 2.0109,
+      "step": 6545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006217337541402916,
+      "loss": 2.068,
+      "step": 6546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006216283896097759,
+      "loss": 2.0897,
+      "step": 6547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006215230193381203,
+      "loss": 1.9568,
+      "step": 6548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006214176433302988,
+      "loss": 1.9562,
+      "step": 6549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006213122615912853,
+      "loss": 2.2228,
+      "step": 6550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000621206874126054,
+      "loss": 2.1491,
+      "step": 6551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006211014809395796,
+      "loss": 2.17,
+      "step": 6552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006209960820368369,
+      "loss": 2.0599,
+      "step": 6553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006208906774228007,
+      "loss": 2.164,
+      "step": 6554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006207852671024466,
+      "loss": 2.1014,
+      "step": 6555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006206798510807501,
+      "loss": 2.0622,
+      "step": 6556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006205744293626872,
+      "loss": 2.1288,
+      "step": 6557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006204690019532337,
+      "loss": 2.0018,
+      "step": 6558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006203635688573666,
+      "loss": 1.9111,
+      "step": 6559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000620258130080062,
+      "loss": 2.2809,
+      "step": 6560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006201526856262973,
+      "loss": 2.0813,
+      "step": 6561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006200472355010493,
+      "loss": 2.0481,
+      "step": 6562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006199417797092959,
+      "loss": 2.0085,
+      "step": 6563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006198363182560147,
+      "loss": 2.1141,
+      "step": 6564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006197308511461836,
+      "loss": 1.9506,
+      "step": 6565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006196253783847808,
+      "loss": 2.1607,
+      "step": 6566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006195198999767854,
+      "loss": 1.9545,
+      "step": 6567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006194144159271756,
+      "loss": 2.0039,
+      "step": 6568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006193089262409308,
+      "loss": 1.9013,
+      "step": 6569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006192034309230304,
+      "loss": 2.147,
+      "step": 6570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006190979299784539,
+      "loss": 2.1944,
+      "step": 6571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006189924234121813,
+      "loss": 2.2091,
+      "step": 6572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006188869112291925,
+      "loss": 2.2085,
+      "step": 6573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006187813934344682,
+      "loss": 2.0474,
+      "step": 6574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000618675870032989,
+      "loss": 2.1272,
+      "step": 6575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006185703410297356,
+      "loss": 2.0457,
+      "step": 6576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006184648064296896,
+      "loss": 2.0475,
+      "step": 6577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006183592662378323,
+      "loss": 2.1893,
+      "step": 6578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006182537204591454,
+      "loss": 2.09,
+      "step": 6579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006181481690986111,
+      "loss": 2.1503,
+      "step": 6580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006180426121612112,
+      "loss": 2.1259,
+      "step": 6581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006179370496519289,
+      "loss": 2.1154,
+      "step": 6582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006178314815757464,
+      "loss": 2.0972,
+      "step": 6583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006177259079376471,
+      "loss": 2.0708,
+      "step": 6584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006176203287426142,
+      "loss": 2.075,
+      "step": 6585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006175147439956315,
+      "loss": 1.8406,
+      "step": 6586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006174091537016823,
+      "loss": 2.147,
+      "step": 6587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006173035578657512,
+      "loss": 2.1506,
+      "step": 6588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006171979564928223,
+      "loss": 1.9134,
+      "step": 6589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006170923495878805,
+      "loss": 2.062,
+      "step": 6590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006169867371559103,
+      "loss": 2.0226,
+      "step": 6591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006168811192018974,
+      "loss": 2.1207,
+      "step": 6592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006167754957308267,
+      "loss": 1.8865,
+      "step": 6593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006166698667476841,
+      "loss": 2.1967,
+      "step": 6594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006165642322574553,
+      "loss": 2.0705,
+      "step": 6595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006164585922651269,
+      "loss": 1.916,
+      "step": 6596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006163529467756851,
+      "loss": 2.0245,
+      "step": 6597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006162472957941166,
+      "loss": 2.1353,
+      "step": 6598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006161416393254083,
+      "loss": 2.0904,
+      "step": 6599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006160359773745477,
+      "loss": 2.0007,
+      "step": 6600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006159303099465221,
+      "loss": 1.9704,
+      "step": 6601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006158246370463193,
+      "loss": 2.1524,
+      "step": 6602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006157189586789272,
+      "loss": 2.0184,
+      "step": 6603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006156132748493341,
+      "loss": 2.1067,
+      "step": 6604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006155075855625285,
+      "loss": 1.9955,
+      "step": 6605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006154018908234994,
+      "loss": 2.0893,
+      "step": 6606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006152961906372356,
+      "loss": 2.2204,
+      "step": 6607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006151904850087265,
+      "loss": 2.0645,
+      "step": 6608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006150847739429615,
+      "loss": 2.0464,
+      "step": 6609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006149790574449306,
+      "loss": 2.0498,
+      "step": 6610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006148733355196236,
+      "loss": 2.0843,
+      "step": 6611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006147676081720313,
+      "loss": 2.2371,
+      "step": 6612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006146618754071438,
+      "loss": 2.0158,
+      "step": 6613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006145561372299521,
+      "loss": 2.0455,
+      "step": 6614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006144503936454472,
+      "loss": 2.1226,
+      "step": 6615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006143446446586207,
+      "loss": 2.0701,
+      "step": 6616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006142388902744639,
+      "loss": 2.08,
+      "step": 6617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006141331304979688,
+      "loss": 2.0423,
+      "step": 6618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006140273653341274,
+      "loss": 2.1618,
+      "step": 6619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006139215947879321,
+      "loss": 1.8411,
+      "step": 6620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006138158188643758,
+      "loss": 2.0689,
+      "step": 6621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006137100375684509,
+      "loss": 2.1337,
+      "step": 6622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006136042509051507,
+      "loss": 2.1648,
+      "step": 6623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006134984588794688,
+      "loss": 2.1829,
+      "step": 6624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006133926614963985,
+      "loss": 2.0912,
+      "step": 6625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006132868587609338,
+      "loss": 2.0702,
+      "step": 6626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006131810506780688,
+      "loss": 2.1531,
+      "step": 6627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006130752372527982,
+      "loss": 2.2091,
+      "step": 6628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000612969418490116,
+      "loss": 2.0327,
+      "step": 6629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006128635943950176,
+      "loss": 2.0596,
+      "step": 6630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000612757764972498,
+      "loss": 2.0955,
+      "step": 6631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006126519302275526,
+      "loss": 2.1347,
+      "step": 6632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000612546090165177,
+      "loss": 2.1994,
+      "step": 6633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0006124402447903672,
+      "loss": 1.8747,
+      "step": 6634
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006123343941081192,
+      "loss": 1.9349,
+      "step": 6635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006122285381234296,
+      "loss": 1.934,
+      "step": 6636
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006121226768412947,
+      "loss": 2.0331,
+      "step": 6637
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006120168102667117,
+      "loss": 2.052,
+      "step": 6638
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006119109384046777,
+      "loss": 2.0493,
+      "step": 6639
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00061180506126019,
+      "loss": 2.008,
+      "step": 6640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006116991788382463,
+      "loss": 2.1829,
+      "step": 6641
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006115932911438445,
+      "loss": 2.1955,
+      "step": 6642
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006114873981819827,
+      "loss": 2.0308,
+      "step": 6643
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006113814999576595,
+      "loss": 2.0274,
+      "step": 6644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000611275596475873,
+      "loss": 2.1873,
+      "step": 6645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006111696877416227,
+      "loss": 2.0065,
+      "step": 6646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006110637737599074,
+      "loss": 2.0213,
+      "step": 6647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006109578545357268,
+      "loss": 1.9459,
+      "step": 6648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006108519300740799,
+      "loss": 2.005,
+      "step": 6649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006107460003799673,
+      "loss": 2.0414,
+      "step": 6650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006106400654583886,
+      "loss": 2.1068,
+      "step": 6651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006105341253143444,
+      "loss": 2.0533,
+      "step": 6652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006104281799528354,
+      "loss": 1.94,
+      "step": 6653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006103222293788624,
+      "loss": 2.0315,
+      "step": 6654
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006102162735974263,
+      "loss": 1.9974,
+      "step": 6655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006101103126135288,
+      "loss": 2.0293,
+      "step": 6656
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006100043464321712,
+      "loss": 2.0734,
+      "step": 6657
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006098983750583556,
+      "loss": 2.1327,
+      "step": 6658
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006097923984970837,
+      "loss": 2.0572,
+      "step": 6659
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006096864167533583,
+      "loss": 2.2314,
+      "step": 6660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006095804298321818,
+      "loss": 2.2431,
+      "step": 6661
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006094744377385568,
+      "loss": 2.0447,
+      "step": 6662
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006093684404774867,
+      "loss": 1.9812,
+      "step": 6663
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006092624380539749,
+      "loss": 2.0975,
+      "step": 6664
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006091564304730247,
+      "loss": 1.9903,
+      "step": 6665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006090504177396397,
+      "loss": 2.0587,
+      "step": 6666
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006089443998588245,
+      "loss": 2.1939,
+      "step": 6667
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006088383768355828,
+      "loss": 1.9167,
+      "step": 6668
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006087323486749196,
+      "loss": 2.0448,
+      "step": 6669
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006086263153818396,
+      "loss": 2.1502,
+      "step": 6670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006085202769613475,
+      "loss": 2.1868,
+      "step": 6671
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006084142334184489,
+      "loss": 1.9148,
+      "step": 6672
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006083081847581491,
+      "loss": 2.0929,
+      "step": 6673
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000608202130985454,
+      "loss": 2.2031,
+      "step": 6674
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006080960721053693,
+      "loss": 2.1974,
+      "step": 6675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006079900081229015,
+      "loss": 2.2595,
+      "step": 6676
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006078839390430571,
+      "loss": 1.9687,
+      "step": 6677
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006077778648708425,
+      "loss": 2.097,
+      "step": 6678
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006076717856112649,
+      "loss": 2.189,
+      "step": 6679
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006075657012693315,
+      "loss": 2.0434,
+      "step": 6680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006074596118500494,
+      "loss": 2.0714,
+      "step": 6681
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006073535173584268,
+      "loss": 2.2135,
+      "step": 6682
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000607247417799471,
+      "loss": 2.0306,
+      "step": 6683
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006071413131781907,
+      "loss": 2.0029,
+      "step": 6684
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006070352034995938,
+      "loss": 1.9907,
+      "step": 6685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006069290887686893,
+      "loss": 2.2045,
+      "step": 6686
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006068229689904859,
+      "loss": 2.0972,
+      "step": 6687
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006067168441699927,
+      "loss": 2.1429,
+      "step": 6688
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000606610714312219,
+      "loss": 2.1001,
+      "step": 6689
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006065045794221745,
+      "loss": 2.14,
+      "step": 6690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006063984395048688,
+      "loss": 2.2128,
+      "step": 6691
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006062922945653122,
+      "loss": 2.1759,
+      "step": 6692
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006061861446085147,
+      "loss": 2.1209,
+      "step": 6693
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006060799896394871,
+      "loss": 2.086,
+      "step": 6694
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00060597382966324,
+      "loss": 2.2392,
+      "step": 6695
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006058676646847845,
+      "loss": 2.0171,
+      "step": 6696
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006057614947091316,
+      "loss": 2.134,
+      "step": 6697
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006056553197412932,
+      "loss": 2.1279,
+      "step": 6698
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006055491397862805,
+      "loss": 2.0091,
+      "step": 6699
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006054429548491058,
+      "loss": 2.0438,
+      "step": 6700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000605336764934781,
+      "loss": 2.1841,
+      "step": 6701
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000605230570048319,
+      "loss": 2.0433,
+      "step": 6702
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006051243701947316,
+      "loss": 2.1307,
+      "step": 6703
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006050181653790325,
+      "loss": 2.0466,
+      "step": 6704
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006049119556062345,
+      "loss": 2.1039,
+      "step": 6705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006048057408813509,
+      "loss": 2.0101,
+      "step": 6706
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006046995212093952,
+      "loss": 2.1206,
+      "step": 6707
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006045932965953813,
+      "loss": 2.0786,
+      "step": 6708
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006044870670443234,
+      "loss": 2.1536,
+      "step": 6709
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006043808325612355,
+      "loss": 2.1923,
+      "step": 6710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006042745931511325,
+      "loss": 2.0469,
+      "step": 6711
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006041683488190288,
+      "loss": 2.032,
+      "step": 6712
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006040620995699394,
+      "loss": 2.1129,
+      "step": 6713
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006039558454088796,
+      "loss": 1.9646,
+      "step": 6714
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006038495863408649,
+      "loss": 1.9466,
+      "step": 6715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006037433223709109,
+      "loss": 2.0688,
+      "step": 6716
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006036370535040335,
+      "loss": 1.9309,
+      "step": 6717
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006035307797452489,
+      "loss": 2.1769,
+      "step": 6718
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006034245010995734,
+      "loss": 1.9918,
+      "step": 6719
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006033182175720235,
+      "loss": 2.219,
+      "step": 6720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006032119291676161,
+      "loss": 2.0507,
+      "step": 6721
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006031056358913684,
+      "loss": 2.0982,
+      "step": 6722
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006029993377482977,
+      "loss": 2.0779,
+      "step": 6723
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006028930347434211,
+      "loss": 1.9898,
+      "step": 6724
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006027867268817568,
+      "loss": 1.9333,
+      "step": 6725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006026804141683227,
+      "loss": 1.9912,
+      "step": 6726
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006025740966081368,
+      "loss": 2.0837,
+      "step": 6727
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006024677742062176,
+      "loss": 2.033,
+      "step": 6728
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006023614469675839,
+      "loss": 2.0695,
+      "step": 6729
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006022551148972545,
+      "loss": 2.1041,
+      "step": 6730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006021487780002486,
+      "loss": 2.0842,
+      "step": 6731
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006020424362815853,
+      "loss": 2.1447,
+      "step": 6732
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006019360897462845,
+      "loss": 2.1712,
+      "step": 6733
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006018297383993658,
+      "loss": 2.174,
+      "step": 6734
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006017233822458492,
+      "loss": 2.2172,
+      "step": 6735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000601617021290755,
+      "loss": 2.0302,
+      "step": 6736
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006015106555391039,
+      "loss": 2.0287,
+      "step": 6737
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006014042849959161,
+      "loss": 2.0346,
+      "step": 6738
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006012979096662132,
+      "loss": 1.9626,
+      "step": 6739
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006011915295550158,
+      "loss": 2.0709,
+      "step": 6740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006010851446673456,
+      "loss": 2.0847,
+      "step": 6741
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006009787550082241,
+      "loss": 2.1757,
+      "step": 6742
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006008723605826731,
+      "loss": 2.0503,
+      "step": 6743
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006007659613957147,
+      "loss": 2.1161,
+      "step": 6744
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006006595574523712,
+      "loss": 2.0083,
+      "step": 6745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006005531487576651,
+      "loss": 2.0251,
+      "step": 6746
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006004467353166191,
+      "loss": 2.3444,
+      "step": 6747
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006003403171342563,
+      "loss": 2.0907,
+      "step": 6748
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006002338942155998,
+      "loss": 1.9438,
+      "step": 6749
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006001274665656728,
+      "loss": 2.0597,
+      "step": 6750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0006000210341894994,
+      "loss": 2.0025,
+      "step": 6751
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000599914597092103,
+      "loss": 2.1127,
+      "step": 6752
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005998081552785079,
+      "loss": 1.9984,
+      "step": 6753
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005997017087537382,
+      "loss": 2.0275,
+      "step": 6754
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005995952575228187,
+      "loss": 2.0871,
+      "step": 6755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000599488801590774,
+      "loss": 1.9908,
+      "step": 6756
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005993823409626291,
+      "loss": 1.9908,
+      "step": 6757
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005992758756434092,
+      "loss": 1.8892,
+      "step": 6758
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005991694056381395,
+      "loss": 2.0858,
+      "step": 6759
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005990629309518458,
+      "loss": 2.0486,
+      "step": 6760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005989564515895542,
+      "loss": 2.0912,
+      "step": 6761
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005988499675562902,
+      "loss": 2.113,
+      "step": 6762
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005987434788570806,
+      "loss": 2.1601,
+      "step": 6763
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005986369854969517,
+      "loss": 2.0766,
+      "step": 6764
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005985304874809302,
+      "loss": 1.8474,
+      "step": 6765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005984239848140431,
+      "loss": 2.1492,
+      "step": 6766
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005983174775013176,
+      "loss": 2.1747,
+      "step": 6767
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000598210965547781,
+      "loss": 2.0349,
+      "step": 6768
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005981044489584611,
+      "loss": 2.2486,
+      "step": 6769
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005979979277383855,
+      "loss": 2.1229,
+      "step": 6770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005978914018925825,
+      "loss": 2.0375,
+      "step": 6771
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00059778487142608,
+      "loss": 2.0236,
+      "step": 6772
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005976783363439069,
+      "loss": 2.1228,
+      "step": 6773
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005975717966510915,
+      "loss": 2.1988,
+      "step": 6774
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000597465252352663,
+      "loss": 2.1325,
+      "step": 6775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005973587034536505,
+      "loss": 2.2339,
+      "step": 6776
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005972521499590833,
+      "loss": 2.0613,
+      "step": 6777
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005971455918739909,
+      "loss": 2.0463,
+      "step": 6778
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005970390292034031,
+      "loss": 2.0018,
+      "step": 6779
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005969324619523501,
+      "loss": 2.0531,
+      "step": 6780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000596825890125862,
+      "loss": 2.1111,
+      "step": 6781
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005967193137289691,
+      "loss": 1.9533,
+      "step": 6782
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005966127327667023,
+      "loss": 2.058,
+      "step": 6783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005965061472440923,
+      "loss": 2.0901,
+      "step": 6784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005963995571661701,
+      "loss": 2.0089,
+      "step": 6785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005962929625379672,
+      "loss": 2.0041,
+      "step": 6786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005961863633645151,
+      "loss": 2.1096,
+      "step": 6787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005960797596508453,
+      "loss": 2.0574,
+      "step": 6788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00059597315140199,
+      "loss": 2.0519,
+      "step": 6789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005958665386229811,
+      "loss": 2.1078,
+      "step": 6790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005957599213188514,
+      "loss": 2.0638,
+      "step": 6791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005956532994946329,
+      "loss": 2.0512,
+      "step": 6792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005955466731553589,
+      "loss": 2.0222,
+      "step": 6793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005954400423060621,
+      "loss": 2.2322,
+      "step": 6794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005953334069517759,
+      "loss": 2.2053,
+      "step": 6795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005952267670975335,
+      "loss": 2.1544,
+      "step": 6796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005951201227483686,
+      "loss": 2.0002,
+      "step": 6797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005950134739093156,
+      "loss": 2.1459,
+      "step": 6798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005949068205854078,
+      "loss": 2.0004,
+      "step": 6799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005948001627816799,
+      "loss": 2.172,
+      "step": 6800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005946935005031662,
+      "loss": 2.0371,
+      "step": 6801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005945868337549017,
+      "loss": 2.1179,
+      "step": 6802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005944801625419211,
+      "loss": 2.3375,
+      "step": 6803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005943734868692595,
+      "loss": 2.064,
+      "step": 6804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005942668067419523,
+      "loss": 2.1184,
+      "step": 6805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005941601221650351,
+      "loss": 2.0601,
+      "step": 6806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005940534331435435,
+      "loss": 1.9691,
+      "step": 6807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005939467396825136,
+      "loss": 2.0466,
+      "step": 6808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005938400417869815,
+      "loss": 2.1845,
+      "step": 6809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005937333394619837,
+      "loss": 2.0923,
+      "step": 6810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005936266327125567,
+      "loss": 2.0463,
+      "step": 6811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005935199215437374,
+      "loss": 2.0607,
+      "step": 6812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005934132059605626,
+      "loss": 2.0368,
+      "step": 6813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005933064859680696,
+      "loss": 2.0496,
+      "step": 6814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000593199761571296,
+      "loss": 2.1157,
+      "step": 6815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005930930327752794,
+      "loss": 2.1073,
+      "step": 6816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005929862995850575,
+      "loss": 1.8432,
+      "step": 6817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005928795620056684,
+      "loss": 2.0321,
+      "step": 6818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005927728200421504,
+      "loss": 2.1921,
+      "step": 6819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000592666073699542,
+      "loss": 2.1559,
+      "step": 6820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005925593229828816,
+      "loss": 2.0478,
+      "step": 6821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005924525678972086,
+      "loss": 2.1764,
+      "step": 6822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005923458084475616,
+      "loss": 2.2385,
+      "step": 6823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005922390446389802,
+      "loss": 2.0144,
+      "step": 6824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005921322764765036,
+      "loss": 2.1312,
+      "step": 6825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005920255039651718,
+      "loss": 2.0359,
+      "step": 6826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005919187271100245,
+      "loss": 2.0704,
+      "step": 6827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000591811945916102,
+      "loss": 2.0859,
+      "step": 6828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005917051603884445,
+      "loss": 1.9447,
+      "step": 6829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005915983705320926,
+      "loss": 2.06,
+      "step": 6830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005914915763520867,
+      "loss": 2.0998,
+      "step": 6831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005913847778534682,
+      "loss": 2.1204,
+      "step": 6832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000591277975041278,
+      "loss": 2.1921,
+      "step": 6833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005911711679205575,
+      "loss": 2.033,
+      "step": 6834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000591064356496348,
+      "loss": 2.1478,
+      "step": 6835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005909575407736918,
+      "loss": 2.0923,
+      "step": 6836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005908507207576302,
+      "loss": 2.0478,
+      "step": 6837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000590743896453206,
+      "loss": 2.25,
+      "step": 6838
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000590637067865461,
+      "loss": 2.1727,
+      "step": 6839
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005905302349994378,
+      "loss": 2.0495,
+      "step": 6840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005904233978601796,
+      "loss": 2.0609,
+      "step": 6841
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005903165564527289,
+      "loss": 2.1927,
+      "step": 6842
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005902097107821291,
+      "loss": 2.1192,
+      "step": 6843
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005901028608534235,
+      "loss": 2.0162,
+      "step": 6844
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005899960066716555,
+      "loss": 2.1475,
+      "step": 6845
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005898891482418692,
+      "loss": 2.1034,
+      "step": 6846
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005897822855691084,
+      "loss": 2.0268,
+      "step": 6847
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005896754186584172,
+      "loss": 2.1437,
+      "step": 6848
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005895685475148401,
+      "loss": 2.1877,
+      "step": 6849
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005894616721434214,
+      "loss": 2.0762,
+      "step": 6850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005893547925492061,
+      "loss": 2.0515,
+      "step": 6851
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005892479087372392,
+      "loss": 2.049,
+      "step": 6852
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005891410207125657,
+      "loss": 2.1381,
+      "step": 6853
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005890341284802311,
+      "loss": 2.118,
+      "step": 6854
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005889272320452807,
+      "loss": 2.2059,
+      "step": 6855
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005888203314127607,
+      "loss": 2.2065,
+      "step": 6856
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005887134265877166,
+      "loss": 2.0327,
+      "step": 6857
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005886065175751949,
+      "loss": 2.1218,
+      "step": 6858
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005884996043802416,
+      "loss": 2.0618,
+      "step": 6859
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005883926870079037,
+      "loss": 1.9991,
+      "step": 6860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005882857654632275,
+      "loss": 2.1199,
+      "step": 6861
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005881788397512602,
+      "loss": 2.0371,
+      "step": 6862
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005880719098770489,
+      "loss": 2.0117,
+      "step": 6863
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000587964975845641,
+      "loss": 2.1687,
+      "step": 6864
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005878580376620838,
+      "loss": 1.9883,
+      "step": 6865
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005877510953314253,
+      "loss": 1.8932,
+      "step": 6866
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005876441488587132,
+      "loss": 2.1339,
+      "step": 6867
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005875371982489958,
+      "loss": 2.0946,
+      "step": 6868
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005874302435073213,
+      "loss": 2.0794,
+      "step": 6869
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005873232846387383,
+      "loss": 2.1216,
+      "step": 6870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005872163216482953,
+      "loss": 2.0999,
+      "step": 6871
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005871093545410415,
+      "loss": 2.0311,
+      "step": 6872
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005870023833220258,
+      "loss": 1.9421,
+      "step": 6873
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005868954079962976,
+      "loss": 2.1157,
+      "step": 6874
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005867884285689061,
+      "loss": 2.0311,
+      "step": 6875
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005866814450449014,
+      "loss": 2.0474,
+      "step": 6876
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000586574457429333,
+      "loss": 2.0467,
+      "step": 6877
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005864674657272513,
+      "loss": 2.0251,
+      "step": 6878
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005863604699437062,
+      "loss": 2.0726,
+      "step": 6879
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005862534700837484,
+      "loss": 2.0322,
+      "step": 6880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005861464661524284,
+      "loss": 2.2296,
+      "step": 6881
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005860394581547972,
+      "loss": 2.1297,
+      "step": 6882
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005859324460959056,
+      "loss": 2.0479,
+      "step": 6883
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000585825429980805,
+      "loss": 2.1847,
+      "step": 6884
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005857184098145467,
+      "loss": 1.9432,
+      "step": 6885
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005856113856021823,
+      "loss": 1.9923,
+      "step": 6886
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005855043573487636,
+      "loss": 2.1345,
+      "step": 6887
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005853973250593426,
+      "loss": 2.1167,
+      "step": 6888
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005852902887389715,
+      "loss": 2.1518,
+      "step": 6889
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005851832483927028,
+      "loss": 2.0925,
+      "step": 6890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005850762040255885,
+      "loss": 2.0785,
+      "step": 6891
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005849691556426822,
+      "loss": 2.1848,
+      "step": 6892
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000584862103249036,
+      "loss": 2.0936,
+      "step": 6893
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005847550468497036,
+      "loss": 2.151,
+      "step": 6894
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005846479864497379,
+      "loss": 2.1095,
+      "step": 6895
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005845409220541926,
+      "loss": 2.1073,
+      "step": 6896
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005844338536681214,
+      "loss": 2.0684,
+      "step": 6897
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005843267812965782,
+      "loss": 2.0334,
+      "step": 6898
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000584219704944617,
+      "loss": 2.0228,
+      "step": 6899
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005841126246172919,
+      "loss": 2.0382,
+      "step": 6900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005840055403196576,
+      "loss": 1.9799,
+      "step": 6901
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005838984520567687,
+      "loss": 2.1188,
+      "step": 6902
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005837913598336797,
+      "loss": 1.9677,
+      "step": 6903
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000583684263655446,
+      "loss": 2.0307,
+      "step": 6904
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005835771635271226,
+      "loss": 2.0847,
+      "step": 6905
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005834700594537647,
+      "loss": 2.0625,
+      "step": 6906
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005833629514404282,
+      "loss": 1.9854,
+      "step": 6907
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005832558394921688,
+      "loss": 2.2454,
+      "step": 6908
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005831487236140421,
+      "loss": 2.1055,
+      "step": 6909
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005830416038111045,
+      "loss": 1.9261,
+      "step": 6910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005829344800884121,
+      "loss": 1.8463,
+      "step": 6911
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005828273524510217,
+      "loss": 2.0604,
+      "step": 6912
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005827202209039897,
+      "loss": 1.9581,
+      "step": 6913
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005826130854523731,
+      "loss": 2.2266,
+      "step": 6914
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005825059461012289,
+      "loss": 2.0318,
+      "step": 6915
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005823988028556143,
+      "loss": 2.0174,
+      "step": 6916
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005822916557205866,
+      "loss": 1.9956,
+      "step": 6917
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005821845047012037,
+      "loss": 2.0385,
+      "step": 6918
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005820773498025231,
+      "loss": 1.909,
+      "step": 6919
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005819701910296028,
+      "loss": 2.0867,
+      "step": 6920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000581863028387501,
+      "loss": 2.125,
+      "step": 6921
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000581755861881276,
+      "loss": 2.0889,
+      "step": 6922
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005816486915159862,
+      "loss": 2.1076,
+      "step": 6923
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005815415172966906,
+      "loss": 2.227,
+      "step": 6924
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005814343392284477,
+      "loss": 2.1765,
+      "step": 6925
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005813271573163168,
+      "loss": 2.026,
+      "step": 6926
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000581219971565357,
+      "loss": 2.2755,
+      "step": 6927
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005811127819806277,
+      "loss": 2.1796,
+      "step": 6928
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005810055885671887,
+      "loss": 2.0855,
+      "step": 6929
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005808983913300994,
+      "loss": 1.9297,
+      "step": 6930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005807911902744203,
+      "loss": 2.0558,
+      "step": 6931
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000580683985405211,
+      "loss": 2.0514,
+      "step": 6932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005805767767275323,
+      "loss": 2.1074,
+      "step": 6933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005804695642464442,
+      "loss": 1.9518,
+      "step": 6934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005803623479670077,
+      "loss": 2.0126,
+      "step": 6935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005802551278942835,
+      "loss": 2.0355,
+      "step": 6936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005801479040333328,
+      "loss": 1.9766,
+      "step": 6937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005800406763892166,
+      "loss": 2.0314,
+      "step": 6938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005799334449669967,
+      "loss": 2.1359,
+      "step": 6939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005798262097717339,
+      "loss": 2.1412,
+      "step": 6940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005797189708084908,
+      "loss": 2.253,
+      "step": 6941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005796117280823288,
+      "loss": 2.1174,
+      "step": 6942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005795044815983103,
+      "loss": 2.1489,
+      "step": 6943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005793972313614973,
+      "loss": 2.1912,
+      "step": 6944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005792899773769525,
+      "loss": 2.1272,
+      "step": 6945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005791827196497382,
+      "loss": 2.1385,
+      "step": 6946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005790754581849176,
+      "loss": 2.2806,
+      "step": 6947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005789681929875535,
+      "loss": 1.9953,
+      "step": 6948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005788609240627091,
+      "loss": 2.203,
+      "step": 6949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005787536514154476,
+      "loss": 2.0932,
+      "step": 6950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005786463750508329,
+      "loss": 1.997,
+      "step": 6951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005785390949739282,
+      "loss": 2.0223,
+      "step": 6952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005784318111897979,
+      "loss": 2.1488,
+      "step": 6953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005783245237035054,
+      "loss": 1.938,
+      "step": 6954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005782172325201155,
+      "loss": 2.1798,
+      "step": 6955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005781099376446922,
+      "loss": 1.9967,
+      "step": 6956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005780026390823003,
+      "loss": 2.0098,
+      "step": 6957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005778953368380045,
+      "loss": 2.1579,
+      "step": 6958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005777880309168696,
+      "loss": 1.9515,
+      "step": 6959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005776807213239606,
+      "loss": 2.1191,
+      "step": 6960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005775734080643432,
+      "loss": 2.0636,
+      "step": 6961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005774660911430823,
+      "loss": 2.0989,
+      "step": 6962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000577358770565244,
+      "loss": 1.9743,
+      "step": 6963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005772514463358937,
+      "loss": 2.0942,
+      "step": 6964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005771441184600974,
+      "loss": 2.1565,
+      "step": 6965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005770367869429214,
+      "loss": 2.0593,
+      "step": 6966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005769294517894319,
+      "loss": 2.0007,
+      "step": 6967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005768221130046953,
+      "loss": 2.0854,
+      "step": 6968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005767147705937784,
+      "loss": 2.1001,
+      "step": 6969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005766074245617478,
+      "loss": 1.9011,
+      "step": 6970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005765000749136706,
+      "loss": 1.9485,
+      "step": 6971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000576392721654614,
+      "loss": 2.1545,
+      "step": 6972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005762853647896454,
+      "loss": 2.0164,
+      "step": 6973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005761780043238319,
+      "loss": 2.0519,
+      "step": 6974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005760706402622415,
+      "loss": 2.1045,
+      "step": 6975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005759632726099419,
+      "loss": 2.1251,
+      "step": 6976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005758559013720011,
+      "loss": 2.1175,
+      "step": 6977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005757485265534874,
+      "loss": 2.1723,
+      "step": 6978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005756411481594691,
+      "loss": 2.0246,
+      "step": 6979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005755337661950146,
+      "loss": 2.0477,
+      "step": 6980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005754263806651927,
+      "loss": 2.1953,
+      "step": 6981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005753189915750721,
+      "loss": 1.9358,
+      "step": 6982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005752115989297219,
+      "loss": 2.051,
+      "step": 6983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005751042027342111,
+      "loss": 2.1057,
+      "step": 6984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005749968029936095,
+      "loss": 2.1145,
+      "step": 6985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005748893997129861,
+      "loss": 2.2182,
+      "step": 6986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005747819928974111,
+      "loss": 1.9648,
+      "step": 6987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005746745825519538,
+      "loss": 2.2084,
+      "step": 6988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005745671686816846,
+      "loss": 2.1095,
+      "step": 6989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005744597512916734,
+      "loss": 1.9094,
+      "step": 6990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000574352330386991,
+      "loss": 1.954,
+      "step": 6991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005742449059727072,
+      "loss": 2.0709,
+      "step": 6992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005741374780538935,
+      "loss": 2.1772,
+      "step": 6993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005740300466356202,
+      "loss": 2.0895,
+      "step": 6994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005739226117229585,
+      "loss": 2.0277,
+      "step": 6995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005738151733209794,
+      "loss": 2.2636,
+      "step": 6996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005737077314347545,
+      "loss": 2.0115,
+      "step": 6997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005736002860693551,
+      "loss": 2.1077,
+      "step": 6998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005734928372298529,
+      "loss": 2.1202,
+      "step": 6999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005733853849213199,
+      "loss": 2.1613,
+      "step": 7000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005732779291488279,
+      "loss": 2.0637,
+      "step": 7001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000573170469917449,
+      "loss": 2.2308,
+      "step": 7002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005730630072322559,
+      "loss": 2.1578,
+      "step": 7003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005729555410983207,
+      "loss": 2.0778,
+      "step": 7004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005728480715207163,
+      "loss": 2.1059,
+      "step": 7005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005727405985045151,
+      "loss": 1.9107,
+      "step": 7006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005726331220547907,
+      "loss": 2.21,
+      "step": 7007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005725256421766158,
+      "loss": 2.2383,
+      "step": 7008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005724181588750638,
+      "loss": 1.9397,
+      "step": 7009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000572310672155208,
+      "loss": 2.0258,
+      "step": 7010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005722031820221224,
+      "loss": 1.9072,
+      "step": 7011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005720956884808805,
+      "loss": 1.991,
+      "step": 7012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005719881915365563,
+      "loss": 2.2824,
+      "step": 7013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005718806911942238,
+      "loss": 2.1321,
+      "step": 7014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005717731874589576,
+      "loss": 1.9561,
+      "step": 7015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005716656803358318,
+      "loss": 2.0814,
+      "step": 7016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005715581698299211,
+      "loss": 2.1341,
+      "step": 7017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005714506559463,
+      "loss": 2.1267,
+      "step": 7018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005713431386900438,
+      "loss": 2.1453,
+      "step": 7019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005712356180662274,
+      "loss": 2.0946,
+      "step": 7020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005711280940799261,
+      "loss": 2.0298,
+      "step": 7021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000571020566736215,
+      "loss": 2.1181,
+      "step": 7022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00057091303604017,
+      "loss": 2.1968,
+      "step": 7023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005708055019968666,
+      "loss": 2.0095,
+      "step": 7024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005706979646113806,
+      "loss": 1.9413,
+      "step": 7025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005705904238887882,
+      "loss": 2.0406,
+      "step": 7026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005704828798341652,
+      "loss": 2.1931,
+      "step": 7027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005703753324525885,
+      "loss": 2.1256,
+      "step": 7028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005702677817491339,
+      "loss": 1.9746,
+      "step": 7029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005701602277288787,
+      "loss": 1.8596,
+      "step": 7030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005700526703968994,
+      "loss": 1.9706,
+      "step": 7031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005699451097582729,
+      "loss": 1.9631,
+      "step": 7032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005698375458180762,
+      "loss": 2.2398,
+      "step": 7033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000569729978581387,
+      "loss": 2.1073,
+      "step": 7034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005696224080532823,
+      "loss": 2.0799,
+      "step": 7035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00056951483423884,
+      "loss": 2.0491,
+      "step": 7036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005694072571431375,
+      "loss": 2.1258,
+      "step": 7037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005692996767712529,
+      "loss": 2.0566,
+      "step": 7038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005691920931282642,
+      "loss": 2.0273,
+      "step": 7039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005690845062192495,
+      "loss": 1.99,
+      "step": 7040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005689769160492872,
+      "loss": 2.0719,
+      "step": 7041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000568869322623456,
+      "loss": 2.2192,
+      "step": 7042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005687617259468343,
+      "loss": 2.1041,
+      "step": 7043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005686541260245012,
+      "loss": 2.124,
+      "step": 7044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005685465228615352,
+      "loss": 2.135,
+      "step": 7045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005684389164630158,
+      "loss": 1.9314,
+      "step": 7046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005683313068340221,
+      "loss": 1.9607,
+      "step": 7047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005682236939796337,
+      "loss": 2.1323,
+      "step": 7048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005681160779049299,
+      "loss": 2.0894,
+      "step": 7049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005680084586149908,
+      "loss": 2.0528,
+      "step": 7050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005679008361148958,
+      "loss": 2.0839,
+      "step": 7051
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005677932104097254,
+      "loss": 2.0212,
+      "step": 7052
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005676855815045594,
+      "loss": 2.0964,
+      "step": 7053
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005675779494044784,
+      "loss": 1.945,
+      "step": 7054
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005674703141145627,
+      "loss": 2.0127,
+      "step": 7055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005673626756398931,
+      "loss": 2.1436,
+      "step": 7056
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005672550339855502,
+      "loss": 2.1234,
+      "step": 7057
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005671473891566152,
+      "loss": 2.1835,
+      "step": 7058
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005670397411581688,
+      "loss": 1.9554,
+      "step": 7059
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005669320899952927,
+      "loss": 2.1327,
+      "step": 7060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005668244356730678,
+      "loss": 2.1232,
+      "step": 7061
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000566716778196576,
+      "loss": 2.1591,
+      "step": 7062
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005666091175708987,
+      "loss": 2.0776,
+      "step": 7063
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005665014538011181,
+      "loss": 2.1209,
+      "step": 7064
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005663937868923157,
+      "loss": 1.9061,
+      "step": 7065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000566286116849574,
+      "loss": 2.1389,
+      "step": 7066
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005661784436779751,
+      "loss": 2.1133,
+      "step": 7067
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005660707673826016,
+      "loss": 2.1184,
+      "step": 7068
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005659630879685357,
+      "loss": 1.972,
+      "step": 7069
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005658554054408606,
+      "loss": 2.0815,
+      "step": 7070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005657477198046588,
+      "loss": 2.1347,
+      "step": 7071
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005656400310650134,
+      "loss": 2.0126,
+      "step": 7072
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005655323392270075,
+      "loss": 2.0586,
+      "step": 7073
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005654246442957248,
+      "loss": 1.9536,
+      "step": 7074
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005653169462762482,
+      "loss": 2.0064,
+      "step": 7075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005652092451736615,
+      "loss": 2.0401,
+      "step": 7076
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005651015409930486,
+      "loss": 2.2378,
+      "step": 7077
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005649938337394932,
+      "loss": 2.0965,
+      "step": 7078
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005648861234180794,
+      "loss": 1.995,
+      "step": 7079
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005647784100338913,
+      "loss": 2.0534,
+      "step": 7080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005646706935920132,
+      "loss": 2.0174,
+      "step": 7081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005645629740975299,
+      "loss": 2.105,
+      "step": 7082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005644552515555255,
+      "loss": 2.112,
+      "step": 7083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000564347525971085,
+      "loss": 2.086,
+      "step": 7084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005642397973492934,
+      "loss": 2.1778,
+      "step": 7085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005641320656952358,
+      "loss": 1.9827,
+      "step": 7086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000564024331013997,
+      "loss": 2.0986,
+      "step": 7087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005639165933106626,
+      "loss": 2.0124,
+      "step": 7088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005638088525903179,
+      "loss": 2.0622,
+      "step": 7089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005637011088580488,
+      "loss": 2.1165,
+      "step": 7090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005635933621189408,
+      "loss": 2.1999,
+      "step": 7091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005634856123780798,
+      "loss": 2.1236,
+      "step": 7092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005633778596405519,
+      "loss": 2.0834,
+      "step": 7093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005632701039114434,
+      "loss": 2.0931,
+      "step": 7094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005631623451958403,
+      "loss": 2.1224,
+      "step": 7095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005630545834988294,
+      "loss": 2.1594,
+      "step": 7096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005629468188254971,
+      "loss": 2.0514,
+      "step": 7097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005628390511809304,
+      "loss": 1.8519,
+      "step": 7098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005627312805702156,
+      "loss": 1.9855,
+      "step": 7099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005626235069984405,
+      "loss": 2.0952,
+      "step": 7100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005625157304706915,
+      "loss": 2.1033,
+      "step": 7101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005624079509920565,
+      "loss": 1.9439,
+      "step": 7102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005623001685676226,
+      "loss": 2.1667,
+      "step": 7103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005621923832024774,
+      "loss": 2.023,
+      "step": 7104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005620845949017088,
+      "loss": 2.1338,
+      "step": 7105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005619768036704045,
+      "loss": 2.1389,
+      "step": 7106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005618690095136524,
+      "loss": 2.1315,
+      "step": 7107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005617612124365409,
+      "loss": 2.1713,
+      "step": 7108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005616534124441582,
+      "loss": 2.1159,
+      "step": 7109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005615456095415926,
+      "loss": 2.1721,
+      "step": 7110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005614378037339328,
+      "loss": 2.0364,
+      "step": 7111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005613299950262673,
+      "loss": 2.1085,
+      "step": 7112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000561222183423685,
+      "loss": 1.9346,
+      "step": 7113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005611143689312748,
+      "loss": 2.0875,
+      "step": 7114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005610065515541258,
+      "loss": 2.097,
+      "step": 7115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005608987312973274,
+      "loss": 2.1256,
+      "step": 7116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005607909081659687,
+      "loss": 2.188,
+      "step": 7117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005606830821651396,
+      "loss": 2.0588,
+      "step": 7118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005605752532999291,
+      "loss": 2.1154,
+      "step": 7119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005604674215754276,
+      "loss": 2.1199,
+      "step": 7120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005603595869967246,
+      "loss": 1.9488,
+      "step": 7121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005602517495689104,
+      "loss": 2.0724,
+      "step": 7122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005601439092970751,
+      "loss": 2.0506,
+      "step": 7123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005600360661863089,
+      "loss": 2.0739,
+      "step": 7124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005599282202417022,
+      "loss": 1.9312,
+      "step": 7125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005598203714683458,
+      "loss": 2.0482,
+      "step": 7126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005597125198713303,
+      "loss": 2.0064,
+      "step": 7127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005596046654557466,
+      "loss": 1.9438,
+      "step": 7128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005594968082266855,
+      "loss": 2.0598,
+      "step": 7129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005593889481892384,
+      "loss": 2.2171,
+      "step": 7130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005592810853484962,
+      "loss": 2.0059,
+      "step": 7131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005591732197095505,
+      "loss": 2.1389,
+      "step": 7132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005590653512774927,
+      "loss": 2.0635,
+      "step": 7133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005589574800574146,
+      "loss": 2.0381,
+      "step": 7134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005588496060544077,
+      "loss": 1.9951,
+      "step": 7135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005587417292735642,
+      "loss": 1.9488,
+      "step": 7136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000558633849719976,
+      "loss": 2.0127,
+      "step": 7137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005585259673987352,
+      "loss": 2.2262,
+      "step": 7138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000558418082314934,
+      "loss": 2.0668,
+      "step": 7139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005583101944736653,
+      "loss": 2.0035,
+      "step": 7140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005582023038800211,
+      "loss": 2.1435,
+      "step": 7141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005580944105390943,
+      "loss": 1.9481,
+      "step": 7142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005579865144559778,
+      "loss": 2.0052,
+      "step": 7143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005578786156357646,
+      "loss": 2.2249,
+      "step": 7144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005577707140835474,
+      "loss": 2.0678,
+      "step": 7145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005576628098044197,
+      "loss": 2.0927,
+      "step": 7146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005575549028034749,
+      "loss": 1.9657,
+      "step": 7147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005574469930858062,
+      "loss": 2.1309,
+      "step": 7148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005573390806565073,
+      "loss": 2.2057,
+      "step": 7149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000557231165520672,
+      "loss": 2.1344,
+      "step": 7150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005571232476833941,
+      "loss": 2.1451,
+      "step": 7151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005570153271497675,
+      "loss": 2.1081,
+      "step": 7152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005569074039248864,
+      "loss": 2.1472,
+      "step": 7153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005567994780138448,
+      "loss": 2.1226,
+      "step": 7154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005566915494217374,
+      "loss": 2.179,
+      "step": 7155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005565836181536583,
+      "loss": 2.1451,
+      "step": 7156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005564756842147025,
+      "loss": 2.0272,
+      "step": 7157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005563677476099645,
+      "loss": 1.999,
+      "step": 7158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005562598083445391,
+      "loss": 2.1933,
+      "step": 7159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005561518664235213,
+      "loss": 2.0266,
+      "step": 7160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005560439218520065,
+      "loss": 1.9955,
+      "step": 7161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005559359746350896,
+      "loss": 2.0331,
+      "step": 7162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005558280247778661,
+      "loss": 2.162,
+      "step": 7163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005557200722854313,
+      "loss": 2.058,
+      "step": 7164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005556121171628811,
+      "loss": 2.0726,
+      "step": 7165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000555504159415311,
+      "loss": 2.1967,
+      "step": 7166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005553961990478171,
+      "loss": 2.0503,
+      "step": 7167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000555288236065495,
+      "loss": 2.0501,
+      "step": 7168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000555180270473441,
+      "loss": 1.9963,
+      "step": 7169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005550723022767515,
+      "loss": 2.1117,
+      "step": 7170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005549643314805226,
+      "loss": 2.0506,
+      "step": 7171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005548563580898508,
+      "loss": 1.9789,
+      "step": 7172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005547483821098328,
+      "loss": 2.0488,
+      "step": 7173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005546404035455653,
+      "loss": 2.1248,
+      "step": 7174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005545324224021451,
+      "loss": 2.1428,
+      "step": 7175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005544244386846691,
+      "loss": 2.04,
+      "step": 7176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005543164523982344,
+      "loss": 1.997,
+      "step": 7177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005542084635479383,
+      "loss": 1.9498,
+      "step": 7178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005541004721388781,
+      "loss": 1.9843,
+      "step": 7179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000553992478176151,
+      "loss": 2.1082,
+      "step": 7180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005538844816648551,
+      "loss": 2.0493,
+      "step": 7181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005537764826100875,
+      "loss": 2.0101,
+      "step": 7182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005536684810169463,
+      "loss": 2.1198,
+      "step": 7183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005535604768905294,
+      "loss": 2.0676,
+      "step": 7184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005534524702359348,
+      "loss": 2.0123,
+      "step": 7185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005533444610582607,
+      "loss": 2.0723,
+      "step": 7186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005532364493626054,
+      "loss": 2.06,
+      "step": 7187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005531284351540671,
+      "loss": 2.1833,
+      "step": 7188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005530204184377447,
+      "loss": 2.1408,
+      "step": 7189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005529123992187364,
+      "loss": 2.1808,
+      "step": 7190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005528043775021414,
+      "loss": 2.1274,
+      "step": 7191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005526963532930582,
+      "loss": 1.8989,
+      "step": 7192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005525883265965861,
+      "loss": 2.0159,
+      "step": 7193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005524802974178239,
+      "loss": 1.9217,
+      "step": 7194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000552372265761871,
+      "loss": 2.0809,
+      "step": 7195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005522642316338268,
+      "loss": 1.9551,
+      "step": 7196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005521561950387905,
+      "loss": 2.0162,
+      "step": 7197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005520481559818619,
+      "loss": 2.0706,
+      "step": 7198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005519401144681408,
+      "loss": 1.9181,
+      "step": 7199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005518320705027267,
+      "loss": 1.9408,
+      "step": 7200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005517240240907198,
+      "loss": 1.9495,
+      "step": 7201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00055161597523722,
+      "loss": 2.1097,
+      "step": 7202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005515079239473274,
+      "loss": 2.076,
+      "step": 7203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005513998702261424,
+      "loss": 2.0285,
+      "step": 7204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005512918140787653,
+      "loss": 1.9965,
+      "step": 7205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005511837555102966,
+      "loss": 2.0333,
+      "step": 7206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005510756945258368,
+      "loss": 2.3058,
+      "step": 7207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005509676311304868,
+      "loss": 1.9879,
+      "step": 7208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005508595653293477,
+      "loss": 2.1603,
+      "step": 7209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005507514971275199,
+      "loss": 1.997,
+      "step": 7210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005506434265301048,
+      "loss": 2.1083,
+      "step": 7211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005505353535422034,
+      "loss": 2.0764,
+      "step": 7212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005504272781689173,
+      "loss": 2.0551,
+      "step": 7213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005503192004153475,
+      "loss": 2.1291,
+      "step": 7214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005502111202865958,
+      "loss": 2.0679,
+      "step": 7215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005501030377877637,
+      "loss": 1.9812,
+      "step": 7216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005499949529239531,
+      "loss": 2.1898,
+      "step": 7217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005498868657002657,
+      "loss": 2.0456,
+      "step": 7218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005497787761218034,
+      "loss": 2.0585,
+      "step": 7219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005496706841936686,
+      "loss": 2.2544,
+      "step": 7220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005495625899209632,
+      "loss": 2.1606,
+      "step": 7221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005494544933087895,
+      "loss": 2.2218,
+      "step": 7222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005493463943622502,
+      "loss": 1.9147,
+      "step": 7223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005492382930864473,
+      "loss": 2.1918,
+      "step": 7224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000549130189486484,
+      "loss": 2.1269,
+      "step": 7225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005490220835674627,
+      "loss": 2.2484,
+      "step": 7226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005489139753344863,
+      "loss": 2.1525,
+      "step": 7227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005488058647926577,
+      "loss": 1.826,
+      "step": 7228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005486977519470801,
+      "loss": 2.1077,
+      "step": 7229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005485896368028566,
+      "loss": 2.0351,
+      "step": 7230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005484815193650906,
+      "loss": 2.0452,
+      "step": 7231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005483733996388852,
+      "loss": 1.9584,
+      "step": 7232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005482652776293444,
+      "loss": 2.1227,
+      "step": 7233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005481571533415713,
+      "loss": 2.2004,
+      "step": 7234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005480490267806698,
+      "loss": 2.0215,
+      "step": 7235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005479408979517437,
+      "loss": 2.0401,
+      "step": 7236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005478327668598971,
+      "loss": 2.2084,
+      "step": 7237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005477246335102338,
+      "loss": 2.0452,
+      "step": 7238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005476164979078581,
+      "loss": 1.9982,
+      "step": 7239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005475083600578742,
+      "loss": 2.1224,
+      "step": 7240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005474002199653864,
+      "loss": 1.9837,
+      "step": 7241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005472920776354993,
+      "loss": 1.8988,
+      "step": 7242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005471839330733173,
+      "loss": 2.1042,
+      "step": 7243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005470757862839452,
+      "loss": 2.0869,
+      "step": 7244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005469676372724877,
+      "loss": 2.0627,
+      "step": 7245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005468594860440498,
+      "loss": 2.2069,
+      "step": 7246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005467513326037363,
+      "loss": 2.0094,
+      "step": 7247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005466431769566524,
+      "loss": 2.0512,
+      "step": 7248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005465350191079033,
+      "loss": 2.0912,
+      "step": 7249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005464268590625943,
+      "loss": 2.0781,
+      "step": 7250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005463186968258307,
+      "loss": 1.9495,
+      "step": 7251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005462105324027183,
+      "loss": 2.1396,
+      "step": 7252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005461023657983622,
+      "loss": 2.1597,
+      "step": 7253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005459941970178687,
+      "loss": 2.1307,
+      "step": 7254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005458860260663431,
+      "loss": 2.1103,
+      "step": 7255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005457778529488917,
+      "loss": 2.193,
+      "step": 7256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005456696776706203,
+      "loss": 2.0907,
+      "step": 7257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005455615002366351,
+      "loss": 2.1393,
+      "step": 7258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005454533206520422,
+      "loss": 2.1278,
+      "step": 7259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005453451389219483,
+      "loss": 2.0454,
+      "step": 7260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005452369550514593,
+      "loss": 1.9818,
+      "step": 7261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005451287690456823,
+      "loss": 1.9545,
+      "step": 7262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005450205809097233,
+      "loss": 1.9771,
+      "step": 7263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005449123906486894,
+      "loss": 2.0944,
+      "step": 7264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005448041982676875,
+      "loss": 2.1593,
+      "step": 7265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005446960037718243,
+      "loss": 2.107,
+      "step": 7266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000544587807166207,
+      "loss": 2.0776,
+      "step": 7267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005444796084559428,
+      "loss": 2.2943,
+      "step": 7268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005443714076461386,
+      "loss": 2.1465,
+      "step": 7269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000544263204741902,
+      "loss": 2.0666,
+      "step": 7270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005441549997483402,
+      "loss": 2.0429,
+      "step": 7271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005440467926705611,
+      "loss": 2.0968,
+      "step": 7272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000543938583513672,
+      "loss": 2.1034,
+      "step": 7273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005438303722827807,
+      "loss": 1.9896,
+      "step": 7274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005437221589829949,
+      "loss": 2.0784,
+      "step": 7275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005436139436194228,
+      "loss": 2.1341,
+      "step": 7276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005435057261971723,
+      "loss": 2.1101,
+      "step": 7277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005433975067213514,
+      "loss": 2.0506,
+      "step": 7278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005432892851970684,
+      "loss": 2.0419,
+      "step": 7279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005431810616294316,
+      "loss": 2.0007,
+      "step": 7280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005430728360235492,
+      "loss": 1.9912,
+      "step": 7281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005429646083845302,
+      "loss": 2.0086,
+      "step": 7282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005428563787174827,
+      "loss": 2.1217,
+      "step": 7283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005427481470275157,
+      "loss": 2.0566,
+      "step": 7284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005426399133197377,
+      "loss": 1.9529,
+      "step": 7285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005425316775992579,
+      "loss": 2.0679,
+      "step": 7286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005424234398711849,
+      "loss": 2.0991,
+      "step": 7287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005423152001406282,
+      "loss": 1.995,
+      "step": 7288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005422069584126965,
+      "loss": 2.0262,
+      "step": 7289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005420987146924995,
+      "loss": 2.0717,
+      "step": 7290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005419904689851463,
+      "loss": 2.0966,
+      "step": 7291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005418822212957464,
+      "loss": 1.922,
+      "step": 7292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005417739716294093,
+      "loss": 2.0847,
+      "step": 7293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005416657199912447,
+      "loss": 2.0631,
+      "step": 7294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005415574663863622,
+      "loss": 2.0688,
+      "step": 7295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005414492108198718,
+      "loss": 1.9356,
+      "step": 7296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005413409532968833,
+      "loss": 2.3078,
+      "step": 7297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005412326938225068,
+      "loss": 1.862,
+      "step": 7298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005411244324018523,
+      "loss": 2.1503,
+      "step": 7299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005410161690400301,
+      "loss": 2.102,
+      "step": 7300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005409079037421502,
+      "loss": 2.0148,
+      "step": 7301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005407996365133234,
+      "loss": 1.9318,
+      "step": 7302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005406913673586597,
+      "loss": 1.9557,
+      "step": 7303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005405830962832701,
+      "loss": 1.8937,
+      "step": 7304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005404748232922649,
+      "loss": 2.2353,
+      "step": 7305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000540366548390755,
+      "loss": 2.1583,
+      "step": 7306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005402582715838511,
+      "loss": 1.9494,
+      "step": 7307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005401499928766643,
+      "loss": 2.0228,
+      "step": 7308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005400417122743054,
+      "loss": 2.0848,
+      "step": 7309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005399334297818859,
+      "loss": 1.9627,
+      "step": 7310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005398251454045163,
+      "loss": 2.119,
+      "step": 7311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005397168591473085,
+      "loss": 1.8756,
+      "step": 7312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005396085710153735,
+      "loss": 2.067,
+      "step": 7313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005395002810138228,
+      "loss": 2.3343,
+      "step": 7314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005393919891477681,
+      "loss": 1.9951,
+      "step": 7315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000539283695422321,
+      "loss": 2.1598,
+      "step": 7316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000539175399842593,
+      "loss": 2.1344,
+      "step": 7317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005390671024136961,
+      "loss": 1.8955,
+      "step": 7318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000538958803140742,
+      "loss": 2.0813,
+      "step": 7319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000538850502028843,
+      "loss": 2.1444,
+      "step": 7320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005387421990831108,
+      "loss": 2.1888,
+      "step": 7321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005386338943086578,
+      "loss": 2.0017,
+      "step": 7322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005385255877105961,
+      "loss": 2.2036,
+      "step": 7323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005384172792940381,
+      "loss": 2.1787,
+      "step": 7324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005383089690640961,
+      "loss": 2.3777,
+      "step": 7325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005382006570258829,
+      "loss": 2.0374,
+      "step": 7326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005380923431845105,
+      "loss": 2.144,
+      "step": 7327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005379840275450921,
+      "loss": 2.0708,
+      "step": 7328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005378757101127401,
+      "loss": 2.1566,
+      "step": 7329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005377673908925675,
+      "loss": 2.0939,
+      "step": 7330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005376590698896874,
+      "loss": 2.1398,
+      "step": 7331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005375507471092124,
+      "loss": 2.1859,
+      "step": 7332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000537442422556256,
+      "loss": 2.2833,
+      "step": 7333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005373340962359309,
+      "loss": 2.1044,
+      "step": 7334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005372257681533508,
+      "loss": 1.8418,
+      "step": 7335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005371174383136286,
+      "loss": 2.1785,
+      "step": 7336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005370091067218782,
+      "loss": 2.0669,
+      "step": 7337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005369007733832126,
+      "loss": 2.2638,
+      "step": 7338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005367924383027459,
+      "loss": 2.0259,
+      "step": 7339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005366841014855914,
+      "loss": 2.1869,
+      "step": 7340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005365757629368628,
+      "loss": 1.9661,
+      "step": 7341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005364674226616743,
+      "loss": 2.0416,
+      "step": 7342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005363590806651395,
+      "loss": 2.248,
+      "step": 7343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005362507369523724,
+      "loss": 2.026,
+      "step": 7344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005361423915284873,
+      "loss": 2.0599,
+      "step": 7345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005360340443985982,
+      "loss": 2.008,
+      "step": 7346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005359256955678192,
+      "loss": 1.976,
+      "step": 7347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005358173450412649,
+      "loss": 2.1567,
+      "step": 7348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005357089928240494,
+      "loss": 2.1779,
+      "step": 7349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005356006389212874,
+      "loss": 1.9948,
+      "step": 7350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005354922833380934,
+      "loss": 1.9025,
+      "step": 7351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000535383926079582,
+      "loss": 2.0858,
+      "step": 7352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005352755671508678,
+      "loss": 2.1705,
+      "step": 7353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005351672065570659,
+      "loss": 2.1757,
+      "step": 7354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005350588443032907,
+      "loss": 2.0284,
+      "step": 7355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005349504803946577,
+      "loss": 1.9195,
+      "step": 7356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005348421148362815,
+      "loss": 2.0547,
+      "step": 7357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005347337476332776,
+      "loss": 2.2517,
+      "step": 7358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005346253787907605,
+      "loss": 1.9468,
+      "step": 7359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005345170083138463,
+      "loss": 2.0951,
+      "step": 7360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005344086362076497,
+      "loss": 1.9243,
+      "step": 7361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005343002624772865,
+      "loss": 2.0764,
+      "step": 7362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005341918871278719,
+      "loss": 2.0814,
+      "step": 7363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005340835101645216,
+      "loss": 2.0919,
+      "step": 7364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005339751315923512,
+      "loss": 2.0074,
+      "step": 7365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005338667514164767,
+      "loss": 2.1106,
+      "step": 7366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005337583696420135,
+      "loss": 2.0857,
+      "step": 7367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005336499862740775,
+      "loss": 1.9281,
+      "step": 7368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005335416013177849,
+      "loss": 2.1201,
+      "step": 7369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005334332147782514,
+      "loss": 2.1167,
+      "step": 7370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005333248266605936,
+      "loss": 2.1246,
+      "step": 7371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000533216436969927,
+      "loss": 2.0703,
+      "step": 7372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005331080457113685,
+      "loss": 2.058,
+      "step": 7373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005329996528900338,
+      "loss": 1.9529,
+      "step": 7374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005328912585110399,
+      "loss": 2.037,
+      "step": 7375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005327828625795028,
+      "loss": 1.7805,
+      "step": 7376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005326744651005392,
+      "loss": 2.0396,
+      "step": 7377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005325660660792657,
+      "loss": 2.0037,
+      "step": 7378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005324576655207991,
+      "loss": 2.0415,
+      "step": 7379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005323492634302559,
+      "loss": 2.0689,
+      "step": 7380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005322408598127531,
+      "loss": 2.2287,
+      "step": 7381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005321324546734077,
+      "loss": 1.905,
+      "step": 7382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005320240480173364,
+      "loss": 1.8931,
+      "step": 7383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005319156398496564,
+      "loss": 2.2122,
+      "step": 7384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005318072301754849,
+      "loss": 1.921,
+      "step": 7385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005316988189999389,
+      "loss": 2.1285,
+      "step": 7386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005315904063281358,
+      "loss": 2.0387,
+      "step": 7387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005314819921651928,
+      "loss": 2.0997,
+      "step": 7388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005313735765162276,
+      "loss": 2.186,
+      "step": 7389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005312651593863571,
+      "loss": 2.2732,
+      "step": 7390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005311567407806994,
+      "loss": 2.0487,
+      "step": 7391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005310483207043718,
+      "loss": 2.0027,
+      "step": 7392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005309398991624921,
+      "loss": 2.0773,
+      "step": 7393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005308314761601779,
+      "loss": 1.9155,
+      "step": 7394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005307230517025473,
+      "loss": 1.8768,
+      "step": 7395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005306146257947179,
+      "loss": 1.9458,
+      "step": 7396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005305061984418078,
+      "loss": 2.0706,
+      "step": 7397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005303977696489349,
+      "loss": 2.2295,
+      "step": 7398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005302893394212175,
+      "loss": 1.8968,
+      "step": 7399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005301809077637734,
+      "loss": 2.1087,
+      "step": 7400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005300724746817214,
+      "loss": 2.0971,
+      "step": 7401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005299640401801792,
+      "loss": 2.0407,
+      "step": 7402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005298556042642654,
+      "loss": 2.1771,
+      "step": 7403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005297471669390984,
+      "loss": 2.0617,
+      "step": 7404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005296387282097969,
+      "loss": 2.037,
+      "step": 7405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005295302880814791,
+      "loss": 2.0086,
+      "step": 7406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005294218465592639,
+      "loss": 1.8444,
+      "step": 7407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005293134036482698,
+      "loss": 2.0159,
+      "step": 7408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005292049593536156,
+      "loss": 1.951,
+      "step": 7409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005290965136804201,
+      "loss": 2.0639,
+      "step": 7410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005289880666338025,
+      "loss": 2.1326,
+      "step": 7411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005288796182188811,
+      "loss": 2.0109,
+      "step": 7412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005287711684407758,
+      "loss": 2.0843,
+      "step": 7413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005286627173046048,
+      "loss": 2.062,
+      "step": 7414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005285542648154879,
+      "loss": 2.0467,
+      "step": 7415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005284458109785438,
+      "loss": 2.1732,
+      "step": 7416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005283373557988921,
+      "loss": 2.0956,
+      "step": 7417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000528228899281652,
+      "loss": 2.1214,
+      "step": 7418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005281204414319431,
+      "loss": 1.9456,
+      "step": 7419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005280119822548844,
+      "loss": 2.2138,
+      "step": 7420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000527903521755596,
+      "loss": 2.1304,
+      "step": 7421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005277950599391971,
+      "loss": 2.1228,
+      "step": 7422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005276865968108075,
+      "loss": 2.0437,
+      "step": 7423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005275781323755466,
+      "loss": 1.9054,
+      "step": 7424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005274696666385348,
+      "loss": 2.0894,
+      "step": 7425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005273611996048914,
+      "loss": 2.0335,
+      "step": 7426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005272527312797365,
+      "loss": 1.9994,
+      "step": 7427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005271442616681898,
+      "loss": 1.7984,
+      "step": 7428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005270357907753717,
+      "loss": 1.9941,
+      "step": 7429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005269273186064022,
+      "loss": 2.112,
+      "step": 7430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005268188451664012,
+      "loss": 2.1872,
+      "step": 7431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005267103704604891,
+      "loss": 2.1787,
+      "step": 7432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000526601894493786,
+      "loss": 2.0242,
+      "step": 7433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005264934172714125,
+      "loss": 2.1963,
+      "step": 7434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005263849387984886,
+      "loss": 2.0178,
+      "step": 7435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005262764590801349,
+      "loss": 2.049,
+      "step": 7436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000526167978121472,
+      "loss": 2.1253,
+      "step": 7437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005260594959276202,
+      "loss": 2.0119,
+      "step": 7438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005259510125037005,
+      "loss": 2.0197,
+      "step": 7439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005258425278548331,
+      "loss": 2.1406,
+      "step": 7440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005257340419861392,
+      "loss": 2.153,
+      "step": 7441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005256255549027392,
+      "loss": 2.0347,
+      "step": 7442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000525517066609754,
+      "loss": 1.9911,
+      "step": 7443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005254085771123047,
+      "loss": 2.103,
+      "step": 7444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000525300086415512,
+      "loss": 1.9785,
+      "step": 7445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005251915945244972,
+      "loss": 2.0321,
+      "step": 7446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005250831014443811,
+      "loss": 2.0248,
+      "step": 7447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005249746071802849,
+      "loss": 1.999,
+      "step": 7448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005248661117373299,
+      "loss": 2.1491,
+      "step": 7449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005247576151206372,
+      "loss": 2.0598,
+      "step": 7450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005246491173353282,
+      "loss": 2.0412,
+      "step": 7451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005245406183865242,
+      "loss": 2.2257,
+      "step": 7452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005244321182793464,
+      "loss": 2.0077,
+      "step": 7453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005243236170189165,
+      "loss": 1.9687,
+      "step": 7454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000524215114610356,
+      "loss": 1.9146,
+      "step": 7455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005241066110587863,
+      "loss": 2.2482,
+      "step": 7456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005239981063693292,
+      "loss": 2.0809,
+      "step": 7457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005238896005471063,
+      "loss": 1.9533,
+      "step": 7458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005237810935972393,
+      "loss": 2.0048,
+      "step": 7459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005236725855248499,
+      "loss": 1.9978,
+      "step": 7460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005235640763350601,
+      "loss": 1.9999,
+      "step": 7461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005234555660329918,
+      "loss": 2.044,
+      "step": 7462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005233470546237667,
+      "loss": 1.9885,
+      "step": 7463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005232385421125071,
+      "loss": 2.0295,
+      "step": 7464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005231300285043346,
+      "loss": 1.9859,
+      "step": 7465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005230215138043719,
+      "loss": 2.0705,
+      "step": 7466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005229129980177405,
+      "loss": 2.1334,
+      "step": 7467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005228044811495631,
+      "loss": 2.1646,
+      "step": 7468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005226959632049617,
+      "loss": 1.8781,
+      "step": 7469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005225874441890587,
+      "loss": 2.1112,
+      "step": 7470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005224789241069762,
+      "loss": 2.0324,
+      "step": 7471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000522370402963837,
+      "loss": 2.1186,
+      "step": 7472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005222618807647632,
+      "loss": 2.1806,
+      "step": 7473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005221533575148774,
+      "loss": 2.2032,
+      "step": 7474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005220448332193023,
+      "loss": 2.1021,
+      "step": 7475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005219363078831603,
+      "loss": 2.0239,
+      "step": 7476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005218277815115742,
+      "loss": 2.1423,
+      "step": 7477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005217192541096666,
+      "loss": 2.0237,
+      "step": 7478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005216107256825602,
+      "loss": 1.9543,
+      "step": 7479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005215021962353779,
+      "loss": 1.9783,
+      "step": 7480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005213936657732424,
+      "loss": 2.0607,
+      "step": 7481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005212851343012768,
+      "loss": 2.1022,
+      "step": 7482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005211766018246038,
+      "loss": 2.174,
+      "step": 7483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005210680683483466,
+      "loss": 2.0478,
+      "step": 7484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005209595338776281,
+      "loss": 2.0468,
+      "step": 7485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005208509984175715,
+      "loss": 1.9616,
+      "step": 7486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005207424619732996,
+      "loss": 2.2232,
+      "step": 7487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005206339245499359,
+      "loss": 2.1098,
+      "step": 7488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005205253861526035,
+      "loss": 1.9808,
+      "step": 7489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005204168467864256,
+      "loss": 2.0095,
+      "step": 7490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005203083064565257,
+      "loss": 1.9623,
+      "step": 7491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005201997651680271,
+      "loss": 2.0297,
+      "step": 7492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005200912229260528,
+      "loss": 2.0944,
+      "step": 7493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005199826797357269,
+      "loss": 1.9921,
+      "step": 7494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005198741356021724,
+      "loss": 2.1685,
+      "step": 7495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005197655905305131,
+      "loss": 1.9922,
+      "step": 7496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005196570445258723,
+      "loss": 1.9504,
+      "step": 7497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000519548497593374,
+      "loss": 1.938,
+      "step": 7498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005194399497381416,
+      "loss": 2.1707,
+      "step": 7499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000519331400965299,
+      "loss": 2.1564,
+      "step": 7500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005192228512799695,
+      "loss": 2.0763,
+      "step": 7501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005191143006872775,
+      "loss": 2.078,
+      "step": 7502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005190057491923463,
+      "loss": 2.0216,
+      "step": 7503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005188971968003003,
+      "loss": 2.1585,
+      "step": 7504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005187886435162629,
+      "loss": 2.0527,
+      "step": 7505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005186800893453585,
+      "loss": 2.1565,
+      "step": 7506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005185715342927108,
+      "loss": 1.9838,
+      "step": 7507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000518462978363444,
+      "loss": 2.224,
+      "step": 7508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005183544215626822,
+      "loss": 2.004,
+      "step": 7509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005182458638955495,
+      "loss": 2.0934,
+      "step": 7510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00051813730536717,
+      "loss": 2.2153,
+      "step": 7511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005180287459826681,
+      "loss": 2.1842,
+      "step": 7512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005179201857471678,
+      "loss": 2.112,
+      "step": 7513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005178116246657937,
+      "loss": 2.0973,
+      "step": 7514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005177030627436697,
+      "loss": 2.2104,
+      "step": 7515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005175944999859206,
+      "loss": 2.0717,
+      "step": 7516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005174859363976706,
+      "loss": 2.1053,
+      "step": 7517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005173773719840442,
+      "loss": 2.2037,
+      "step": 7518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005172688067501659,
+      "loss": 2.0792,
+      "step": 7519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005171602407011602,
+      "loss": 2.0355,
+      "step": 7520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005170516738421516,
+      "loss": 2.0115,
+      "step": 7521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005169431061782649,
+      "loss": 2.0204,
+      "step": 7522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005168345377146244,
+      "loss": 2.0976,
+      "step": 7523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005167259684563553,
+      "loss": 2.0375,
+      "step": 7524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005166173984085818,
+      "loss": 2.0859,
+      "step": 7525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000516508827576429,
+      "loss": 2.1258,
+      "step": 7526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005164002559650215,
+      "loss": 2.2602,
+      "step": 7527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0005162916835794844,
+      "loss": 2.1572,
+      "step": 7528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000516183110424942,
+      "loss": 1.9928,
+      "step": 7529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005160745365065199,
+      "loss": 2.061,
+      "step": 7530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005159659618293425,
+      "loss": 2.0932,
+      "step": 7531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005158573863985352,
+      "loss": 1.9597,
+      "step": 7532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005157488102192226,
+      "loss": 2.0683,
+      "step": 7533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005156402332965301,
+      "loss": 2.1008,
+      "step": 7534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005155316556355825,
+      "loss": 2.0689,
+      "step": 7535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005154230772415052,
+      "loss": 2.0634,
+      "step": 7536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000515314498119423,
+      "loss": 1.9311,
+      "step": 7537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005152059182744613,
+      "loss": 2.0295,
+      "step": 7538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005150973377117452,
+      "loss": 2.2241,
+      "step": 7539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005149887564364004,
+      "loss": 1.945,
+      "step": 7540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005148801744535514,
+      "loss": 2.0787,
+      "step": 7541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005147715917683243,
+      "loss": 2.0437,
+      "step": 7542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005146630083858438,
+      "loss": 2.1325,
+      "step": 7543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005145544243112357,
+      "loss": 2.1179,
+      "step": 7544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005144458395496253,
+      "loss": 2.1163,
+      "step": 7545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005143372541061382,
+      "loss": 2.1995,
+      "step": 7546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005142286679858995,
+      "loss": 2.1489,
+      "step": 7547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005141200811940352,
+      "loss": 2.0771,
+      "step": 7548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005140114937356704,
+      "loss": 2.1489,
+      "step": 7549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005139029056159309,
+      "loss": 1.9431,
+      "step": 7550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005137943168399422,
+      "loss": 2.203,
+      "step": 7551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005136857274128301,
+      "loss": 2.1474,
+      "step": 7552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005135771373397203,
+      "loss": 2.0552,
+      "step": 7553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005134685466257384,
+      "loss": 2.0751,
+      "step": 7554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005133599552760101,
+      "loss": 2.2534,
+      "step": 7555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005132513632956611,
+      "loss": 2.0913,
+      "step": 7556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005131427706898173,
+      "loss": 2.1795,
+      "step": 7557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005130341774636045,
+      "loss": 2.0718,
+      "step": 7558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005129255836221486,
+      "loss": 2.1094,
+      "step": 7559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005128169891705752,
+      "loss": 2.0085,
+      "step": 7560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005127083941140108,
+      "loss": 1.8911,
+      "step": 7561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005125997984575807,
+      "loss": 2.0565,
+      "step": 7562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005124912022064113,
+      "loss": 2.0983,
+      "step": 7563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005123826053656283,
+      "loss": 2.0676,
+      "step": 7564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005122740079403581,
+      "loss": 2.1274,
+      "step": 7565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005121654099357262,
+      "loss": 2.0959,
+      "step": 7566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000512056811356859,
+      "loss": 1.9968,
+      "step": 7567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005119482122088828,
+      "loss": 2.1397,
+      "step": 7568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005118396124969232,
+      "loss": 2.0847,
+      "step": 7569
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005117310122261069,
+      "loss": 2.0058,
+      "step": 7570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005116224114015598,
+      "loss": 2.1931,
+      "step": 7571
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005115138100284083,
+      "loss": 2.0234,
+      "step": 7572
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005114052081117782,
+      "loss": 2.067,
+      "step": 7573
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005112966056567962,
+      "loss": 2.1293,
+      "step": 7574
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005111880026685883,
+      "loss": 2.1496,
+      "step": 7575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005110793991522812,
+      "loss": 2.0572,
+      "step": 7576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005109707951130007,
+      "loss": 2.181,
+      "step": 7577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005108621905558737,
+      "loss": 2.0458,
+      "step": 7578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005107535854860262,
+      "loss": 1.8929,
+      "step": 7579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005106449799085847,
+      "loss": 1.9677,
+      "step": 7580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005105363738286757,
+      "loss": 2.073,
+      "step": 7581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005104277672514257,
+      "loss": 2.0612,
+      "step": 7582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000510319160181961,
+      "loss": 2.0693,
+      "step": 7583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005102105526254083,
+      "loss": 2.1321,
+      "step": 7584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005101019445868938,
+      "loss": 2.1736,
+      "step": 7585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005099933360715447,
+      "loss": 2.096,
+      "step": 7586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005098847270844867,
+      "loss": 2.0368,
+      "step": 7587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005097761176308471,
+      "loss": 2.031,
+      "step": 7588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005096675077157524,
+      "loss": 1.9438,
+      "step": 7589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005095588973443289,
+      "loss": 2.0877,
+      "step": 7590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005094502865217034,
+      "loss": 1.9046,
+      "step": 7591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005093416752530028,
+      "loss": 2.2363,
+      "step": 7592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005092330635433536,
+      "loss": 2.0463,
+      "step": 7593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005091244513978825,
+      "loss": 1.9555,
+      "step": 7594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005090158388217164,
+      "loss": 2.0658,
+      "step": 7595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005089072258199819,
+      "loss": 2.0317,
+      "step": 7596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000508798612397806,
+      "loss": 2.0911,
+      "step": 7597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005086899985603152,
+      "loss": 1.991,
+      "step": 7598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005085813843126365,
+      "loss": 2.206,
+      "step": 7599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005084727696598968,
+      "loss": 2.2067,
+      "step": 7600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005083641546072229,
+      "loss": 1.9879,
+      "step": 7601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005082555391597417,
+      "loss": 2.1243,
+      "step": 7602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00050814692332258,
+      "loss": 2.2614,
+      "step": 7603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005080383071008647,
+      "loss": 2.1237,
+      "step": 7604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005079296904997229,
+      "loss": 2.0684,
+      "step": 7605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005078210735242814,
+      "loss": 2.0099,
+      "step": 7606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005077124561796674,
+      "loss": 2.0173,
+      "step": 7607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005076038384710077,
+      "loss": 2.1651,
+      "step": 7608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005074952204034293,
+      "loss": 2.1113,
+      "step": 7609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005073866019820592,
+      "loss": 2.1126,
+      "step": 7610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005072779832120246,
+      "loss": 1.9564,
+      "step": 7611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005071693640984524,
+      "loss": 2.0651,
+      "step": 7612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005070607446464696,
+      "loss": 2.0729,
+      "step": 7613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005069521248612036,
+      "loss": 2.1038,
+      "step": 7614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005068435047477813,
+      "loss": 2.091,
+      "step": 7615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005067348843113298,
+      "loss": 2.0311,
+      "step": 7616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005066262635569763,
+      "loss": 2.0706,
+      "step": 7617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005065176424898478,
+      "loss": 2.1543,
+      "step": 7618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005064090211150718,
+      "loss": 1.9777,
+      "step": 7619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005063003994377751,
+      "loss": 2.066,
+      "step": 7620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005061917774630851,
+      "loss": 2.0291,
+      "step": 7621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005060831551961289,
+      "loss": 2.027,
+      "step": 7622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000505974532642034,
+      "loss": 1.9586,
+      "step": 7623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005058659098059272,
+      "loss": 2.0595,
+      "step": 7624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005057572866929361,
+      "loss": 2.14,
+      "step": 7625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005056486633081877,
+      "loss": 2.1049,
+      "step": 7626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005055400396568095,
+      "loss": 1.9953,
+      "step": 7627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005054314157439286,
+      "loss": 2.009,
+      "step": 7628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005053227915746726,
+      "loss": 2.0468,
+      "step": 7629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005052141671541684,
+      "loss": 2.06,
+      "step": 7630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005051055424875435,
+      "loss": 2.1035,
+      "step": 7631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005049969175799253,
+      "loss": 2.1128,
+      "step": 7632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005048882924364412,
+      "loss": 2.0625,
+      "step": 7633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005047796670622183,
+      "loss": 2.1257,
+      "step": 7634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005046710414623842,
+      "loss": 1.9978,
+      "step": 7635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005045624156420662,
+      "loss": 2.1114,
+      "step": 7636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005044537896063917,
+      "loss": 2.0749,
+      "step": 7637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005043451633604881,
+      "loss": 2.0446,
+      "step": 7638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005042365369094828,
+      "loss": 2.4037,
+      "step": 7639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005041279102585032,
+      "loss": 2.1555,
+      "step": 7640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005040192834126767,
+      "loss": 1.9996,
+      "step": 7641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005039106563771307,
+      "loss": 2.1789,
+      "step": 7642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005038020291569929,
+      "loss": 2.0646,
+      "step": 7643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005036934017573905,
+      "loss": 2.1195,
+      "step": 7644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005035847741834511,
+      "loss": 2.1293,
+      "step": 7645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000503476146440302,
+      "loss": 2.0511,
+      "step": 7646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000503367518533071,
+      "loss": 1.8867,
+      "step": 7647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005032588904668851,
+      "loss": 1.9755,
+      "step": 7648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005031502622468723,
+      "loss": 2.0864,
+      "step": 7649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005030416338781598,
+      "loss": 1.8637,
+      "step": 7650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005029330053658752,
+      "loss": 2.0176,
+      "step": 7651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005028243767151459,
+      "loss": 2.0466,
+      "step": 7652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005027157479310998,
+      "loss": 2.0696,
+      "step": 7653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005026071190188639,
+      "loss": 2.1414,
+      "step": 7654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005024984899835661,
+      "loss": 2.0461,
+      "step": 7655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005023898608303337,
+      "loss": 2.0839,
+      "step": 7656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005022812315642945,
+      "loss": 2.095,
+      "step": 7657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005021726021905758,
+      "loss": 2.0178,
+      "step": 7658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005020639727143053,
+      "loss": 1.9717,
+      "step": 7659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005019553431406107,
+      "loss": 2.0697,
+      "step": 7660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005018467134746192,
+      "loss": 2.1685,
+      "step": 7661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005017380837214586,
+      "loss": 2.0201,
+      "step": 7662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005016294538862567,
+      "loss": 1.8668,
+      "step": 7663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005015208239741406,
+      "loss": 1.9579,
+      "step": 7664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005014121939902383,
+      "loss": 2.1326,
+      "step": 7665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000501303563939677,
+      "loss": 2.0678,
+      "step": 7666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005011949338275846,
+      "loss": 1.9874,
+      "step": 7667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005010863036590885,
+      "loss": 2.0485,
+      "step": 7668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005009776734393165,
+      "loss": 2.116,
+      "step": 7669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005008690431733959,
+      "loss": 2.1361,
+      "step": 7670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005007604128664545,
+      "loss": 2.2148,
+      "step": 7671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005006517825236199,
+      "loss": 2.399,
+      "step": 7672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005005431521500197,
+      "loss": 1.9175,
+      "step": 7673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005004345217507813,
+      "loss": 2.1021,
+      "step": 7674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005003258913310327,
+      "loss": 2.0792,
+      "step": 7675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005002172608959011,
+      "loss": 2.0856,
+      "step": 7676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0005001086304505145,
+      "loss": 1.9789,
+      "step": 7677
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0005,
+      "loss": 1.9313,
+      "step": 7678
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004998913695494857,
+      "loss": 2.0825,
+      "step": 7679
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000499782739104099,
+      "loss": 2.1995,
+      "step": 7680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004996741086689674,
+      "loss": 2.0781,
+      "step": 7681
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004995654782492188,
+      "loss": 2.0791,
+      "step": 7682
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004994568478499804,
+      "loss": 2.1835,
+      "step": 7683
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004993482174763803,
+      "loss": 2.0947,
+      "step": 7684
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004992395871335455,
+      "loss": 2.0501,
+      "step": 7685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004991309568266042,
+      "loss": 1.8729,
+      "step": 7686
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004990223265606837,
+      "loss": 2.124,
+      "step": 7687
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004989136963409116,
+      "loss": 2.0309,
+      "step": 7688
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004988050661724155,
+      "loss": 2.1414,
+      "step": 7689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004986964360603231,
+      "loss": 1.9598,
+      "step": 7690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004985878060097619,
+      "loss": 2.105,
+      "step": 7691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004984791760258595,
+      "loss": 2.0779,
+      "step": 7692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004983705461137434,
+      "loss": 2.0434,
+      "step": 7693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004982619162785414,
+      "loss": 2.0831,
+      "step": 7694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004981532865253808,
+      "loss": 2.188,
+      "step": 7695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004980446568593895,
+      "loss": 2.0761,
+      "step": 7696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004979360272856947,
+      "loss": 1.8971,
+      "step": 7697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004978273978094242,
+      "loss": 2.17,
+      "step": 7698
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004977187684357057,
+      "loss": 2.111,
+      "step": 7699
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004976101391696665,
+      "loss": 1.9714,
+      "step": 7700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000497501510016434,
+      "loss": 1.9771,
+      "step": 7701
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004973928809811363,
+      "loss": 2.1749,
+      "step": 7702
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004972842520689003,
+      "loss": 2.0294,
+      "step": 7703
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004971756232848542,
+      "loss": 2.0333,
+      "step": 7704
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004970669946341248,
+      "loss": 2.1351,
+      "step": 7705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004969583661218403,
+      "loss": 2.054,
+      "step": 7706
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004968497377531277,
+      "loss": 2.2126,
+      "step": 7707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004967411095331149,
+      "loss": 2.1116,
+      "step": 7708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004966324814669291,
+      "loss": 2.2823,
+      "step": 7709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000496523853559698,
+      "loss": 2.0276,
+      "step": 7710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000496415225816549,
+      "loss": 1.9757,
+      "step": 7711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004963065982426096,
+      "loss": 2.0269,
+      "step": 7712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004961979708430072,
+      "loss": 2.031,
+      "step": 7713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004960893436228693,
+      "loss": 2.0765,
+      "step": 7714
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004959807165873234,
+      "loss": 2.0648,
+      "step": 7715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000495872089741497,
+      "loss": 2.0628,
+      "step": 7716
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004957634630905173,
+      "loss": 2.0564,
+      "step": 7717
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000495654836639512,
+      "loss": 2.1293,
+      "step": 7718
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004955462103936084,
+      "loss": 2.05,
+      "step": 7719
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004954375843579339,
+      "loss": 1.8544,
+      "step": 7720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004953289585376159,
+      "loss": 2.099,
+      "step": 7721
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004952203329377818,
+      "loss": 2.1183,
+      "step": 7722
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004951117075635589,
+      "loss": 2.0255,
+      "step": 7723
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004950030824200747,
+      "loss": 1.9598,
+      "step": 7724
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004948944575124566,
+      "loss": 2.0626,
+      "step": 7725
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004947858328458318,
+      "loss": 2.0747,
+      "step": 7726
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004946772084253276,
+      "loss": 2.0449,
+      "step": 7727
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004945685842560714,
+      "loss": 2.0357,
+      "step": 7728
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004944599603431905,
+      "loss": 2.0512,
+      "step": 7729
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004943513366918124,
+      "loss": 2.1377,
+      "step": 7730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000494242713307064,
+      "loss": 1.9493,
+      "step": 7731
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004941340901940729,
+      "loss": 2.0241,
+      "step": 7732
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004940254673579661,
+      "loss": 2.0999,
+      "step": 7733
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004939168448038711,
+      "loss": 2.0408,
+      "step": 7734
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004938082225369149,
+      "loss": 2.1269,
+      "step": 7735
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004936996005622251,
+      "loss": 1.9834,
+      "step": 7736
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004935909788849283,
+      "loss": 2.186,
+      "step": 7737
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004934823575101523,
+      "loss": 2.1322,
+      "step": 7738
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004933737364430238,
+      "loss": 2.2757,
+      "step": 7739
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004932651156886703,
+      "loss": 1.9866,
+      "step": 7740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004931564952522187,
+      "loss": 1.9543,
+      "step": 7741
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004930478751387964,
+      "loss": 2.0826,
+      "step": 7742
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004929392553535304,
+      "loss": 2.1573,
+      "step": 7743
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004928306359015478,
+      "loss": 2.1283,
+      "step": 7744
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004927220167879755,
+      "loss": 2.0472,
+      "step": 7745
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004926133980179409,
+      "loss": 2.063,
+      "step": 7746
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004925047795965708,
+      "loss": 1.8885,
+      "step": 7747
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004923961615289924,
+      "loss": 2.1268,
+      "step": 7748
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004922875438203326,
+      "loss": 2.0205,
+      "step": 7749
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004921789264757186,
+      "loss": 2.2371,
+      "step": 7750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004920703095002771,
+      "loss": 1.8968,
+      "step": 7751
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004919616928991354,
+      "loss": 2.1659,
+      "step": 7752
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004918530766774202,
+      "loss": 2.0108,
+      "step": 7753
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004917444608402584,
+      "loss": 2.1702,
+      "step": 7754
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004916358453927772,
+      "loss": 1.9479,
+      "step": 7755
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004915272303401033,
+      "loss": 2.0752,
+      "step": 7756
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004914186156873635,
+      "loss": 2.1237,
+      "step": 7757
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004913100014396849,
+      "loss": 2.2139,
+      "step": 7758
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004912013876021941,
+      "loss": 1.9559,
+      "step": 7759
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004910927741800182,
+      "loss": 2.066,
+      "step": 7760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004909841611782837,
+      "loss": 2.033,
+      "step": 7761
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004908755486021176,
+      "loss": 2.0621,
+      "step": 7762
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004907669364566465,
+      "loss": 2.0492,
+      "step": 7763
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004906583247469974,
+      "loss": 2.0834,
+      "step": 7764
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004905497134782966,
+      "loss": 2.1736,
+      "step": 7765
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004904411026556713,
+      "loss": 2.0393,
+      "step": 7766
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004903324922842478,
+      "loss": 2.1372,
+      "step": 7767
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000490223882369153,
+      "loss": 2.1058,
+      "step": 7768
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004901152729155133,
+      "loss": 1.8365,
+      "step": 7769
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004900066639284555,
+      "loss": 2.1038,
+      "step": 7770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004898980554131062,
+      "loss": 2.089,
+      "step": 7771
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004897894473745918,
+      "loss": 2.0951,
+      "step": 7772
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004896808398180392,
+      "loss": 2.0871,
+      "step": 7773
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004895722327485744,
+      "loss": 2.1705,
+      "step": 7774
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004894636261713244,
+      "loss": 2.1023,
+      "step": 7775
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004893550200914153,
+      "loss": 2.0153,
+      "step": 7776
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000489246414513974,
+      "loss": 2.0661,
+      "step": 7777
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004891378094441264,
+      "loss": 2.1445,
+      "step": 7778
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004890292048869993,
+      "loss": 2.1647,
+      "step": 7779
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004889206008477189,
+      "loss": 2.2304,
+      "step": 7780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004888119973314117,
+      "loss": 2.0832,
+      "step": 7781
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004887033943432039,
+      "loss": 2.0521,
+      "step": 7782
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000488594791888222,
+      "loss": 2.0922,
+      "step": 7783
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004884861899715918,
+      "loss": 2.1629,
+      "step": 7784
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004883775885984403,
+      "loss": 2.0704,
+      "step": 7785
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004882689877738931,
+      "loss": 2.0424,
+      "step": 7786
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004881603875030768,
+      "loss": 2.1072,
+      "step": 7787
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004880517877911173,
+      "loss": 2.0648,
+      "step": 7788
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048794318864314105,
+      "loss": 2.0475,
+      "step": 7789
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004878345900642739,
+      "loss": 2.0075,
+      "step": 7790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048772599205964217,
+      "loss": 2.2534,
+      "step": 7791
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004876173946343717,
+      "loss": 2.107,
+      "step": 7792
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004875087977935889,
+      "loss": 1.9695,
+      "step": 7793
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048740020154241934,
+      "loss": 1.9983,
+      "step": 7794
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048729160588598943,
+      "loss": 1.928,
+      "step": 7795
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004871830108294248,
+      "loss": 2.0831,
+      "step": 7796
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004870744163778516,
+      "loss": 2.0853,
+      "step": 7797
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048696582253639567,
+      "loss": 1.9995,
+      "step": 7798
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048685722931018287,
+      "loss": 2.0249,
+      "step": 7799
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048674863670433906,
+      "loss": 1.9336,
+      "step": 7800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004866400447239901,
+      "loss": 2.0593,
+      "step": 7801
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048653145337426175,
+      "loss": 2.0639,
+      "step": 7802
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048642286266027977,
+      "loss": 2.0074,
+      "step": 7803
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048631427258716986,
+      "loss": 2.0149,
+      "step": 7804
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048620568316005785,
+      "loss": 2.0786,
+      "step": 7805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048609709438406925,
+      "loss": 2.0604,
+      "step": 7806
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048598850626432985,
+      "loss": 1.9873,
+      "step": 7807
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048587991880596506,
+      "loss": 2.0953,
+      "step": 7808
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048577133201410064,
+      "loss": 2.0832,
+      "step": 7809
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000485662745893862,
+      "loss": 1.9011,
+      "step": 7810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004855541604503748,
+      "loss": 1.9611,
+      "step": 7811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004854455756887644,
+      "loss": 2.1284,
+      "step": 7812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048533699161415635,
+      "loss": 2.041,
+      "step": 7813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000485228408231676,
+      "loss": 2.1021,
+      "step": 7814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048511982554644875,
+      "loss": 2.1202,
+      "step": 7815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004850112435635999,
+      "loss": 2.0367,
+      "step": 7816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004849026622882549,
+      "loss": 2.0282,
+      "step": 7817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048479408172553886,
+      "loss": 2.0251,
+      "step": 7818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048468550188057724,
+      "loss": 1.9964,
+      "step": 7819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048457692275849507,
+      "loss": 2.0121,
+      "step": 7820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004844683443644177,
+      "loss": 2.1041,
+      "step": 7821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004843597667034701,
+      "loss": 2.0821,
+      "step": 7822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004842511897807775,
+      "loss": 2.0386,
+      "step": 7823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000484142613601465,
+      "loss": 2.1736,
+      "step": 7824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0004840340381706576,
+      "loss": 2.1191,
+      "step": 7825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00048392546349348026,
+      "loss": 2.0411,
+      "step": 7826
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048381688957505807,
+      "loss": 1.9806,
+      "step": 7827
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004837083164205159,
+      "loss": 2.0887,
+      "step": 7828
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004835997440349787,
+      "loss": 2.1646,
+      "step": 7829
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048349117242357114,
+      "loss": 2.1346,
+      "step": 7830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048338260159141837,
+      "loss": 2.0654,
+      "step": 7831
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048327403154364493,
+      "loss": 2.0252,
+      "step": 7832
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004831654622853757,
+      "loss": 2.0996,
+      "step": 7833
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048305689382173537,
+      "loss": 2.0644,
+      "step": 7834
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048294832615784864,
+      "loss": 2.0834,
+      "step": 7835
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048283975929884005,
+      "loss": 2.1687,
+      "step": 7836
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004827311932498344,
+      "loss": 2.201,
+      "step": 7837
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048262262801595596,
+      "loss": 2.2043,
+      "step": 7838
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048251406360232964,
+      "loss": 2.1097,
+      "step": 7839
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048240550001407955,
+      "loss": 2.1322,
+      "step": 7840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004822969372563305,
+      "loss": 2.1206,
+      "step": 7841
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048218837533420657,
+      "loss": 1.9759,
+      "step": 7842
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048207981425283237,
+      "loss": 2.0029,
+      "step": 7843
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004819712540173321,
+      "loss": 2.0781,
+      "step": 7844
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048186269463283015,
+      "loss": 2.0435,
+      "step": 7845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048175413610445066,
+      "loss": 2.1614,
+      "step": 7846
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000481645578437318,
+      "loss": 2.0353,
+      "step": 7847
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004815370216365561,
+      "loss": 2.1847,
+      "step": 7848
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004814284657072894,
+      "loss": 1.973,
+      "step": 7849
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004813199106546417,
+      "loss": 2.0482,
+      "step": 7850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048121135648373733,
+      "loss": 1.9932,
+      "step": 7851
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048110280319969996,
+      "loss": 2.0397,
+      "step": 7852
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048099425080765383,
+      "loss": 2.0362,
+      "step": 7853
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048088569931272276,
+      "loss": 2.0657,
+      "step": 7854
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004807771487200307,
+      "loss": 1.9956,
+      "step": 7855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004806685990347013,
+      "loss": 2.1829,
+      "step": 7856
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004805600502618586,
+      "loss": 2.1785,
+      "step": 7857
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048045150240662613,
+      "loss": 2.0941,
+      "step": 7858
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048034295547412774,
+      "loss": 2.1192,
+      "step": 7859
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048023440946948703,
+      "loss": 2.0306,
+      "step": 7860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00048012586439782765,
+      "loss": 2.1328,
+      "step": 7861
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004800173202642732,
+      "loss": 2.2553,
+      "step": 7862
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004799087770739472,
+      "loss": 2.0131,
+      "step": 7863
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047980023483197316,
+      "loss": 2.0129,
+      "step": 7864
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004796916935434744,
+      "loss": 2.1619,
+      "step": 7865
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047958315321357454,
+      "loss": 2.141,
+      "step": 7866
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047947461384739666,
+      "loss": 1.9795,
+      "step": 7867
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047936607545006433,
+      "loss": 2.0058,
+      "step": 7868
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004792575380267005,
+      "loss": 1.9973,
+      "step": 7869
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004791490015824288,
+      "loss": 2.1431,
+      "step": 7870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000479040466122372,
+      "loss": 1.9976,
+      "step": 7871
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047893193165165354,
+      "loss": 1.9804,
+      "step": 7872
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047882339817539627,
+      "loss": 1.9774,
+      "step": 7873
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047871486569872336,
+      "loss": 1.9512,
+      "step": 7874
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047860633422675767,
+      "loss": 2.1013,
+      "step": 7875
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004784978037646223,
+      "loss": 1.8915,
+      "step": 7876
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004783892743174399,
+      "loss": 2.0917,
+      "step": 7877
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004782807458903336,
+      "loss": 1.9913,
+      "step": 7878
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004781722184884259,
+      "loss": 2.1331,
+      "step": 7879
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004780636921168398,
+      "loss": 2.0979,
+      "step": 7880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004779551667806978,
+      "loss": 2.1848,
+      "step": 7881
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047784664248512265,
+      "loss": 2.0132,
+      "step": 7882
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047773811923523693,
+      "loss": 1.8768,
+      "step": 7883
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004776295970361632,
+      "loss": 2.1607,
+      "step": 7884
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047752107589302386,
+      "loss": 1.9663,
+      "step": 7885
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047741255581094157,
+      "loss": 2.1597,
+      "step": 7886
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047730403679503843,
+      "loss": 1.8634,
+      "step": 7887
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047719551885043703,
+      "loss": 2.1751,
+      "step": 7888
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004770870019822595,
+      "loss": 2.174,
+      "step": 7889
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004769784861956283,
+      "loss": 2.0985,
+      "step": 7890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004768699714956654,
+      "loss": 1.9801,
+      "step": 7891
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047676145788749314,
+      "loss": 1.9594,
+      "step": 7892
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004766529453762334,
+      "loss": 2.0992,
+      "step": 7893
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047654443396700835,
+      "loss": 2.1253,
+      "step": 7894
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047643592366493993,
+      "loss": 2.032,
+      "step": 7895
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004763274144751502,
+      "loss": 2.0614,
+      "step": 7896
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047621890640276083,
+      "loss": 1.9601,
+      "step": 7897
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004761103994528939,
+      "loss": 1.9062,
+      "step": 7898
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004760018936306709,
+      "loss": 1.8853,
+      "step": 7899
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004758933889412139,
+      "loss": 2.0588,
+      "step": 7900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004757848853896442,
+      "loss": 1.9348,
+      "step": 7901
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004756763829810837,
+      "loss": 2.1337,
+      "step": 7902
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004755678817206538,
+      "loss": 2.0115,
+      "step": 7903
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004754593816134759,
+      "loss": 2.0321,
+      "step": 7904
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047535088266467186,
+      "loss": 2.0249,
+      "step": 7905
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047524238487936277,
+      "loss": 2.0016,
+      "step": 7906
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004751338882626701,
+      "loss": 2.0906,
+      "step": 7907
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047502539281971503,
+      "loss": 2.093,
+      "step": 7908
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047491689855561894,
+      "loss": 1.9818,
+      "step": 7909
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004748084054755028,
+      "loss": 2.0304,
+      "step": 7910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000474699913584488,
+      "loss": 1.9594,
+      "step": 7911
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047459142288769536,
+      "loss": 2.0277,
+      "step": 7912
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047448293339024604,
+      "loss": 1.9354,
+      "step": 7913
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047437444509726087,
+      "loss": 2.0979,
+      "step": 7914
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004742659580138609,
+      "loss": 1.9104,
+      "step": 7915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004741574721451669,
+      "loss": 2.0532,
+      "step": 7916
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004740489874962996,
+      "loss": 2.0605,
+      "step": 7917
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004739405040723797,
+      "loss": 2.0709,
+      "step": 7918
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004738320218785281,
+      "loss": 2.0504,
+      "step": 7919
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047372354091986506,
+      "loss": 1.9076,
+      "step": 7920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004736150612015115,
+      "loss": 2.1095,
+      "step": 7921
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004735065827285876,
+      "loss": 1.867,
+      "step": 7922
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000473398105506214,
+      "loss": 2.0374,
+      "step": 7923
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004732896295395109,
+      "loss": 2.0685,
+      "step": 7924
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004731811548335989,
+      "loss": 2.0726,
+      "step": 7925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047307268139359785,
+      "loss": 2.0299,
+      "step": 7926
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004729642092246282,
+      "loss": 1.9532,
+      "step": 7927
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047285573833181013,
+      "loss": 2.047,
+      "step": 7928
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047274726872026363,
+      "loss": 2.131,
+      "step": 7929
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047263880039510864,
+      "loss": 1.9813,
+      "step": 7930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004725303333614653,
+      "loss": 2.0322,
+      "step": 7931
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004724218676244533,
+      "loss": 2.1597,
+      "step": 7932
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004723134031891926,
+      "loss": 2.2088,
+      "step": 7933
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047220494006080296,
+      "loss": 2.0292,
+      "step": 7934
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004720964782444041,
+      "loss": 2.0533,
+      "step": 7935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004719880177451155,
+      "loss": 1.9456,
+      "step": 7936
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000471879558568057,
+      "loss": 2.0294,
+      "step": 7937
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047177110071834795,
+      "loss": 2.0716,
+      "step": 7938
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047166264420110785,
+      "loss": 1.9163,
+      "step": 7939
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004715541890214562,
+      "loss": 2.0491,
+      "step": 7940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004714457351845122,
+      "loss": 2.0487,
+      "step": 7941
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047133728269539515,
+      "loss": 1.9822,
+      "step": 7942
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004712288315592242,
+      "loss": 2.0646,
+      "step": 7943
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004711203817811187,
+      "loss": 2.1522,
+      "step": 7944
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004710119333661975,
+      "loss": 1.8146,
+      "step": 7945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004709034863195798,
+      "loss": 2.0352,
+      "step": 7946
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047079504064638435,
+      "loss": 1.9237,
+      "step": 7947
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047068659635173026,
+      "loss": 2.0832,
+      "step": 7948
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004705781534407361,
+      "loss": 2.1672,
+      "step": 7949
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004704697119185209,
+      "loss": 2.1955,
+      "step": 7950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004703612717902031,
+      "loss": 2.1455,
+      "step": 7951
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004702528330609015,
+      "loss": 2.1502,
+      "step": 7952
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004701443957357345,
+      "loss": 1.9503,
+      "step": 7953
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00047003595981982084,
+      "loss": 2.0792,
+      "step": 7954
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004699275253182786,
+      "loss": 2.1612,
+      "step": 7955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004698190922362264,
+      "loss": 2.0144,
+      "step": 7956
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046971066057878246,
+      "loss": 2.0446,
+      "step": 7957
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046960223035106505,
+      "loss": 2.1207,
+      "step": 7958
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046949380155819213,
+      "loss": 1.9676,
+      "step": 7959
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004693853742052821,
+      "loss": 2.0371,
+      "step": 7960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046927694829745263,
+      "loss": 2.1189,
+      "step": 7961
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000469168523839822,
+      "loss": 1.9585,
+      "step": 7962
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046906010083750787,
+      "loss": 2.0566,
+      "step": 7963
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004689516792956282,
+      "loss": 2.1825,
+      "step": 7964
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004688432592193006,
+      "loss": 2.1048,
+      "step": 7965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004687348406136429,
+      "loss": 2.0237,
+      "step": 7966
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004686264234837725,
+      "loss": 2.078,
+      "step": 7967
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046851800783480715,
+      "loss": 1.9915,
+      "step": 7968
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046840959367186414,
+      "loss": 2.1425,
+      "step": 7969
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004683011810000611,
+      "loss": 1.9808,
+      "step": 7970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004681927698245151,
+      "loss": 2.0332,
+      "step": 7971
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00046808436015034353,
+      "loss": 1.8887,
+      "step": 7972
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004679759519826636,
+      "loss": 1.9288,
+      "step": 7973
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004678675453265924,
+      "loss": 2.0732,
+      "step": 7974
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0004677591401872468,
+      "loss": 2.2251,
+      "step": 7975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004676507365697441,
+      "loss": 2.2004,
+      "step": 7976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004675423344792009,
+      "loss": 2.1918,
+      "step": 7977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004674339339207343,
+      "loss": 2.0643,
+      "step": 7978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004673255348994607,
+      "loss": 2.0347,
+      "step": 7979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004672171374204972,
+      "loss": 2.0269,
+      "step": 7980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004671087414889601,
+      "loss": 2.1253,
+      "step": 7981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046700034710996606,
+      "loss": 2.1869,
+      "step": 7982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004668919542886315,
+      "loss": 2.1145,
+      "step": 7983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004667835630300729,
+      "loss": 2.0922,
+      "step": 7984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046667517333940647,
+      "loss": 2.0687,
+      "step": 7985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004665667852217485,
+      "loss": 2.1837,
+      "step": 7986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004664583986822151,
+      "loss": 1.8509,
+      "step": 7987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046635001372592256,
+      "loss": 2.1533,
+      "step": 7988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004662416303579866,
+      "loss": 2.0154,
+      "step": 7989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004661332485835234,
+      "loss": 2.0831,
+      "step": 7990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004660248684076487,
+      "loss": 2.1356,
+      "step": 7991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004659164898354784,
+      "loss": 2.2671,
+      "step": 7992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004658081128721281,
+      "loss": 1.9375,
+      "step": 7993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004656997375227136,
+      "loss": 1.9487,
+      "step": 7994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004655913637923503,
+      "loss": 2.0886,
+      "step": 7995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046548299168615374,
+      "loss": 2.1212,
+      "step": 7996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046537462120923933,
+      "loss": 2.0586,
+      "step": 7997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046526625236672256,
+      "loss": 2.0717,
+      "step": 7998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004651578851637184,
+      "loss": 2.0357,
+      "step": 7999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004650495196053423,
+      "loss": 2.2208,
+      "step": 8000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004649411556967091,
+      "loss": 2.1251,
+      "step": 8001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046483279344293417,
+      "loss": 2.0384,
+      "step": 8002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046472443284913215,
+      "loss": 2.1476,
+      "step": 8003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004646160739204181,
+      "loss": 2.038,
+      "step": 8004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004645077166619066,
+      "loss": 2.0436,
+      "step": 8005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004643993610787126,
+      "loss": 1.946,
+      "step": 8006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046429100717595057,
+      "loss": 2.1309,
+      "step": 8007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004641826549587352,
+      "loss": 1.8554,
+      "step": 8008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004640743044321808,
+      "loss": 2.059,
+      "step": 8009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004639659556014019,
+      "loss": 2.0229,
+      "step": 8010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004638576084715127,
+      "loss": 2.1378,
+      "step": 8011
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004637492630476275,
+      "loss": 1.9992,
+      "step": 8012
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046364091933486054,
+      "loss": 2.0302,
+      "step": 8013
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046353257733832575,
+      "loss": 2.117,
+      "step": 8014
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004634242370631371,
+      "loss": 2.2409,
+      "step": 8015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004633158985144087,
+      "loss": 2.1541,
+      "step": 8016
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004632075616972541,
+      "loss": 2.1643,
+      "step": 8017
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046309922661678737,
+      "loss": 1.9648,
+      "step": 8018
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046299089327812187,
+      "loss": 1.9456,
+      "step": 8019
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004628825616863714,
+      "loss": 2.139,
+      "step": 8020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046277423184664935,
+      "loss": 1.9685,
+      "step": 8021
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004626659037640692,
+      "loss": 1.9855,
+      "step": 8022
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004625575774437442,
+      "loss": 2.1509,
+      "step": 8023
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046244925289078766,
+      "loss": 1.9141,
+      "step": 8024
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004623409301103127,
+      "loss": 2.0156,
+      "step": 8025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046223260910743254,
+      "loss": 1.8755,
+      "step": 8026
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046212428988726,
+      "loss": 1.9811,
+      "step": 8027
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000462015972454908,
+      "loss": 2.0212,
+      "step": 8028
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046190765681548965,
+      "loss": 2.1581,
+      "step": 8029
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004617993429741173,
+      "loss": 2.1517,
+      "step": 8030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000461691030935904,
+      "loss": 2.1853,
+      "step": 8031
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046158272070596196,
+      "loss": 2.0485,
+      "step": 8032
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000461474412289404,
+      "loss": 2.0947,
+      "step": 8033
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046136610569134223,
+      "loss": 2.0662,
+      "step": 8034
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004612578009168893,
+      "loss": 2.1666,
+      "step": 8035
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004611494979711571,
+      "loss": 2.2029,
+      "step": 8036
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046104119685925804,
+      "loss": 2.1572,
+      "step": 8037
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046093289758630395,
+      "loss": 1.97,
+      "step": 8038
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00046082460015740715,
+      "loss": 2.1697,
+      "step": 8039
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004607163045776791,
+      "loss": 1.9798,
+      "step": 8040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000460608010852232,
+      "loss": 2.0999,
+      "step": 8041
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004604997189861772,
+      "loss": 1.9996,
+      "step": 8042
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004603914289846267,
+      "loss": 2.0584,
+      "step": 8043
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004602831408526916,
+      "loss": 1.8845,
+      "step": 8044
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004601748545954837,
+      "loss": 2.1185,
+      "step": 8045
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004600665702181143,
+      "loss": 2.1051,
+      "step": 8046
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045995828772569456,
+      "loss": 1.9479,
+      "step": 8047
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045985000712333576,
+      "loss": 2.0833,
+      "step": 8048
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045974172841614896,
+      "loss": 2.1046,
+      "step": 8049
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004596334516092451,
+      "loss": 1.959,
+      "step": 8050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004595251767077352,
+      "loss": 2.0523,
+      "step": 8051
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045941690371673,
+      "loss": 2.0652,
+      "step": 8052
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045930863264134037,
+      "loss": 2.0703,
+      "step": 8053
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045920036348667673,
+      "loss": 2.1179,
+      "step": 8054
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045909209625784986,
+      "loss": 1.9483,
+      "step": 8055
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045898383095997,
+      "loss": 2.169,
+      "step": 8056
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045887556759814776,
+      "loss": 2.0843,
+      "step": 8057
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045876730617749325,
+      "loss": 1.9828,
+      "step": 8058
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045865904670311673,
+      "loss": 1.999,
+      "step": 8059
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004585507891801282,
+      "loss": 2.1187,
+      "step": 8060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045844253361363785,
+      "loss": 2.1653,
+      "step": 8061
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045833428000875546,
+      "loss": 1.9594,
+      "step": 8062
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045822602837059087,
+      "loss": 2.1635,
+      "step": 8063
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045811777870425377,
+      "loss": 1.9867,
+      "step": 8064
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004580095310148539,
+      "loss": 1.935,
+      "step": 8065
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004579012853075006,
+      "loss": 2.0323,
+      "step": 8066
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045779304158730355,
+      "loss": 2.0361,
+      "step": 8067
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045768479985937196,
+      "loss": 2.0351,
+      "step": 8068
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045757656012881514,
+      "loss": 2.1347,
+      "step": 8069
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004574683224007422,
+      "loss": 2.0445,
+      "step": 8070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045736008668026233,
+      "loss": 2.1774,
+      "step": 8071
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004572518529724844,
+      "loss": 2.0543,
+      "step": 8072
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045714362128251737,
+      "loss": 2.0823,
+      "step": 8073
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004570353916154698,
+      "loss": 2.089,
+      "step": 8074
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045692716397645073,
+      "loss": 2.1382,
+      "step": 8075
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004568189383705684,
+      "loss": 2.0046,
+      "step": 8076
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045671071480293167,
+      "loss": 2.0854,
+      "step": 8077
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045660249327864863,
+      "loss": 2.0705,
+      "step": 8078
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045649427380282785,
+      "loss": 2.1776,
+      "step": 8079
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045638605638057716,
+      "loss": 2.1153,
+      "step": 8080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004562778410170051,
+      "loss": 1.9908,
+      "step": 8081
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004561696277172194,
+      "loss": 1.8983,
+      "step": 8082
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045606141648632813,
+      "loss": 2.0854,
+      "step": 8083
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000455953207329439,
+      "loss": 2.1126,
+      "step": 8084
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045584500025165983,
+      "loss": 2.1039,
+      "step": 8085
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004557367952580981,
+      "loss": 1.8963,
+      "step": 8086
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004556285923538615,
+      "loss": 1.9931,
+      "step": 8087
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045552039154405727,
+      "loss": 2.1702,
+      "step": 8088
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000455412192833793,
+      "loss": 2.0354,
+      "step": 8089
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045530399622817564,
+      "loss": 2.0318,
+      "step": 8090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045519580173231257,
+      "loss": 2.084,
+      "step": 8091
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045508760935131053,
+      "loss": 1.994,
+      "step": 8092
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004549794190902768,
+      "loss": 2.0811,
+      "step": 8093
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004548712309543179,
+      "loss": 2.0721,
+      "step": 8094
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004547630449485408,
+      "loss": 2.1366,
+      "step": 8095
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004546548610780518,
+      "loss": 2.0709,
+      "step": 8096
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004545466793479578,
+      "loss": 2.0495,
+      "step": 8097
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000454438499763365,
+      "loss": 1.9169,
+      "step": 8098
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045433032232937984,
+      "loss": 2.153,
+      "step": 8099
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004542221470511084,
+      "loss": 2.0827,
+      "step": 8100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000454113973933657,
+      "loss": 2.124,
+      "step": 8101
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045400580298213147,
+      "loss": 2.1551,
+      "step": 8102
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045389763420163786,
+      "loss": 2.013,
+      "step": 8103
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045378946759728193,
+      "loss": 2.0415,
+      "step": 8104
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004536813031741694,
+      "loss": 2.0814,
+      "step": 8105
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004535731409374058,
+      "loss": 2.0626,
+      "step": 8106
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004534649808920968,
+      "loss": 2.0245,
+      "step": 8107
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004533568230433477,
+      "loss": 2.1399,
+      "step": 8108
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045324866739626386,
+      "loss": 1.9966,
+      "step": 8109
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045314051395595027,
+      "loss": 2.0633,
+      "step": 8110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004530323627275124,
+      "loss": 1.9795,
+      "step": 8111
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045292421371605484,
+      "loss": 2.1747,
+      "step": 8112
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004528160669266828,
+      "loss": 2.0263,
+      "step": 8113
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045270792236450075,
+      "loss": 2.0604,
+      "step": 8114
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045259978003461365,
+      "loss": 2.2824,
+      "step": 8115
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004524916399421259,
+      "loss": 1.9705,
+      "step": 8116
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000452383502092142,
+      "loss": 2.072,
+      "step": 8117
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045227536648976624,
+      "loss": 2.0254,
+      "step": 8118
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045216723314010294,
+      "loss": 2.1196,
+      "step": 8119
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045205910204825634,
+      "loss": 1.8927,
+      "step": 8120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004519509732193303,
+      "loss": 2.0095,
+      "step": 8121
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004518428466584289,
+      "loss": 2.0155,
+      "step": 8122
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0004517347223706557,
+      "loss": 2.0848,
+      "step": 8123
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00045162660036111483,
+      "loss": 2.0439,
+      "step": 8124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045151848063490946,
+      "loss": 1.9956,
+      "step": 8125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045141036319714344,
+      "loss": 2.0062,
+      "step": 8126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045130224805292,
+      "loss": 2.0198,
+      "step": 8127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004511941352073424,
+      "loss": 2.0667,
+      "step": 8128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045108602466551384,
+      "loss": 2.1229,
+      "step": 8129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045097791643253747,
+      "loss": 2.15,
+      "step": 8130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045086981051351607,
+      "loss": 2.2261,
+      "step": 8131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045076170691355275,
+      "loss": 2.0715,
+      "step": 8132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045065360563774994,
+      "loss": 1.9717,
+      "step": 8133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045054550669121053,
+      "loss": 2.0677,
+      "step": 8134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045043741007903685,
+      "loss": 2.0538,
+      "step": 8135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004503293158063315,
+      "loss": 2.1435,
+      "step": 8136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004502212238781965,
+      "loss": 1.9609,
+      "step": 8137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045011313429973445,
+      "loss": 1.9339,
+      "step": 8138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00045000504707604696,
+      "loss": 2.0012,
+      "step": 8139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044989696221223635,
+      "loss": 2.1085,
+      "step": 8140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044978887971340424,
+      "loss": 2.2353,
+      "step": 8141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044968079958465266,
+      "loss": 1.9621,
+      "step": 8142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044957272183108283,
+      "loss": 1.9905,
+      "step": 8143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044946464645779664,
+      "loss": 1.9988,
+      "step": 8144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044935657346989536,
+      "loss": 2.0856,
+      "step": 8145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004492485028724803,
+      "loss": 2.159,
+      "step": 8146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004491404346706524,
+      "loss": 2.1646,
+      "step": 8147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044903236886951315,
+      "loss": 2.1329,
+      "step": 8148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004489243054741632,
+      "loss": 2.1254,
+      "step": 8149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004488162444897036,
+      "loss": 2.1238,
+      "step": 8150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044870818592123487,
+      "loss": 2.0275,
+      "step": 8151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004486001297738578,
+      "loss": 2.2069,
+      "step": 8152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044849207605267267,
+      "loss": 2.0663,
+      "step": 8153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004483840247627801,
+      "loss": 2.1003,
+      "step": 8154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004482759759092802,
+      "loss": 2.1542,
+      "step": 8155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044816792949727333,
+      "loss": 1.9852,
+      "step": 8156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004480598855318592,
+      "loss": 1.9679,
+      "step": 8157
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004479518440181381,
+      "loss": 2.0879,
+      "step": 8158
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044784380496120947,
+      "loss": 2.1762,
+      "step": 8159
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044773576836617336,
+      "loss": 2.0235,
+      "step": 8160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044762773423812904,
+      "loss": 2.1831,
+      "step": 8161
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044751970258217615,
+      "loss": 2.05,
+      "step": 8162
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044741167340341397,
+      "loss": 1.9387,
+      "step": 8163
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044730364670694177,
+      "loss": 1.9665,
+      "step": 8164
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004471956224978586,
+      "loss": 1.8922,
+      "step": 8165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004470876007812636,
+      "loss": 1.9924,
+      "step": 8166
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044697958156225535,
+      "loss": 2.0719,
+      "step": 8167
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004468715648459329,
+      "loss": 2.0794,
+      "step": 8168
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044676355063739465,
+      "loss": 1.9166,
+      "step": 8169
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004466555389417394,
+      "loss": 2.0029,
+      "step": 8170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004465475297640652,
+      "loss": 2.0298,
+      "step": 8171
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004464395231094707,
+      "loss": 1.938,
+      "step": 8172
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044633151898305375,
+      "loss": 2.1454,
+      "step": 8173
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044622351738991264,
+      "loss": 2.0295,
+      "step": 8174
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000446115518335145,
+      "loss": 2.2093,
+      "step": 8175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000446007521823849,
+      "loss": 2.0476,
+      "step": 8176
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000445899527861122,
+      "loss": 2.1148,
+      "step": 8177
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004457915364520618,
+      "loss": 2.2511,
+      "step": 8178
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004456835476017657,
+      "loss": 1.9821,
+      "step": 8179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044557556131533105,
+      "loss": 2.142,
+      "step": 8180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004454675775978551,
+      "loss": 1.9336,
+      "step": 8181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004453595964544349,
+      "loss": 2.04,
+      "step": 8182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004452516178901673,
+      "loss": 2.0192,
+      "step": 8183
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044514364191014933,
+      "loss": 1.9965,
+      "step": 8184
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004450356685194775,
+      "loss": 2.0731,
+      "step": 8185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004449276977232487,
+      "loss": 2.1612,
+      "step": 8186
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000444819729526559,
+      "loss": 2.1608,
+      "step": 8187
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004447117639345052,
+      "loss": 2.1334,
+      "step": 8188
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004446038009521831,
+      "loss": 2.0137,
+      "step": 8189
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044449584058468914,
+      "loss": 2.0267,
+      "step": 8190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044438788283711903,
+      "loss": 2.1551,
+      "step": 8191
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004442799277145688,
+      "loss": 2.0485,
+      "step": 8192
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044417197522213405,
+      "loss": 2.0911,
+      "step": 8193
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004440640253649105,
+      "loss": 2.0424,
+      "step": 8194
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044395607814799357,
+      "loss": 2.091,
+      "step": 8195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004438481335764787,
+      "loss": 2.0739,
+      "step": 8196
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044374019165546096,
+      "loss": 2.1464,
+      "step": 8197
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004436322523900357,
+      "loss": 1.9567,
+      "step": 8198
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004435243157852976,
+      "loss": 2.0229,
+      "step": 8199
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044341638184634173,
+      "loss": 2.09,
+      "step": 8200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004433084505782627,
+      "loss": 2.069,
+      "step": 8201
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004432005219861553,
+      "loss": 1.9175,
+      "step": 8202
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044309259607511377,
+      "loss": 1.975,
+      "step": 8203
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004429846728502326,
+      "loss": 2.0856,
+      "step": 8204
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000442876752316606,
+      "loss": 2.0922,
+      "step": 8205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044276883447932815,
+      "loss": 1.8738,
+      "step": 8206
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004426609193434927,
+      "loss": 2.0056,
+      "step": 8207
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000442553006914194,
+      "loss": 2.0076,
+      "step": 8208
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044244509719652527,
+      "loss": 1.9986,
+      "step": 8209
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044233719019558026,
+      "loss": 2.0749,
+      "step": 8210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044222928591645266,
+      "loss": 2.0264,
+      "step": 8211
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044212138436423555,
+      "loss": 2.0569,
+      "step": 8212
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004420134855440222,
+      "loss": 2.1,
+      "step": 8213
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044190558946090563,
+      "loss": 2.0601,
+      "step": 8214
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000441797696119979,
+      "loss": 2.1143,
+      "step": 8215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004416898055263348,
+      "loss": 1.991,
+      "step": 8216
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044158191768506593,
+      "loss": 2.0608,
+      "step": 8217
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044147403260126486,
+      "loss": 1.9694,
+      "step": 8218
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044136615028002416,
+      "loss": 2.118,
+      "step": 8219
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004412582707264358,
+      "loss": 2.0134,
+      "step": 8220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004411503939455923,
+      "loss": 2.031,
+      "step": 8221
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004410425199425855,
+      "loss": 2.2464,
+      "step": 8222
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004409346487225074,
+      "loss": 1.9168,
+      "step": 8223
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004408267802904496,
+      "loss": 2.0861,
+      "step": 8224
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044071891465150394,
+      "loss": 2.0343,
+      "step": 8225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004406110518107617,
+      "loss": 1.9313,
+      "step": 8226
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044050319177331453,
+      "loss": 2.0364,
+      "step": 8227
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044039533454425347,
+      "loss": 2.2105,
+      "step": 8228
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044028748012866977,
+      "loss": 2.1424,
+      "step": 8229
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004401796285316542,
+      "loss": 2.1134,
+      "step": 8230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00044007177975829786,
+      "loss": 2.0176,
+      "step": 8231
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004399639338136912,
+      "loss": 2.2478,
+      "step": 8232
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000439856090702925,
+      "loss": 2.0515,
+      "step": 8233
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004397482504310896,
+      "loss": 2.1137,
+      "step": 8234
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043964041300327544,
+      "loss": 2.0502,
+      "step": 8235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004395325784245725,
+      "loss": 2.0843,
+      "step": 8236
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004394247467000709,
+      "loss": 2.0953,
+      "step": 8237
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004393169178348606,
+      "loss": 2.2986,
+      "step": 8238
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043920909183403135,
+      "loss": 2.1058,
+      "step": 8239
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004391012687026727,
+      "loss": 1.9386,
+      "step": 8240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004389934484458743,
+      "loss": 1.9604,
+      "step": 8241
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004388856310687253,
+      "loss": 2.1027,
+      "step": 8242
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004387778165763152,
+      "loss": 2.0372,
+      "step": 8243
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043867000497373285,
+      "loss": 2.0655,
+      "step": 8244
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043856219626606734,
+      "loss": 2.1448,
+      "step": 8245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043845439045840743,
+      "loss": 2.1684,
+      "step": 8246
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043834658755584186,
+      "loss": 2.0691,
+      "step": 8247
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043823878756345914,
+      "loss": 1.8283,
+      "step": 8248
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004381309904863477,
+      "loss": 2.1595,
+      "step": 8249
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004380231963295957,
+      "loss": 2.0308,
+      "step": 8250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043791540509829143,
+      "loss": 2.0486,
+      "step": 8251
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043780761679752274,
+      "loss": 1.966,
+      "step": 8252
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004376998314323777,
+      "loss": 2.1643,
+      "step": 8253
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004375920490079437,
+      "loss": 2.0353,
+      "step": 8254
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004374842695293087,
+      "loss": 1.9992,
+      "step": 8255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004373764930015598,
+      "loss": 2.017,
+      "step": 8256
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004372687194297845,
+      "loss": 2.1053,
+      "step": 8257
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004371609488190699,
+      "loss": 2.1816,
+      "step": 8258
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043705318117450304,
+      "loss": 1.8956,
+      "step": 8259
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004369454165011707,
+      "loss": 1.8819,
+      "step": 8260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004368376548041598,
+      "loss": 2.0344,
+      "step": 8261
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004367298960885568,
+      "loss": 2.0707,
+      "step": 8262
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043662214035944825,
+      "loss": 2.0361,
+      "step": 8263
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043651438762192034,
+      "loss": 2.0824,
+      "step": 8264
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004364066378810595,
+      "loss": 2.0784,
+      "step": 8265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004362988911419514,
+      "loss": 1.9507,
+      "step": 8266
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004361911474096822,
+      "loss": 2.1106,
+      "step": 8267
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004360834066893376,
+      "loss": 1.944,
+      "step": 8268
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043597566898600327,
+      "loss": 2.0734,
+      "step": 8269
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043586793430476446,
+      "loss": 2.0436,
+      "step": 8270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0004357602026507067,
+      "loss": 2.1833,
+      "step": 8271
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043565247402891505,
+      "loss": 2.0217,
+      "step": 8272
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00043554474844447475,
+      "loss": 2.0172,
+      "step": 8273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004354370259024704,
+      "loss": 2.0277,
+      "step": 8274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043532930640798696,
+      "loss": 2.0633,
+      "step": 8275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004352215899661089,
+      "loss": 1.9496,
+      "step": 8276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004351138765819209,
+      "loss": 1.9097,
+      "step": 8277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043500616626050703,
+      "loss": 1.9616,
+      "step": 8278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043489845900695167,
+      "loss": 2.1954,
+      "step": 8279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043479075482633864,
+      "loss": 2.1949,
+      "step": 8280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043468305372375206,
+      "loss": 2.0519,
+      "step": 8281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004345753557042754,
+      "loss": 2.0683,
+      "step": 8282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004344676607729926,
+      "loss": 2.0811,
+      "step": 8283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043435996893498676,
+      "loss": 2.1922,
+      "step": 8284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043425228019534147,
+      "loss": 1.9727,
+      "step": 8285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043414459455913963,
+      "loss": 2.1086,
+      "step": 8286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043403691203146446,
+      "loss": 2.0515,
+      "step": 8287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043392923261739865,
+      "loss": 2.1149,
+      "step": 8288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043382155632202516,
+      "loss": 2.0757,
+      "step": 8289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043371388315042623,
+      "loss": 1.9667,
+      "step": 8290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043360621310768456,
+      "loss": 2.1701,
+      "step": 8291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004334985461988822,
+      "loss": 1.9201,
+      "step": 8292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004333908824291015,
+      "loss": 1.9416,
+      "step": 8293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004332832218034242,
+      "loss": 2.1534,
+      "step": 8294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004331755643269324,
+      "loss": 2.0528,
+      "step": 8295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004330679100047076,
+      "loss": 2.0867,
+      "step": 8296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043296025884183135,
+      "loss": 1.9802,
+      "step": 8297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000432852610843385,
+      "loss": 1.9706,
+      "step": 8298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043274496601444995,
+      "loss": 1.9437,
+      "step": 8299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043263732436010706,
+      "loss": 2.0005,
+      "step": 8300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004325296858854374,
+      "loss": 2.3119,
+      "step": 8301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043242205059552176,
+      "loss": 2.2409,
+      "step": 8302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043231441849544066,
+      "loss": 2.1218,
+      "step": 8303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004322067895902748,
+      "loss": 2.1583,
+      "step": 8304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004320991638851043,
+      "loss": 1.991,
+      "step": 8305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004319915413850094,
+      "loss": 2.0783,
+      "step": 8306
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004318839220950701,
+      "loss": 2.095,
+      "step": 8307
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004317763060203664,
+      "loss": 1.9902,
+      "step": 8308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000431668693165978,
+      "loss": 2.0471,
+      "step": 8309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004315610835369843,
+      "loss": 1.9796,
+      "step": 8310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004314534771384649,
+      "loss": 2.1229,
+      "step": 8311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004313458739754991,
+      "loss": 2.1532,
+      "step": 8312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043123827405316576,
+      "loss": 2.0306,
+      "step": 8313
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004311306773765441,
+      "loss": 1.9112,
+      "step": 8314
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004310230839507128,
+      "loss": 1.9718,
+      "step": 8315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004309154937807507,
+      "loss": 1.976,
+      "step": 8316
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000430807906871736,
+      "loss": 2.1327,
+      "step": 8317
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004307003232287473,
+      "loss": 1.9974,
+      "step": 8318
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043059274285686265,
+      "loss": 1.9618,
+      "step": 8319
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043048516576116026,
+      "loss": 2.3466,
+      "step": 8320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004303775919467178,
+      "loss": 2.124,
+      "step": 8321
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00043027002141861315,
+      "loss": 2.0292,
+      "step": 8322
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004301624541819238,
+      "loss": 2.1248,
+      "step": 8323
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004300548902417273,
+      "loss": 2.1877,
+      "step": 8324
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004299473296031008,
+      "loss": 2.0067,
+      "step": 8325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042983977227112146,
+      "loss": 1.9484,
+      "step": 8326
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004297322182508661,
+      "loss": 2.093,
+      "step": 8327
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004296246675474118,
+      "loss": 2.1849,
+      "step": 8328
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004295171201658349,
+      "loss": 1.9403,
+      "step": 8329
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042940957611121213,
+      "loss": 1.9222,
+      "step": 8330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042930203538861956,
+      "loss": 2.0592,
+      "step": 8331
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042919449800313367,
+      "loss": 2.0836,
+      "step": 8332
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042908696395983016,
+      "loss": 2.0008,
+      "step": 8333
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004289794332637851,
+      "loss": 1.9967,
+      "step": 8334
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004288719059200741,
+      "loss": 1.9283,
+      "step": 8335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004287643819337727,
+      "loss": 1.9486,
+      "step": 8336
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004286568613099562,
+      "loss": 2.0266,
+      "step": 8337
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004285493440537001,
+      "loss": 2.0327,
+      "step": 8338
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042844183017007916,
+      "loss": 2.0746,
+      "step": 8339
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042833431966416845,
+      "loss": 2.0266,
+      "step": 8340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042822681254104256,
+      "loss": 2.0718,
+      "step": 8341
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042811930880577633,
+      "loss": 2.0139,
+      "step": 8342
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004280118084634439,
+      "loss": 2.0561,
+      "step": 8343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042790431151911975,
+      "loss": 2.0332,
+      "step": 8344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004277968179778778,
+      "loss": 2.0217,
+      "step": 8345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042768932784479214,
+      "loss": 2.0556,
+      "step": 8346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004275818411249365,
+      "loss": 2.2187,
+      "step": 8347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004274743578233845,
+      "loss": 2.1144,
+      "step": 8348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004273668779452095,
+      "loss": 2.1903,
+      "step": 8349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000427259401495485,
+      "loss": 2.2938,
+      "step": 8350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000427151928479284,
+      "loss": 2.0419,
+      "step": 8351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042704445890167953,
+      "loss": 2.0211,
+      "step": 8352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004269369927677443,
+      "loss": 2.0247,
+      "step": 8353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004268295300825511,
+      "loss": 1.9364,
+      "step": 8354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004267220708511723,
+      "loss": 2.0643,
+      "step": 8355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004266146150786802,
+      "loss": 2.1598,
+      "step": 8356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004265071627701471,
+      "loss": 2.0596,
+      "step": 8357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004263997139306449,
+      "loss": 2.0819,
+      "step": 8358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004262922685652456,
+      "loss": 1.9926,
+      "step": 8359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042618482667902056,
+      "loss": 2.0237,
+      "step": 8360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042607738827704156,
+      "loss": 1.9995,
+      "step": 8361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004259699533643798,
+      "loss": 2.0869,
+      "step": 8362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004258625219461065,
+      "loss": 2.1683,
+      "step": 8363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004257550940272926,
+      "loss": 2.0257,
+      "step": 8364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042564766961300916,
+      "loss": 2.1108,
+      "step": 8365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004255402487083265,
+      "loss": 2.0657,
+      "step": 8366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042543283131831546,
+      "loss": 2.0706,
+      "step": 8367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004253254174480462,
+      "loss": 2.149,
+      "step": 8368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000425218007102589,
+      "loss": 2.1409,
+      "step": 8369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004251106002870138,
+      "loss": 2.0016,
+      "step": 8370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042500319700639055,
+      "loss": 2.209,
+      "step": 8371
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004248957972657888,
+      "loss": 2.0733,
+      "step": 8372
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004247884010702782,
+      "loss": 2.0363,
+      "step": 8373
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042468100842492796,
+      "loss": 1.9957,
+      "step": 8374
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004245736193348074,
+      "loss": 1.9654,
+      "step": 8375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004244662338049854,
+      "loss": 2.1456,
+      "step": 8376
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042435885184053095,
+      "loss": 2.1006,
+      "step": 8377
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042425147344651257,
+      "loss": 2.0029,
+      "step": 8378
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004241440986279988,
+      "loss": 2.109,
+      "step": 8379
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004240367273900581,
+      "loss": 2.068,
+      "step": 8380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004239293597377585,
+      "loss": 2.0036,
+      "step": 8381
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004238219956761682,
+      "loss": 2.0544,
+      "step": 8382
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004237146352103547,
+      "loss": 2.0404,
+      "step": 8383
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042360727834538596,
+      "loss": 2.0418,
+      "step": 8384
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004234999250863293,
+      "loss": 2.0821,
+      "step": 8385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004233925754382522,
+      "loss": 2.1012,
+      "step": 8386
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004232852294062216,
+      "loss": 2.0678,
+      "step": 8387
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004231778869953047,
+      "loss": 2.0568,
+      "step": 8388
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004230705482105681,
+      "loss": 2.2133,
+      "step": 8389
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042296321305707865,
+      "loss": 1.9897,
+      "step": 8390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004228558815399025,
+      "loss": 1.9895,
+      "step": 8391
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042274855366410634,
+      "loss": 2.1543,
+      "step": 8392
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042264122943475605,
+      "loss": 1.9837,
+      "step": 8393
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004225339088569176,
+      "loss": 2.1487,
+      "step": 8394
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004224265919356568,
+      "loss": 2.1633,
+      "step": 8395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004223192786760393,
+      "loss": 2.1644,
+      "step": 8396
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042221196908313045,
+      "loss": 1.9736,
+      "step": 8397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004221046631619956,
+      "loss": 2.02,
+      "step": 8398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004219973609176996,
+      "loss": 2.1069,
+      "step": 8399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004218900623553078,
+      "loss": 1.9245,
+      "step": 8400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004217827674798845,
+      "loss": 2.1259,
+      "step": 8401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042167547629649453,
+      "loss": 1.9153,
+      "step": 8402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042156818881020216,
+      "loss": 2.0724,
+      "step": 8403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004214609050260717,
+      "loss": 2.178,
+      "step": 8404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004213536249491671,
+      "loss": 2.0599,
+      "step": 8405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004212463485845523,
+      "loss": 1.9627,
+      "step": 8406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004211390759372909,
+      "loss": 1.9187,
+      "step": 8407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042103180701244655,
+      "loss": 2.127,
+      "step": 8408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042092454181508237,
+      "loss": 2.1073,
+      "step": 8409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004208172803502618,
+      "loss": 2.0703,
+      "step": 8410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042071002262304756,
+      "loss": 2.1257,
+      "step": 8411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004206027686385028,
+      "loss": 2.1911,
+      "step": 8412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042049551840168974,
+      "loss": 1.9337,
+      "step": 8413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042038827191767116,
+      "loss": 1.9906,
+      "step": 8414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004202810291915091,
+      "loss": 1.8548,
+      "step": 8415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00042017379022826594,
+      "loss": 2.106,
+      "step": 8416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004200665550330034,
+      "loss": 1.9329,
+      "step": 8417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004199593236107833,
+      "loss": 2.148,
+      "step": 8418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004198520959666672,
+      "loss": 1.9891,
+      "step": 8419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004197448721057164,
+      "loss": 2.0024,
+      "step": 8420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00041963765203299226,
+      "loss": 2.0902,
+      "step": 8421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00041953043575355584,
+      "loss": 2.0156,
+      "step": 8422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0004194232232724677,
+      "loss": 1.9822,
+      "step": 8423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004193160145947889,
+      "loss": 2.2152,
+      "step": 8424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004192088097255796,
+      "loss": 2.0649,
+      "step": 8425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004191016086699005,
+      "loss": 2.0081,
+      "step": 8426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041899441143281133,
+      "loss": 2.0192,
+      "step": 8427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004188872180193723,
+      "loss": 2.0915,
+      "step": 8428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041878002843464304,
+      "loss": 1.9709,
+      "step": 8429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041867284268368334,
+      "loss": 2.1759,
+      "step": 8430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041856566077155236,
+      "loss": 1.9624,
+      "step": 8431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041845848270330953,
+      "loss": 2.0243,
+      "step": 8432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004183513084840138,
+      "loss": 1.9821,
+      "step": 8433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041824413811872416,
+      "loss": 1.9564,
+      "step": 8434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041813697161249906,
+      "loss": 2.0484,
+      "step": 8435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041802980897039734,
+      "loss": 1.9394,
+      "step": 8436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000417922650197477,
+      "loss": 2.1105,
+      "step": 8437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004178154952987964,
+      "loss": 1.9629,
+      "step": 8438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041770834427941333,
+      "loss": 2.015,
+      "step": 8439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041760119714438575,
+      "loss": 2.105,
+      "step": 8440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004174940538987711,
+      "loss": 2.1186,
+      "step": 8441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041738691454762687,
+      "loss": 2.0286,
+      "step": 8442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004172797790960102,
+      "loss": 2.0627,
+      "step": 8443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004171726475489783,
+      "loss": 2.1104,
+      "step": 8444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004170655199115878,
+      "loss": 1.9784,
+      "step": 8445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041695839618889564,
+      "loss": 2.0733,
+      "step": 8446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000416851276385958,
+      "loss": 2.1713,
+      "step": 8447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004167441605078314,
+      "loss": 1.9941,
+      "step": 8448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004166370485595718,
+      "loss": 2.1922,
+      "step": 8449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041652994054623525,
+      "loss": 1.9747,
+      "step": 8450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041642283647287746,
+      "loss": 2.1497,
+      "step": 8451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004163157363445541,
+      "loss": 2.1781,
+      "step": 8452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004162086401663203,
+      "loss": 2.0354,
+      "step": 8453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004161015479432314,
+      "loss": 2.1823,
+      "step": 8454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004159944596803424,
+      "loss": 2.1064,
+      "step": 8455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004158873753827081,
+      "loss": 2.0395,
+      "step": 8456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041578029505538305,
+      "loss": 2.0956,
+      "step": 8457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041567321870342183,
+      "loss": 2.1118,
+      "step": 8458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004155661463318785,
+      "loss": 1.9896,
+      "step": 8459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004154590779458074,
+      "loss": 1.9192,
+      "step": 8460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004153520135502621,
+      "loss": 2.0714,
+      "step": 8461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041524495315029653,
+      "loss": 2.0103,
+      "step": 8462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004151378967509639,
+      "loss": 2.143,
+      "step": 8463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041503084435731794,
+      "loss": 1.9959,
+      "step": 8464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004149237959744113,
+      "loss": 2.0879,
+      "step": 8465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004148167516072973,
+      "loss": 2.0815,
+      "step": 8466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004147097112610284,
+      "loss": 1.9658,
+      "step": 8467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041460267494065737,
+      "loss": 2.1119,
+      "step": 8468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004144956426512364,
+      "loss": 1.9743,
+      "step": 8469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041438861439781774,
+      "loss": 2.0839,
+      "step": 8470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004142815901854534,
+      "loss": 2.0786,
+      "step": 8471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041417457001919505,
+      "loss": 1.9858,
+      "step": 8472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041406755390409447,
+      "loss": 2.0059,
+      "step": 8473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004139605418452029,
+      "loss": 2.1586,
+      "step": 8474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004138535338475716,
+      "loss": 1.8632,
+      "step": 8475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004137465299162516,
+      "loss": 1.9192,
+      "step": 8476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004136395300562938,
+      "loss": 2.0723,
+      "step": 8477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004135325342727488,
+      "loss": 2.1748,
+      "step": 8478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000413425542570667,
+      "loss": 2.1508,
+      "step": 8479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004133185549550986,
+      "loss": 2.244,
+      "step": 8480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041321157143109393,
+      "loss": 1.9143,
+      "step": 8481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041310459200370254,
+      "loss": 2.0174,
+      "step": 8482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041299761667797433,
+      "loss": 1.9426,
+      "step": 8483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041289064545895854,
+      "loss": 1.989,
+      "step": 8484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041278367835170473,
+      "loss": 1.9368,
+      "step": 8485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041267671536126177,
+      "loss": 1.9765,
+      "step": 8486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041256975649267883,
+      "loss": 2.0407,
+      "step": 8487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041246280175100423,
+      "loss": 1.983,
+      "step": 8488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004123558511412869,
+      "loss": 2.0178,
+      "step": 8489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004122489046685748,
+      "loss": 2.0451,
+      "step": 8490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004121419623379162,
+      "loss": 2.1097,
+      "step": 8491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041203502415435906,
+      "loss": 1.9883,
+      "step": 8492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004119280901229512,
+      "loss": 2.1012,
+      "step": 8493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041182116024873983,
+      "loss": 2.2592,
+      "step": 8494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004117142345367726,
+      "loss": 2.0362,
+      "step": 8495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004116073129920964,
+      "loss": 2.051,
+      "step": 8496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041150039561975844,
+      "loss": 2.0224,
+      "step": 8497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004113934824248052,
+      "loss": 2.0831,
+      "step": 8498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004112865734122835,
+      "loss": 1.9712,
+      "step": 8499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004111796685872394,
+      "loss": 2.0174,
+      "step": 8500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004110727679547193,
+      "loss": 1.9384,
+      "step": 8501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000410965871519769,
+      "loss": 2.1767,
+      "step": 8502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041085897928743446,
+      "loss": 2.0959,
+      "step": 8503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041075209126276087,
+      "loss": 2.28,
+      "step": 8504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041064520745079397,
+      "loss": 1.9945,
+      "step": 8505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041053832785657864,
+      "loss": 2.1506,
+      "step": 8506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041043145248516013,
+      "loss": 2.1377,
+      "step": 8507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004103245813415829,
+      "loss": 2.0672,
+      "step": 8508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004102177144308917,
+      "loss": 2.1067,
+      "step": 8509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041011085175813087,
+      "loss": 2.1282,
+      "step": 8510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00041000399332834453,
+      "loss": 2.1022,
+      "step": 8511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004098971391465767,
+      "loss": 2.1609,
+      "step": 8512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004097902892178711,
+      "loss": 2.0081,
+      "step": 8513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004096834435472712,
+      "loss": 2.0281,
+      "step": 8514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004095766021398206,
+      "loss": 1.9948,
+      "step": 8515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004094697650005622,
+      "loss": 2.0882,
+      "step": 8516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040936293213453923,
+      "loss": 2.1232,
+      "step": 8517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004092561035467942,
+      "loss": 2.1176,
+      "step": 8518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004091492792423698,
+      "loss": 2.0481,
+      "step": 8519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040904245922630827,
+      "loss": 1.9576,
+      "step": 8520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000408935643503652,
+      "loss": 2.1758,
+      "step": 8521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004088288320794426,
+      "loss": 2.0302,
+      "step": 8522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040872202495872214,
+      "loss": 1.9742,
+      "step": 8523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040861522214653183,
+      "loss": 2.1203,
+      "step": 8524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040850842364791343,
+      "loss": 1.9541,
+      "step": 8525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004084016294679076,
+      "loss": 1.9792,
+      "step": 8526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040829483961155565,
+      "loss": 2.0259,
+      "step": 8527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004081880540838981,
+      "loss": 2.0154,
+      "step": 8528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040808127288997556,
+      "loss": 2.0711,
+      "step": 8529
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004079744960348283,
+      "loss": 2.0609,
+      "step": 8530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004078677235234965,
+      "loss": 2.1036,
+      "step": 8531
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004077609553610199,
+      "loss": 2.0022,
+      "step": 8532
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040765419155243854,
+      "loss": 2.0488,
+      "step": 8533
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004075474321027915,
+      "loss": 1.9713,
+      "step": 8534
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004074406770171184,
+      "loss": 2.2318,
+      "step": 8535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004073339263004581,
+      "loss": 1.9436,
+      "step": 8536
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040722717995784975,
+      "loss": 2.093,
+      "step": 8537
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040712043799433163,
+      "loss": 1.925,
+      "step": 8538
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040701370041494264,
+      "loss": 2.244,
+      "step": 8539
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040690696722472063,
+      "loss": 2.0642,
+      "step": 8540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040680023842870405,
+      "loss": 1.9579,
+      "step": 8541
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004066935140319304,
+      "loss": 2.1108,
+      "step": 8542
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040658679403943757,
+      "loss": 2.0102,
+      "step": 8543
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004064800784562627,
+      "loss": 1.9784,
+      "step": 8544
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040637336728744337,
+      "loss": 2.0474,
+      "step": 8545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004062666605380164,
+      "loss": 2.0363,
+      "step": 8546
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040615995821301855,
+      "loss": 2.1501,
+      "step": 8547
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040605326031748645,
+      "loss": 2.1146,
+      "step": 8548
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004059465668564566,
+      "loss": 2.142,
+      "step": 8549
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000405839877834965,
+      "loss": 2.0665,
+      "step": 8550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040573319325804783,
+      "loss": 2.0349,
+      "step": 8551
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004056265131307405,
+      "loss": 2.0097,
+      "step": 8552
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000405519837458079,
+      "loss": 2.0333,
+      "step": 8553
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004054131662450983,
+      "loss": 1.9631,
+      "step": 8554
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004053064994968338,
+      "loss": 1.936,
+      "step": 8555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004051998372183201,
+      "loss": 1.9385,
+      "step": 8556
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040509317941459227,
+      "loss": 2.0705,
+      "step": 8557
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040498652609068455,
+      "loss": 1.9189,
+      "step": 8558
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040487987725163133,
+      "loss": 2.0521,
+      "step": 8559
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004047732329024666,
+      "loss": 2.136,
+      "step": 8560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004046665930482242,
+      "loss": 2.0143,
+      "step": 8561
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040455995769393807,
+      "loss": 1.8744,
+      "step": 8562
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004044533268446412,
+      "loss": 1.9494,
+      "step": 8563
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004043467005053671,
+      "loss": 2.0983,
+      "step": 8564
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004042400786811488,
+      "loss": 2.0451,
+      "step": 8565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004041334613770189,
+      "loss": 2.1547,
+      "step": 8566
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004040268485980101,
+      "loss": 1.9779,
+      "step": 8567
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004039202403491548,
+      "loss": 1.9086,
+      "step": 8568
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040381363663548504,
+      "loss": 2.1003,
+      "step": 8569
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040370703746203294,
+      "loss": 2.1026,
+      "step": 8570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00040360044283383,
+      "loss": 2.0627,
+      "step": 8571
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0004034938527559079,
+      "loss": 1.8915,
+      "step": 8572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004033872672332978,
+      "loss": 1.9391,
+      "step": 8573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000403280686271031,
+      "loss": 2.0317,
+      "step": 8574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040317410987413807,
+      "loss": 2.0165,
+      "step": 8575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040306753804764995,
+      "loss": 2.0772,
+      "step": 8576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040296097079659687,
+      "loss": 2.0307,
+      "step": 8577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040285440812600924,
+      "loss": 2.0484,
+      "step": 8578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004027478500409168,
+      "loss": 2.0257,
+      "step": 8579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004026412965463496,
+      "loss": 1.8999,
+      "step": 8580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000402534747647337,
+      "loss": 1.945,
+      "step": 8581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004024282033489086,
+      "loss": 2.0341,
+      "step": 8582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004023216636560932,
+      "loss": 2.07,
+      "step": 8583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040221512857392005,
+      "loss": 2.1013,
+      "step": 8584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040210859810741763,
+      "loss": 2.0509,
+      "step": 8585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040200207226161454,
+      "loss": 2.0971,
+      "step": 8586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000401895551041539,
+      "loss": 2.1329,
+      "step": 8587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000401789034452219,
+      "loss": 2.0589,
+      "step": 8588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040168252249868244,
+      "loss": 1.9883,
+      "step": 8589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000401576015185957,
+      "loss": 2.0197,
+      "step": 8590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004014695125190699,
+      "loss": 1.9991,
+      "step": 8591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040136301450304846,
+      "loss": 2.1611,
+      "step": 8592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040125652114291946,
+      "loss": 2.1895,
+      "step": 8593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040115003244370985,
+      "loss": 2.0913,
+      "step": 8594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000401043548410446,
+      "loss": 2.0615,
+      "step": 8595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040093706904815427,
+      "loss": 2.0017,
+      "step": 8596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004008305943618606,
+      "loss": 2.2213,
+      "step": 8597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000400724124356591,
+      "loss": 2.0176,
+      "step": 8598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000400617659037371,
+      "loss": 2.0954,
+      "step": 8599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00040051119840922605,
+      "loss": 2.0659,
+      "step": 8600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004004047424771813,
+      "loss": 2.0676,
+      "step": 8601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004002982912462618,
+      "loss": 2.1389,
+      "step": 8602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004001918447214922,
+      "loss": 1.9123,
+      "step": 8603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0004000854029078971,
+      "loss": 2.0002,
+      "step": 8604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039997896581050064,
+      "loss": 2.1756,
+      "step": 8605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039987253343432716,
+      "loss": 2.0147,
+      "step": 8606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039976610578440025,
+      "loss": 1.9681,
+      "step": 8607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039965968286574373,
+      "loss": 1.8924,
+      "step": 8608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003995532646833808,
+      "loss": 1.8718,
+      "step": 8609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039944685124233495,
+      "loss": 2.049,
+      "step": 8610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003993404425476288,
+      "loss": 2.0187,
+      "step": 8611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003992340386042854,
+      "loss": 1.9896,
+      "step": 8612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000399127639417327,
+      "loss": 2.023,
+      "step": 8613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039902124499177605,
+      "loss": 2.0279,
+      "step": 8614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039891485533265447,
+      "loss": 2.1258,
+      "step": 8615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039880847044498425,
+      "loss": 2.0153,
+      "step": 8616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003987020903337869,
+      "loss": 2.0326,
+      "step": 8617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003985957150040839,
+      "loss": 2.1111,
+      "step": 8618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039848934446089625,
+      "loss": 1.9508,
+      "step": 8619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039838297870924507,
+      "loss": 2.0919,
+      "step": 8620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003982766177541509,
+      "loss": 2.0604,
+      "step": 8621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039817026160063443,
+      "loss": 2.0084,
+      "step": 8622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039806391025371565,
+      "loss": 2.0116,
+      "step": 8623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003979575637184148,
+      "loss": 1.9303,
+      "step": 8624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039785122199975155,
+      "loss": 2.0665,
+      "step": 8625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039774488510274566,
+      "loss": 2.1432,
+      "step": 8626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039763855303241616,
+      "loss": 1.9305,
+      "step": 8627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003975322257937825,
+      "loss": 1.9799,
+      "step": 8628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039742590339186335,
+      "loss": 2.2258,
+      "step": 8629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003973195858316775,
+      "loss": 2.0724,
+      "step": 8630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039721327311824324,
+      "loss": 2.1223,
+      "step": 8631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000397106965256579,
+      "loss": 1.9318,
+      "step": 8632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003970006622517025,
+      "loss": 2.1731,
+      "step": 8633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039689436410863166,
+      "loss": 1.9961,
+      "step": 8634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039678807083238387,
+      "loss": 2.1637,
+      "step": 8635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039668178242797665,
+      "loss": 1.9565,
+      "step": 8636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039657549890042675,
+      "loss": 2.1239,
+      "step": 8637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003964692202547512,
+      "loss": 2.0064,
+      "step": 8638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003963629464959665,
+      "loss": 2.0085,
+      "step": 8639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003962566776290892,
+      "loss": 2.0006,
+      "step": 8640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003961504136591351,
+      "loss": 2.0964,
+      "step": 8641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003960441545911204,
+      "loss": 2.0557,
+      "step": 8642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003959379004300606,
+      "loss": 2.1699,
+      "step": 8643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039583165118097137,
+      "loss": 2.1476,
+      "step": 8644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003957254068488676,
+      "loss": 1.9582,
+      "step": 8645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039561916743876445,
+      "loss": 2.0313,
+      "step": 8646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003955129329556766,
+      "loss": 2.0592,
+      "step": 8647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003954067034046188,
+      "loss": 2.0074,
+      "step": 8648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039530047879060494,
+      "loss": 1.9989,
+      "step": 8649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003951942591186492,
+      "loss": 1.9457,
+      "step": 8650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003950880443937656,
+      "loss": 2.0331,
+      "step": 8651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003949818346209675,
+      "loss": 1.9541,
+      "step": 8652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039487562980526837,
+      "loss": 2.0352,
+      "step": 8653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003947694299516812,
+      "loss": 1.9653,
+      "step": 8654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000394663235065219,
+      "loss": 2.0649,
+      "step": 8655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003945570451508942,
+      "loss": 1.9229,
+      "step": 8656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039445086021371956,
+      "loss": 2.042,
+      "step": 8657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003943446802587069,
+      "loss": 2.1668,
+      "step": 8658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039423850529086843,
+      "loss": 1.9755,
+      "step": 8659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003941323353152156,
+      "loss": 2.1617,
+      "step": 8660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039402617033676006,
+      "loss": 2.1066,
+      "step": 8661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003939200103605129,
+      "loss": 2.0638,
+      "step": 8662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003938138553914854,
+      "loss": 1.9769,
+      "step": 8663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003937077054346879,
+      "loss": 1.9818,
+      "step": 8664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003936015604951313,
+      "loss": 2.014,
+      "step": 8665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003934954205778256,
+      "loss": 2.0226,
+      "step": 8666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003933892856877811,
+      "loss": 2.0272,
+      "step": 8667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003932831558300074,
+      "loss": 2.2035,
+      "step": 8668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039317703100951416,
+      "loss": 2.0013,
+      "step": 8669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003930709112313107,
+      "loss": 2.056,
+      "step": 8670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039296479650040625,
+      "loss": 2.0459,
+      "step": 8671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039285868682180944,
+      "loss": 2.0334,
+      "step": 8672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039275258220052906,
+      "loss": 1.9455,
+      "step": 8673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003926464826415733,
+      "loss": 2.2463,
+      "step": 8674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039254038814995064,
+      "loss": 1.9551,
+      "step": 8675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003924342987306686,
+      "loss": 2.1436,
+      "step": 8676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039232821438873523,
+      "loss": 2.1522,
+      "step": 8677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039222213512915755,
+      "loss": 2.2549,
+      "step": 8678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003921160609569431,
+      "loss": 2.1315,
+      "step": 8679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039200999187709854,
+      "loss": 1.9403,
+      "step": 8680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003919039278946308,
+      "loss": 2.1154,
+      "step": 8681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003917978690145462,
+      "loss": 2.2072,
+      "step": 8682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000391691815241851,
+      "loss": 2.0234,
+      "step": 8683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003915857665815512,
+      "loss": 2.0752,
+      "step": 8684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003914797230386526,
+      "loss": 2.1692,
+      "step": 8685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039137368461816054,
+      "loss": 1.9957,
+      "step": 8686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039126765132508045,
+      "loss": 2.0563,
+      "step": 8687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039116162316441715,
+      "loss": 1.9806,
+      "step": 8688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003910556001411757,
+      "loss": 2.0498,
+      "step": 8689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003909495822603603,
+      "loss": 2.1922,
+      "step": 8690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039084356952697553,
+      "loss": 2.0286,
+      "step": 8691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003907375619460252,
+      "loss": 2.024,
+      "step": 8692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003906315595225133,
+      "loss": 2.0405,
+      "step": 8693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003905255622614432,
+      "loss": 1.9286,
+      "step": 8694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003904195701678184,
+      "loss": 2.084,
+      "step": 8695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039031358324664183,
+      "loss": 2.0386,
+      "step": 8696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039020760150291643,
+      "loss": 2.0494,
+      "step": 8697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00039010162494164467,
+      "loss": 2.133,
+      "step": 8698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038999565356782906,
+      "loss": 1.9483,
+      "step": 8699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003898896873864714,
+      "loss": 2.0003,
+      "step": 8700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003897837264025739,
+      "loss": 2.1554,
+      "step": 8701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003896777706211378,
+      "loss": 2.1105,
+      "step": 8702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038957182004716474,
+      "loss": 1.9635,
+      "step": 8703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038946587468565564,
+      "loss": 2.0905,
+      "step": 8704
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038935993454161156,
+      "loss": 2.0238,
+      "step": 8705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038925399962003296,
+      "loss": 2.0417,
+      "step": 8706
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003891480699259202,
+      "loss": 2.1796,
+      "step": 8707
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003890421454642735,
+      "loss": 2.1955,
+      "step": 8708
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038893622624009273,
+      "loss": 2.0203,
+      "step": 8709
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038883031225837747,
+      "loss": 1.9364,
+      "step": 8710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038872440352412716,
+      "loss": 1.9102,
+      "step": 8711
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003886185000423408,
+      "loss": 2.0323,
+      "step": 8712
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003885126018180175,
+      "loss": 2.0396,
+      "step": 8713
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003884067088561556,
+      "loss": 1.999,
+      "step": 8714
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003883008211617539,
+      "loss": 2.036,
+      "step": 8715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038819493873981013,
+      "loss": 2.1857,
+      "step": 8716
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038808906159532253,
+      "loss": 2.098,
+      "step": 8717
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003879831897332884,
+      "loss": 2.1675,
+      "step": 8718
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003878773231587055,
+      "loss": 2.0208,
+      "step": 8719
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00038777146187657065,
+      "loss": 2.0673,
+      "step": 8720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000387665605891881,
+      "loss": 2.052,
+      "step": 8721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038755975520963294,
+      "loss": 2.0872,
+      "step": 8722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038745390983482316,
+      "loss": 1.8591,
+      "step": 8723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003873480697724475,
+      "loss": 2.1255,
+      "step": 8724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038724223502750217,
+      "loss": 2.0692,
+      "step": 8725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038713640560498254,
+      "loss": 2.1215,
+      "step": 8726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003870305815098842,
+      "loss": 1.9567,
+      "step": 8727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003869247627472021,
+      "loss": 2.0661,
+      "step": 8728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003868189493219313,
+      "loss": 2.0229,
+      "step": 8729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038671314123906635,
+      "loss": 2.0518,
+      "step": 8730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038660733850360174,
+      "loss": 1.9765,
+      "step": 8731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003865015411205314,
+      "loss": 2.0179,
+      "step": 8732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003863957490948494,
+      "loss": 2.0337,
+      "step": 8733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003862899624315492,
+      "loss": 2.2019,
+      "step": 8734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003861841811356245,
+      "loss": 2.029,
+      "step": 8735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038607840521206793,
+      "loss": 1.9989,
+      "step": 8736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003859726346658728,
+      "loss": 1.9856,
+      "step": 8737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038586686950203136,
+      "loss": 2.0598,
+      "step": 8738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038576110972553636,
+      "loss": 2.2377,
+      "step": 8739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003856553553413795,
+      "loss": 2.0177,
+      "step": 8740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038554960635455284,
+      "loss": 1.9687,
+      "step": 8741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003854438627700481,
+      "loss": 2.0674,
+      "step": 8742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003853381245928563,
+      "loss": 1.9074,
+      "step": 8743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003852323918279689,
+      "loss": 2.0996,
+      "step": 8744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003851266644803764,
+      "loss": 2.0305,
+      "step": 8745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003850209425550696,
+      "loss": 2.1279,
+      "step": 8746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003849152260570386,
+      "loss": 2.0163,
+      "step": 8747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003848095149912737,
+      "loss": 2.0427,
+      "step": 8748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003847038093627645,
+      "loss": 1.9675,
+      "step": 8749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038459810917650073,
+      "loss": 2.0151,
+      "step": 8750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038449241443747154,
+      "loss": 2.0207,
+      "step": 8751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003843867251506661,
+      "loss": 1.9172,
+      "step": 8752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038428104132107296,
+      "loss": 2.1237,
+      "step": 8753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038417536295368095,
+      "loss": 2.0301,
+      "step": 8754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000384069690053478,
+      "loss": 2.1169,
+      "step": 8755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038396402262545244,
+      "loss": 2.0623,
+      "step": 8756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003838583606745917,
+      "loss": 2.0825,
+      "step": 8757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003837527042058836,
+      "loss": 2.0774,
+      "step": 8758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000383647053224315,
+      "loss": 1.9325,
+      "step": 8759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003835414077348732,
+      "loss": 1.9954,
+      "step": 8760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003834357677425447,
+      "loss": 2.1363,
+      "step": 8761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003833301332523161,
+      "loss": 2.0013,
+      "step": 8762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038322450426917344,
+      "loss": 2.074,
+      "step": 8763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003831188807981028,
+      "loss": 1.9557,
+      "step": 8764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003830132628440897,
+      "loss": 2.0986,
+      "step": 8765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038290765041211966,
+      "loss": 2.0052,
+      "step": 8766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038280204350717774,
+      "loss": 2.0581,
+      "step": 8767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000382696442134249,
+      "loss": 2.0446,
+      "step": 8768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003825908462983178,
+      "loss": 2.2278,
+      "step": 8769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003824852560043688,
+      "loss": 2.0097,
+      "step": 8770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003823796712573858,
+      "loss": 2.1819,
+      "step": 8771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038227409206235296,
+      "loss": 2.1612,
+      "step": 8772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003821685184242536,
+      "loss": 2.1363,
+      "step": 8773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038206295034807125,
+      "loss": 1.9068,
+      "step": 8774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038195738783878874,
+      "loss": 2.0799,
+      "step": 8775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038185183090138914,
+      "loss": 1.9421,
+      "step": 8776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003817462795408547,
+      "loss": 1.9623,
+      "step": 8777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003816407337621679,
+      "loss": 2.1511,
+      "step": 8778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003815351935703105,
+      "loss": 2.1859,
+      "step": 8779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038142965897026456,
+      "loss": 2.1155,
+      "step": 8780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038132412996701127,
+      "loss": 2.0542,
+      "step": 8781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003812186065655321,
+      "loss": 2.1593,
+      "step": 8782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038111308877080763,
+      "loss": 2.0139,
+      "step": 8783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000381007576587819,
+      "loss": 1.9191,
+      "step": 8784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038090207002154625,
+      "loss": 2.037,
+      "step": 8785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003807965690769698,
+      "loss": 2.0576,
+      "step": 8786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038069107375906933,
+      "loss": 1.9909,
+      "step": 8787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003805855840728246,
+      "loss": 2.0749,
+      "step": 8788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038048010002321485,
+      "loss": 2.0131,
+      "step": 8789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003803746216152193,
+      "loss": 1.9837,
+      "step": 8790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038026914885381667,
+      "loss": 2.1233,
+      "step": 8791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038016368174398563,
+      "loss": 2.052,
+      "step": 8792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00038005822029070426,
+      "loss": 2.0892,
+      "step": 8793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003799527644989509,
+      "loss": 2.0964,
+      "step": 8794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037984731437370296,
+      "loss": 2.1301,
+      "step": 8795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003797418699199382,
+      "loss": 1.8605,
+      "step": 8796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003796364311426336,
+      "loss": 2.0255,
+      "step": 8797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037953099804676643,
+      "loss": 1.9147,
+      "step": 8798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037942557063731303,
+      "loss": 2.3038,
+      "step": 8799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003793201489192501,
+      "loss": 2.074,
+      "step": 8800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003792147328975536,
+      "loss": 2.1001,
+      "step": 8801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003791093225771995,
+      "loss": 1.958,
+      "step": 8802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037900391796316337,
+      "loss": 2.1283,
+      "step": 8803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037889851906042053,
+      "loss": 2.0344,
+      "step": 8804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003787931258739461,
+      "loss": 1.9443,
+      "step": 8805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000378687738408715,
+      "loss": 2.0254,
+      "step": 8806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003785823566697014,
+      "loss": 1.8224,
+      "step": 8807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037847698066187976,
+      "loss": 2.0408,
+      "step": 8808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037837161039022427,
+      "loss": 2.1074,
+      "step": 8809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003782662458597084,
+      "loss": 2.0485,
+      "step": 8810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037816088707530573,
+      "loss": 2.1767,
+      "step": 8811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003780555340419893,
+      "loss": 2.069,
+      "step": 8812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003779501867647322,
+      "loss": 2.0951,
+      "step": 8813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003778448452485069,
+      "loss": 2.0068,
+      "step": 8814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037773950949828583,
+      "loss": 2.1317,
+      "step": 8815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003776341795190411,
+      "loss": 2.115,
+      "step": 8816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037752885531574447,
+      "loss": 2.1138,
+      "step": 8817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003774235368933675,
+      "loss": 1.9578,
+      "step": 8818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037731822425688137,
+      "loss": 2.0447,
+      "step": 8819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003772129174112574,
+      "loss": 2.1276,
+      "step": 8820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037710761636146594,
+      "loss": 2.0109,
+      "step": 8821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037700232111247773,
+      "loss": 1.9927,
+      "step": 8822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003768970316692627,
+      "loss": 1.8428,
+      "step": 8823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037679174803679095,
+      "loss": 2.0231,
+      "step": 8824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000376686470220032,
+      "loss": 2.0768,
+      "step": 8825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003765811982239553,
+      "loss": 1.9867,
+      "step": 8826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003764759320535297,
+      "loss": 2.0527,
+      "step": 8827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037637067171372436,
+      "loss": 2.0853,
+      "step": 8828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003762654172095075,
+      "loss": 2.1079,
+      "step": 8829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003761601685458476,
+      "loss": 2.0247,
+      "step": 8830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003760549257277124,
+      "loss": 1.9454,
+      "step": 8831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037594968876006985,
+      "loss": 2.0606,
+      "step": 8832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037584445764788716,
+      "loss": 2.0566,
+      "step": 8833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037573923239613173,
+      "loss": 2.0978,
+      "step": 8834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037563401300977014,
+      "loss": 2.0869,
+      "step": 8835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003755287994937693,
+      "loss": 2.0231,
+      "step": 8836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037542359185309523,
+      "loss": 2.1805,
+      "step": 8837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003753183900927142,
+      "loss": 1.9866,
+      "step": 8838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037521319421759177,
+      "loss": 1.9803,
+      "step": 8839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037510800423269365,
+      "loss": 2.0591,
+      "step": 8840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037500282014298487,
+      "loss": 2.0816,
+      "step": 8841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037489764195343053,
+      "loss": 2.2059,
+      "step": 8842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037479246966899503,
+      "loss": 2.0288,
+      "step": 8843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037468730329464305,
+      "loss": 2.133,
+      "step": 8844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037458214283533843,
+      "loss": 1.9748,
+      "step": 8845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037447698829604515,
+      "loss": 2.041,
+      "step": 8846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003743718396817266,
+      "loss": 1.992,
+      "step": 8847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003742666969973463,
+      "loss": 2.039,
+      "step": 8848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037416156024786684,
+      "loss": 1.7912,
+      "step": 8849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037405642943825123,
+      "loss": 2.0612,
+      "step": 8850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037395130457346175,
+      "loss": 2.0074,
+      "step": 8851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003738461856584606,
+      "loss": 2.0177,
+      "step": 8852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037374107269820957,
+      "loss": 2.1191,
+      "step": 8853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037363596569767036,
+      "loss": 2.0836,
+      "step": 8854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037353086466180397,
+      "loss": 2.0052,
+      "step": 8855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037342576959557177,
+      "loss": 2.011,
+      "step": 8856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003733206805039342,
+      "loss": 2.2166,
+      "step": 8857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037321559739185195,
+      "loss": 2.1627,
+      "step": 8858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000373110520264285,
+      "loss": 2.0256,
+      "step": 8859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037300544912619336,
+      "loss": 1.9667,
+      "step": 8860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003729003839825364,
+      "loss": 2.1583,
+      "step": 8861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037279532483827383,
+      "loss": 1.9908,
+      "step": 8862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037269027169836434,
+      "loss": 1.9802,
+      "step": 8863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037258522456776676,
+      "loss": 1.9836,
+      "step": 8864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037248018345143965,
+      "loss": 1.99,
+      "step": 8865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037237514835434117,
+      "loss": 2.0421,
+      "step": 8866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037227011928142904,
+      "loss": 1.8606,
+      "step": 8867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003721650962376612,
+      "loss": 2.0772,
+      "step": 8868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003720600792279947,
+      "loss": 2.0307,
+      "step": 8869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00037195506825738665,
+      "loss": 2.1118,
+      "step": 8870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037185006333079386,
+      "loss": 2.0231,
+      "step": 8871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037174506445317293,
+      "loss": 1.9573,
+      "step": 8872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037164007162947965,
+      "loss": 2.1256,
+      "step": 8873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037153508486467034,
+      "loss": 2.1188,
+      "step": 8874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037143010416370036,
+      "loss": 2.156,
+      "step": 8875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003713251295315253,
+      "loss": 2.0971,
+      "step": 8876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003712201609730998,
+      "loss": 1.8586,
+      "step": 8877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037111519849337905,
+      "loss": 2.1469,
+      "step": 8878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003710102420973172,
+      "loss": 2.1751,
+      "step": 8879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003709052917898687,
+      "loss": 2.0487,
+      "step": 8880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003708003475759871,
+      "loss": 2.1164,
+      "step": 8881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003706954094606264,
+      "loss": 2.0961,
+      "step": 8882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003705904774487396,
+      "loss": 2.0125,
+      "step": 8883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037048555154528,
+      "loss": 2.0157,
+      "step": 8884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037038063175520003,
+      "loss": 2.0439,
+      "step": 8885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003702757180834525,
+      "loss": 2.0805,
+      "step": 8886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00037017081053498924,
+      "loss": 1.9994,
+      "step": 8887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003700659091147624,
+      "loss": 2.0966,
+      "step": 8888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003699610138277234,
+      "loss": 2.0929,
+      "step": 8889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003698561246788237,
+      "loss": 2.1192,
+      "step": 8890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000369751241673014,
+      "loss": 2.1658,
+      "step": 8891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003696463648152454,
+      "loss": 2.0757,
+      "step": 8892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000369541494110468,
+      "loss": 2.0364,
+      "step": 8893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003694366295636323,
+      "loss": 1.9177,
+      "step": 8894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036933177117968774,
+      "loss": 2.0656,
+      "step": 8895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036922691896358416,
+      "loss": 2.1083,
+      "step": 8896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003691220729202707,
+      "loss": 1.9605,
+      "step": 8897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003690172330546965,
+      "loss": 2.0068,
+      "step": 8898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036891239937181,
+      "loss": 2.1708,
+      "step": 8899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003688075718765598,
+      "loss": 2.0318,
+      "step": 8900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036870275057389367,
+      "loss": 1.9843,
+      "step": 8901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036859793546876,
+      "loss": 2.0107,
+      "step": 8902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003684931265661057,
+      "loss": 1.9856,
+      "step": 8903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036838832387087834,
+      "loss": 2.1497,
+      "step": 8904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036828352738802476,
+      "loss": 2.0603,
+      "step": 8905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036817873712249156,
+      "loss": 2.0155,
+      "step": 8906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003680739530792251,
+      "loss": 2.2024,
+      "step": 8907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036796917526317154,
+      "loss": 2.0615,
+      "step": 8908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036786440367927645,
+      "loss": 1.9991,
+      "step": 8909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003677596383324853,
+      "loss": 2.0691,
+      "step": 8910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036765487922774343,
+      "loss": 2.0933,
+      "step": 8911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036755012636999565,
+      "loss": 2.0911,
+      "step": 8912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036744537976418645,
+      "loss": 2.0587,
+      "step": 8913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036734063941526007,
+      "loss": 2.0683,
+      "step": 8914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036723590532816074,
+      "loss": 2.1956,
+      "step": 8915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036713117750783187,
+      "loss": 2.0662,
+      "step": 8916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003670264559592171,
+      "loss": 2.0873,
+      "step": 8917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003669217406872592,
+      "loss": 1.9322,
+      "step": 8918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003668170316969015,
+      "loss": 2.2251,
+      "step": 8919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003667123289930859,
+      "loss": 2.0514,
+      "step": 8920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000366607632580755,
+      "loss": 2.0288,
+      "step": 8921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036650294246485057,
+      "loss": 1.9537,
+      "step": 8922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003663982586503143,
+      "loss": 2.201,
+      "step": 8923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036629358114208737,
+      "loss": 1.9382,
+      "step": 8924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000366188909945111,
+      "loss": 2.0498,
+      "step": 8925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003660842450643257,
+      "loss": 2.1223,
+      "step": 8926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003659795865046721,
+      "loss": 2.1492,
+      "step": 8927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036587493427109015,
+      "loss": 2.1363,
+      "step": 8928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003657702883685198,
+      "loss": 1.957,
+      "step": 8929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036566564880190043,
+      "loss": 2.1681,
+      "step": 8930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003655610155761715,
+      "loss": 2.1943,
+      "step": 8931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003654563886962717,
+      "loss": 2.0053,
+      "step": 8932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003653517681671398,
+      "loss": 1.8438,
+      "step": 8933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000365247153993714,
+      "loss": 1.9605,
+      "step": 8934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036514254618093256,
+      "loss": 2.0533,
+      "step": 8935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000365037944733733,
+      "loss": 1.9755,
+      "step": 8936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036493334965705283,
+      "loss": 2.0488,
+      "step": 8937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036482876095582904,
+      "loss": 1.9143,
+      "step": 8938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003647241786349988,
+      "loss": 1.9843,
+      "step": 8939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003646196026994982,
+      "loss": 2.0874,
+      "step": 8940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036451503315426394,
+      "loss": 2.0178,
+      "step": 8941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036441047000423146,
+      "loss": 1.9578,
+      "step": 8942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003643059132543367,
+      "loss": 2.0498,
+      "step": 8943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036420136290951486,
+      "loss": 1.8516,
+      "step": 8944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036409681897470105,
+      "loss": 2.2352,
+      "step": 8945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036399228145482976,
+      "loss": 2.1025,
+      "step": 8946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003638877503548358,
+      "loss": 1.9546,
+      "step": 8947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003637832256796528,
+      "loss": 1.9511,
+      "step": 8948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003636787074342149,
+      "loss": 2.0785,
+      "step": 8949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003635741956234555,
+      "loss": 1.9711,
+      "step": 8950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003634696902523079,
+      "loss": 2.1155,
+      "step": 8951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003633651913257047,
+      "loss": 2.0756,
+      "step": 8952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036326069884857885,
+      "loss": 1.7961,
+      "step": 8953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036315621282586235,
+      "loss": 1.9748,
+      "step": 8954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036305173326248744,
+      "loss": 2.0368,
+      "step": 8955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036294726016338554,
+      "loss": 2.1345,
+      "step": 8956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003628427935334882,
+      "loss": 2.1521,
+      "step": 8957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036273833337772634,
+      "loss": 2.0914,
+      "step": 8958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003626338797010309,
+      "loss": 2.1235,
+      "step": 8959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036252943250833215,
+      "loss": 1.9702,
+      "step": 8960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003624249918045604,
+      "loss": 1.9748,
+      "step": 8961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036232055759464533,
+      "loss": 2.0329,
+      "step": 8962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036221612988351665,
+      "loss": 2.0096,
+      "step": 8963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036211170867610345,
+      "loss": 2.1043,
+      "step": 8964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003620072939773347,
+      "loss": 1.9649,
+      "step": 8965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036190288579213896,
+      "loss": 1.914,
+      "step": 8966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003617984841254447,
+      "loss": 2.0917,
+      "step": 8967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003616940889821797,
+      "loss": 2.136,
+      "step": 8968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003615897003672719,
+      "loss": 2.0545,
+      "step": 8969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036148531828564837,
+      "loss": 1.9363,
+      "step": 8970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003613809427422365,
+      "loss": 1.9338,
+      "step": 8971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036127657374196274,
+      "loss": 2.0602,
+      "step": 8972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003611722112897539,
+      "loss": 2.0567,
+      "step": 8973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003610678553905358,
+      "loss": 2.0515,
+      "step": 8974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036096350604923453,
+      "loss": 2.0445,
+      "step": 8975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036085916327077553,
+      "loss": 1.9883,
+      "step": 8976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036075482706008396,
+      "loss": 2.0427,
+      "step": 8977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036065049742208465,
+      "loss": 2.0734,
+      "step": 8978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036054617436170254,
+      "loss": 1.9749,
+      "step": 8979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003604418578838615,
+      "loss": 2.0091,
+      "step": 8980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036033754799348584,
+      "loss": 1.9982,
+      "step": 8981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000360233244695499,
+      "loss": 2.2243,
+      "step": 8982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00036012894799482455,
+      "loss": 2.0085,
+      "step": 8983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003600246578963853,
+      "loss": 2.0744,
+      "step": 8984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035992037440510415,
+      "loss": 1.9702,
+      "step": 8985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035981609752590337,
+      "loss": 2.0684,
+      "step": 8986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003597118272637053,
+      "loss": 1.94,
+      "step": 8987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003596075636234315,
+      "loss": 2.0474,
+      "step": 8988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003595033066100036,
+      "loss": 2.155,
+      "step": 8989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003593990562283427,
+      "loss": 1.9791,
+      "step": 8990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035929481248336983,
+      "loss": 2.0907,
+      "step": 8991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035919057538000515,
+      "loss": 2.1046,
+      "step": 8992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003590863449231693,
+      "loss": 2.0578,
+      "step": 8993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035898212111778197,
+      "loss": 2.0599,
+      "step": 8994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003588779039687628,
+      "loss": 2.0232,
+      "step": 8995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035877369348103107,
+      "loss": 2.2309,
+      "step": 8996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035866948965950586,
+      "loss": 2.05,
+      "step": 8997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003585652925091056,
+      "loss": 2.0957,
+      "step": 8998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000358461102034749,
+      "loss": 2.1031,
+      "step": 8999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035835691824135377,
+      "loss": 2.0287,
+      "step": 9000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035825274113383766,
+      "loss": 2.0276,
+      "step": 9001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003581485707171183,
+      "loss": 2.0812,
+      "step": 9002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003580444069961124,
+      "loss": 1.9837,
+      "step": 9003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035794024997573716,
+      "loss": 1.9326,
+      "step": 9004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003578360996609087,
+      "loss": 2.0855,
+      "step": 9005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003577319560565433,
+      "loss": 1.9367,
+      "step": 9006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003576278191675567,
+      "loss": 2.0111,
+      "step": 9007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003575236889988646,
+      "loss": 2.0209,
+      "step": 9008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003574195655553818,
+      "loss": 1.9061,
+      "step": 9009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003573154488420235,
+      "loss": 1.9026,
+      "step": 9010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035721133886370415,
+      "loss": 2.2188,
+      "step": 9011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000357107235625338,
+      "loss": 2.0562,
+      "step": 9012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035700313913183886,
+      "loss": 2.189,
+      "step": 9013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003568990493881205,
+      "loss": 2.0365,
+      "step": 9014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003567949663990959,
+      "loss": 2.0161,
+      "step": 9015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003566908901696785,
+      "loss": 2.0106,
+      "step": 9016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035658682070478044,
+      "loss": 2.005,
+      "step": 9017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003564827580093143,
+      "loss": 2.0741,
+      "step": 9018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000356378702088192,
+      "loss": 1.935,
+      "step": 9019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035627465294632533,
+      "loss": 1.9863,
+      "step": 9020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035617061058862545,
+      "loss": 2.0366,
+      "step": 9021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003560665750200036,
+      "loss": 2.097,
+      "step": 9022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035596254624537026,
+      "loss": 1.9612,
+      "step": 9023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003558585242696362,
+      "loss": 1.8572,
+      "step": 9024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003557545090977111,
+      "loss": 2.0563,
+      "step": 9025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000355650500734505,
+      "loss": 2.1001,
+      "step": 9026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003555464991849271,
+      "loss": 1.9107,
+      "step": 9027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003554425044538868,
+      "loss": 2.114,
+      "step": 9028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003553385165462926,
+      "loss": 1.8998,
+      "step": 9029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003552345354670531,
+      "loss": 2.1482,
+      "step": 9030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035513056122107644,
+      "loss": 2.0546,
+      "step": 9031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003550265938132705,
+      "loss": 1.9513,
+      "step": 9032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003549226332485426,
+      "loss": 2.0509,
+      "step": 9033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035481867953180026,
+      "loss": 2.0306,
+      "step": 9034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035471473266794986,
+      "loss": 2.1053,
+      "step": 9035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003546107926618984,
+      "loss": 2.1824,
+      "step": 9036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003545068595185517,
+      "loss": 1.9881,
+      "step": 9037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035440293324281595,
+      "loss": 2.0728,
+      "step": 9038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003542990138395965,
+      "loss": 2.1478,
+      "step": 9039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035419510131379873,
+      "loss": 2.0273,
+      "step": 9040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003540911956703274,
+      "loss": 2.0976,
+      "step": 9041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003539872969140873,
+      "loss": 2.0824,
+      "step": 9042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003538834050499824,
+      "loss": 2.1536,
+      "step": 9043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003537795200829169,
+      "loss": 2.0873,
+      "step": 9044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003536756420177943,
+      "loss": 1.948,
+      "step": 9045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035357177085951793,
+      "loss": 1.946,
+      "step": 9046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003534679066129906,
+      "loss": 2.0172,
+      "step": 9047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003533640492831152,
+      "loss": 2.0961,
+      "step": 9048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003532601988747938,
+      "loss": 2.1505,
+      "step": 9049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003531563553929285,
+      "loss": 2.0353,
+      "step": 9050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035305251884242087,
+      "loss": 1.9227,
+      "step": 9051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003529486892281724,
+      "loss": 2.2149,
+      "step": 9052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035284486655508376,
+      "loss": 2.1758,
+      "step": 9053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000352741050828056,
+      "loss": 1.9858,
+      "step": 9054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035263724205198923,
+      "loss": 2.0189,
+      "step": 9055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035253344023178357,
+      "loss": 2.0706,
+      "step": 9056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035242964537233855,
+      "loss": 2.0759,
+      "step": 9057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003523258574785537,
+      "loss": 1.9559,
+      "step": 9058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000352222076555328,
+      "loss": 2.1155,
+      "step": 9059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003521183026075601,
+      "loss": 2.0346,
+      "step": 9060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003520145356401484,
+      "loss": 2.0687,
+      "step": 9061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035191077565799095,
+      "loss": 2.127,
+      "step": 9062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003518070226659854,
+      "loss": 1.9921,
+      "step": 9063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035170327666902924,
+      "loss": 2.0582,
+      "step": 9064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035159953767201934,
+      "loss": 2.0165,
+      "step": 9065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035149580567985276,
+      "loss": 2.047,
+      "step": 9066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003513920806974255,
+      "loss": 2.0568,
+      "step": 9067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035128836272963377,
+      "loss": 1.9809,
+      "step": 9068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003511846517813733,
+      "loss": 2.0367,
+      "step": 9069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035108094785753956,
+      "loss": 1.9649,
+      "step": 9070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035097725096302747,
+      "loss": 1.8775,
+      "step": 9071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035087356110273196,
+      "loss": 1.9141,
+      "step": 9072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00035076987828154717,
+      "loss": 1.9828,
+      "step": 9073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003506662025043674,
+      "loss": 2.1264,
+      "step": 9074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003505625337760862,
+      "loss": 2.0586,
+      "step": 9075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003504588721015972,
+      "loss": 2.0918,
+      "step": 9076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003503552174857932,
+      "loss": 2.0063,
+      "step": 9077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003502515699335672,
+      "loss": 2.0293,
+      "step": 9078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003501479294498113,
+      "loss": 2.114,
+      "step": 9079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003500442960394179,
+      "loss": 2.156,
+      "step": 9080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034994066970727843,
+      "loss": 2.1932,
+      "step": 9081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003498370504582846,
+      "loss": 1.9965,
+      "step": 9082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034973343829732716,
+      "loss": 1.9233,
+      "step": 9083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034962983322929715,
+      "loss": 2.1434,
+      "step": 9084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003495262352590847,
+      "loss": 1.9882,
+      "step": 9085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034942264439158003,
+      "loss": 1.9799,
+      "step": 9086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034931906063167276,
+      "loss": 2.1177,
+      "step": 9087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034921548398425247,
+      "loss": 1.9864,
+      "step": 9088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000349111914454208,
+      "loss": 2.0631,
+      "step": 9089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034900835204642824,
+      "loss": 2.0281,
+      "step": 9090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003489047967658013,
+      "loss": 2.1612,
+      "step": 9091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003488012486172155,
+      "loss": 2.1801,
+      "step": 9092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034869770760555865,
+      "loss": 2.1487,
+      "step": 9093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034859417373571774,
+      "loss": 2.0763,
+      "step": 9094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034849064701258014,
+      "loss": 2.0659,
+      "step": 9095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034838712744103236,
+      "loss": 2.0662,
+      "step": 9096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034828361502596084,
+      "loss": 2.1144,
+      "step": 9097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003481801097722516,
+      "loss": 1.9901,
+      "step": 9098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034807661168479034,
+      "loss": 2.2592,
+      "step": 9099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034797312076846235,
+      "loss": 2.095,
+      "step": 9100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034786963702815267,
+      "loss": 2.1392,
+      "step": 9101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034776616046874593,
+      "loss": 2.0287,
+      "step": 9102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003476626910951266,
+      "loss": 2.2118,
+      "step": 9103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003475592289121785,
+      "loss": 1.9604,
+      "step": 9104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034745577392478546,
+      "loss": 1.9362,
+      "step": 9105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034735232613783055,
+      "loss": 2.1446,
+      "step": 9106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000347248885556197,
+      "loss": 1.9912,
+      "step": 9107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003471454521847672,
+      "loss": 2.037,
+      "step": 9108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003470420260284237,
+      "loss": 1.8412,
+      "step": 9109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034693860709204827,
+      "loss": 1.9961,
+      "step": 9110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034683519538052267,
+      "loss": 2.0738,
+      "step": 9111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003467317908987279,
+      "loss": 2.18,
+      "step": 9112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003466283936515452,
+      "loss": 2.0458,
+      "step": 9113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000346525003643855,
+      "loss": 1.9424,
+      "step": 9114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034642162088053754,
+      "loss": 2.1199,
+      "step": 9115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003463182453664727,
+      "loss": 2.0946,
+      "step": 9116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034621487710654025,
+      "loss": 2.1115,
+      "step": 9117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003461115161056191,
+      "loss": 1.9669,
+      "step": 9118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034600816236858843,
+      "loss": 1.9694,
+      "step": 9119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034590481590032654,
+      "loss": 2.0622,
+      "step": 9120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003458014767057117,
+      "loss": 1.9955,
+      "step": 9121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034569814478962183,
+      "loss": 2.0346,
+      "step": 9122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034559482015693435,
+      "loss": 2.1454,
+      "step": 9123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034549150281252633,
+      "loss": 1.9508,
+      "step": 9124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034538819276127484,
+      "loss": 2.0949,
+      "step": 9125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003452848900080561,
+      "loss": 1.999,
+      "step": 9126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003451815945577465,
+      "loss": 2.1198,
+      "step": 9127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034507830641522145,
+      "loss": 2.0638,
+      "step": 9128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003449750255853568,
+      "loss": 2.0305,
+      "step": 9129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003448717520730273,
+      "loss": 1.9948,
+      "step": 9130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034476848588310796,
+      "loss": 1.9956,
+      "step": 9131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034466522702047297,
+      "loss": 1.8821,
+      "step": 9132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034456197548999657,
+      "loss": 1.9949,
+      "step": 9133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034445873129655235,
+      "loss": 2.1097,
+      "step": 9134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034435549444501376,
+      "loss": 1.9275,
+      "step": 9135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003442522649402536,
+      "loss": 2.1412,
+      "step": 9136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003441490427871449,
+      "loss": 1.9846,
+      "step": 9137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034404582799055965,
+      "loss": 2.0567,
+      "step": 9138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003439426205553701,
+      "loss": 2.064,
+      "step": 9139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034383942048644755,
+      "loss": 2.0066,
+      "step": 9140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034373622778866367,
+      "loss": 2.0263,
+      "step": 9141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034363304246688904,
+      "loss": 2.1325,
+      "step": 9142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034352986452599455,
+      "loss": 2.1477,
+      "step": 9143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034342669397085016,
+      "loss": 2.0545,
+      "step": 9144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034332353080632596,
+      "loss": 1.9555,
+      "step": 9145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034322037503729125,
+      "loss": 2.0375,
+      "step": 9146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034311722666861556,
+      "loss": 2.0009,
+      "step": 9147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034301408570516746,
+      "loss": 2.0287,
+      "step": 9148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034291095215181554,
+      "loss": 1.9896,
+      "step": 9149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034280782601342794,
+      "loss": 2.0203,
+      "step": 9150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003427047072948725,
+      "loss": 1.9141,
+      "step": 9151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003426015960010165,
+      "loss": 2.1114,
+      "step": 9152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034249849213672726,
+      "loss": 1.9607,
+      "step": 9153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003423953957068713,
+      "loss": 1.9616,
+      "step": 9154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034229230671631517,
+      "loss": 1.9901,
+      "step": 9155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034218922516992466,
+      "loss": 2.1639,
+      "step": 9156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034208615107256593,
+      "loss": 2.0505,
+      "step": 9157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034198308442910376,
+      "loss": 1.9908,
+      "step": 9158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034188002524440355,
+      "loss": 2.1736,
+      "step": 9159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003417769735233297,
+      "loss": 2.0387,
+      "step": 9160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003416739292707466,
+      "loss": 1.9469,
+      "step": 9161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000341570892491518,
+      "loss": 1.9038,
+      "step": 9162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003414678631905078,
+      "loss": 1.9821,
+      "step": 9163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003413648413725789,
+      "loss": 1.962,
+      "step": 9164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003412618270425944,
+      "loss": 1.7956,
+      "step": 9165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034115882020541654,
+      "loss": 2.2409,
+      "step": 9166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034105582086590785,
+      "loss": 2.0764,
+      "step": 9167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003409528290289298,
+      "loss": 2.1372,
+      "step": 9168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003408498446993441,
+      "loss": 2.077,
+      "step": 9169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003407468678820116,
+      "loss": 2.2654,
+      "step": 9170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034064389858179324,
+      "loss": 2.0707,
+      "step": 9171
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034054093680354925,
+      "loss": 2.0158,
+      "step": 9172
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003404379825521399,
+      "loss": 2.013,
+      "step": 9173
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003403350358324245,
+      "loss": 2.037,
+      "step": 9174
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003402320966492628,
+      "loss": 1.9828,
+      "step": 9175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003401291650075134,
+      "loss": 2.1172,
+      "step": 9176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034002624091203517,
+      "loss": 2.0934,
+      "step": 9177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033992332436768614,
+      "loss": 2.0636,
+      "step": 9178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003398204153793245,
+      "loss": 1.922,
+      "step": 9179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003397175139518074,
+      "loss": 2.2071,
+      "step": 9180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033961462008999234,
+      "loss": 2.0098,
+      "step": 9181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033951173379873604,
+      "loss": 2.1218,
+      "step": 9182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000339408855082895,
+      "loss": 2.2298,
+      "step": 9183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033930598394732523,
+      "loss": 1.9768,
+      "step": 9184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003392031203968826,
+      "loss": 1.9601,
+      "step": 9185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033910026443642263,
+      "loss": 2.1362,
+      "step": 9186
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033899741607079995,
+      "loss": 1.9835,
+      "step": 9187
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033889457530486967,
+      "loss": 1.9351,
+      "step": 9188
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033879174214348573,
+      "loss": 1.9527,
+      "step": 9189
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003386889165915025,
+      "loss": 2.0036,
+      "step": 9190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033858609865377323,
+      "loss": 1.9265,
+      "step": 9191
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033848328833515143,
+      "loss": 1.9962,
+      "step": 9192
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003383804856404896,
+      "loss": 2.0037,
+      "step": 9193
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003382776905746408,
+      "loss": 2.0652,
+      "step": 9194
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033817490314245677,
+      "loss": 2.0398,
+      "step": 9195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033807212334878954,
+      "loss": 2.1318,
+      "step": 9196
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033796935119849035,
+      "loss": 2.0305,
+      "step": 9197
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003378665866964105,
+      "loss": 2.1063,
+      "step": 9198
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003377638298474006,
+      "loss": 2.0089,
+      "step": 9199
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000337661080656311,
+      "loss": 1.9974,
+      "step": 9200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003375583391279917,
+      "loss": 1.9949,
+      "step": 9201
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003374556052672924,
+      "loss": 2.0098,
+      "step": 9202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003373528790790623,
+      "loss": 2.0887,
+      "step": 9203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003372501605681504,
+      "loss": 2.0308,
+      "step": 9204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033714744973940505,
+      "loss": 2.0349,
+      "step": 9205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003370447465976747,
+      "loss": 2.0647,
+      "step": 9206
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033694205114780703,
+      "loss": 1.9904,
+      "step": 9207
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033683936339464957,
+      "loss": 2.0655,
+      "step": 9208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003367366833430493,
+      "loss": 2.0714,
+      "step": 9209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033663401099785306,
+      "loss": 2.1076,
+      "step": 9210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033653134636390717,
+      "loss": 1.9347,
+      "step": 9211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003364286894460577,
+      "loss": 1.9719,
+      "step": 9212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033632604024915017,
+      "loss": 1.9865,
+      "step": 9213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033622339877803004,
+      "loss": 2.0792,
+      "step": 9214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000336120765037542,
+      "loss": 2.0784,
+      "step": 9215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033601813903253087,
+      "loss": 1.9917,
+      "step": 9216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033591552076784047,
+      "loss": 2.0067,
+      "step": 9217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033581291024831504,
+      "loss": 1.9718,
+      "step": 9218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003357103074787977,
+      "loss": 2.1179,
+      "step": 9219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033560771246413177,
+      "loss": 2.0282,
+      "step": 9220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033550512520915977,
+      "loss": 2.2048,
+      "step": 9221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033540254571872417,
+      "loss": 2.0046,
+      "step": 9222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033529997399766687,
+      "loss": 2.0359,
+      "step": 9223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003351974100508297,
+      "loss": 1.9532,
+      "step": 9224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003350948538830536,
+      "loss": 1.8913,
+      "step": 9225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033499230549917977,
+      "loss": 2.0183,
+      "step": 9226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003348897649040484,
+      "loss": 2.0724,
+      "step": 9227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033478723210250005,
+      "loss": 2.1232,
+      "step": 9228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033468470709937414,
+      "loss": 2.0898,
+      "step": 9229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003345821898995103,
+      "loss": 2.0516,
+      "step": 9230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003344796805077474,
+      "loss": 2.0083,
+      "step": 9231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033437717892892437,
+      "loss": 2.0104,
+      "step": 9232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033427468516787916,
+      "loss": 1.9608,
+      "step": 9233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003341721992294502,
+      "loss": 2.0171,
+      "step": 9234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033406972111847455,
+      "loss": 1.9574,
+      "step": 9235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003339672508397898,
+      "loss": 2.0612,
+      "step": 9236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003338647883982325,
+      "loss": 2.0769,
+      "step": 9237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003337623337986394,
+      "loss": 1.9901,
+      "step": 9238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033365988704584627,
+      "loss": 2.0959,
+      "step": 9239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003335574481446891,
+      "loss": 2.1428,
+      "step": 9240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003334550171000031,
+      "loss": 2.1061,
+      "step": 9241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003333525939166234,
+      "loss": 2.1102,
+      "step": 9242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033325017859938436,
+      "loss": 2.0237,
+      "step": 9243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003331477711531205,
+      "loss": 2.181,
+      "step": 9244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003330453715826653,
+      "loss": 1.8562,
+      "step": 9245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003329429798928527,
+      "loss": 2.0301,
+      "step": 9246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033284059608851556,
+      "loss": 2.1962,
+      "step": 9247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003327382201744868,
+      "loss": 2.0417,
+      "step": 9248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033263585215559854,
+      "loss": 2.0488,
+      "step": 9249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003325334920366831,
+      "loss": 2.0349,
+      "step": 9250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003324311398225718,
+      "loss": 2.0844,
+      "step": 9251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033232879551809623,
+      "loss": 2.0099,
+      "step": 9252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003322264591280869,
+      "loss": 1.9174,
+      "step": 9253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033212413065737467,
+      "loss": 1.9912,
+      "step": 9254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003320218101107895,
+      "loss": 2.1476,
+      "step": 9255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003319194974931612,
+      "loss": 1.9805,
+      "step": 9256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003318171928093191,
+      "loss": 2.0907,
+      "step": 9257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003317148960640923,
+      "loss": 2.0966,
+      "step": 9258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033161260726230935,
+      "loss": 2.0002,
+      "step": 9259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033151032640879846,
+      "loss": 1.9524,
+      "step": 9260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033140805350838777,
+      "loss": 1.9835,
+      "step": 9261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003313057885659047,
+      "loss": 2.0431,
+      "step": 9262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003312035315861763,
+      "loss": 2.0686,
+      "step": 9263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033110128257402943,
+      "loss": 1.9998,
+      "step": 9264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003309990415342904,
+      "loss": 2.163,
+      "step": 9265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033089680847178535,
+      "loss": 2.0153,
+      "step": 9266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033079458339133966,
+      "loss": 1.9749,
+      "step": 9267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003306923662977789,
+      "loss": 2.1117,
+      "step": 9268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003305901571959277,
+      "loss": 2.2059,
+      "step": 9269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003304879560906108,
+      "loss": 2.131,
+      "step": 9270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033038576298665204,
+      "loss": 2.0723,
+      "step": 9271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00033028357788887543,
+      "loss": 2.2148,
+      "step": 9272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003301814008021041,
+      "loss": 2.2126,
+      "step": 9273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003300792317311614,
+      "loss": 2.081,
+      "step": 9274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032997707068086954,
+      "loss": 2.0518,
+      "step": 9275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032987491765605105,
+      "loss": 2.0837,
+      "step": 9276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003297727726615276,
+      "loss": 2.0346,
+      "step": 9277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003296706357021208,
+      "loss": 2.0486,
+      "step": 9278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003295685067826516,
+      "loss": 2.0924,
+      "step": 9279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032946638590794096,
+      "loss": 2.0347,
+      "step": 9280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000329364273082809,
+      "loss": 1.9346,
+      "step": 9281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003292621683120758,
+      "loss": 1.9753,
+      "step": 9282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003291600716005608,
+      "loss": 1.9869,
+      "step": 9283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003290579829530834,
+      "loss": 2.0804,
+      "step": 9284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032895590237446223,
+      "loss": 2.0949,
+      "step": 9285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003288538298695159,
+      "loss": 2.0165,
+      "step": 9286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003287517654430623,
+      "loss": 2.0592,
+      "step": 9287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003286497090999193,
+      "loss": 2.076,
+      "step": 9288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032854766084490397,
+      "loss": 2.0485,
+      "step": 9289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003284456206828333,
+      "loss": 2.0792,
+      "step": 9290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003283435886185239,
+      "loss": 2.068,
+      "step": 9291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000328241564656792,
+      "loss": 2.1148,
+      "step": 9292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032813954880245305,
+      "loss": 1.998,
+      "step": 9293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032803754106032273,
+      "loss": 1.9385,
+      "step": 9294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032793554143521585,
+      "loss": 1.9776,
+      "step": 9295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032783354993194723,
+      "loss": 2.0552,
+      "step": 9296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003277315665553308,
+      "loss": 2.0226,
+      "step": 9297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003276295913101808,
+      "loss": 1.9429,
+      "step": 9298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003275276242013103,
+      "loss": 2.0828,
+      "step": 9299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032742566523353265,
+      "loss": 2.0321,
+      "step": 9300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032732371441166044,
+      "loss": 2.0267,
+      "step": 9301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003272217717405061,
+      "loss": 1.9941,
+      "step": 9302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003271198372248814,
+      "loss": 1.9608,
+      "step": 9303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032701791086959793,
+      "loss": 2.0316,
+      "step": 9304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032691599267946684,
+      "loss": 1.9318,
+      "step": 9305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032681408265929903,
+      "loss": 1.9989,
+      "step": 9306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003267121808139047,
+      "loss": 2.089,
+      "step": 9307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032661028714809406,
+      "loss": 2.2306,
+      "step": 9308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003265084016666764,
+      "loss": 1.9679,
+      "step": 9309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003264065243744614,
+      "loss": 2.1569,
+      "step": 9310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032630465527625743,
+      "loss": 2.066,
+      "step": 9311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032620279437687335,
+      "loss": 2.0529,
+      "step": 9312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00032610094168111694,
+      "loss": 1.992,
+      "step": 9313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003259990971937961,
+      "loss": 2.0645,
+      "step": 9314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003258972609197178,
+      "loss": 2.0709,
+      "step": 9315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003257954328636894,
+      "loss": 2.0018,
+      "step": 9316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003256936130305171,
+      "loss": 2.1979,
+      "step": 9317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032559180142500716,
+      "loss": 2.0333,
+      "step": 9318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032548999805196517,
+      "loss": 2.1038,
+      "step": 9319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032538820291619667,
+      "loss": 2.0232,
+      "step": 9320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003252864160225065,
+      "loss": 1.9506,
+      "step": 9321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003251846373756994,
+      "loss": 2.0678,
+      "step": 9322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003250828669805792,
+      "loss": 2.1563,
+      "step": 9323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003249811048419501,
+      "loss": 2.1552,
+      "step": 9324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003248793509646153,
+      "loss": 2.0575,
+      "step": 9325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003247776053533779,
+      "loss": 2.0358,
+      "step": 9326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032467586801304037,
+      "loss": 2.0941,
+      "step": 9327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032457413894840514,
+      "loss": 1.9323,
+      "step": 9328
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003244724181642739,
+      "loss": 1.9961,
+      "step": 9329
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003243707056654483,
+      "loss": 1.9971,
+      "step": 9330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003242690014567292,
+      "loss": 1.8669,
+      "step": 9331
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003241673055429174,
+      "loss": 2.0414,
+      "step": 9332
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032406561792881305,
+      "loss": 2.0693,
+      "step": 9333
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032396393861921623,
+      "loss": 2.0933,
+      "step": 9334
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003238622676189262,
+      "loss": 1.9485,
+      "step": 9335
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003237606049327424,
+      "loss": 2.0351,
+      "step": 9336
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032365895056546315,
+      "loss": 2.0985,
+      "step": 9337
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032355730452188707,
+      "loss": 2.0262,
+      "step": 9338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003234556668068119,
+      "loss": 2.1116,
+      "step": 9339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032335403742503533,
+      "loss": 2.1512,
+      "step": 9340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003232524163813543,
+      "loss": 2.0228,
+      "step": 9341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003231508036805658,
+      "loss": 1.9212,
+      "step": 9342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000323049199327466,
+      "loss": 2.1103,
+      "step": 9343
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000322947603326851,
+      "loss": 2.0443,
+      "step": 9344
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032284601568351623,
+      "loss": 1.9491,
+      "step": 9345
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003227444364022569,
+      "loss": 1.9208,
+      "step": 9346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032264286548786783,
+      "loss": 1.913,
+      "step": 9347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003225413029451434,
+      "loss": 2.015,
+      "step": 9348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032243974877887744,
+      "loss": 1.8868,
+      "step": 9349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003223382029938637,
+      "loss": 2.0516,
+      "step": 9350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003222366655948955,
+      "loss": 2.0826,
+      "step": 9351
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003221351365867653,
+      "loss": 2.0016,
+      "step": 9352
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032203361597426584,
+      "loss": 2.1209,
+      "step": 9353
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003219321037621888,
+      "loss": 1.8897,
+      "step": 9354
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032183059995532614,
+      "loss": 1.9475,
+      "step": 9355
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003217291045584687,
+      "loss": 1.9983,
+      "step": 9356
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032162761757640766,
+      "loss": 2.097,
+      "step": 9357
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032152613901393316,
+      "loss": 2.0992,
+      "step": 9358
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003214246688758354,
+      "loss": 2.2055,
+      "step": 9359
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003213232071669039,
+      "loss": 1.9241,
+      "step": 9360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000321221753891928,
+      "loss": 2.2383,
+      "step": 9361
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032112030905569633,
+      "loss": 2.0425,
+      "step": 9362
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003210188726629975,
+      "loss": 2.0082,
+      "step": 9363
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003209174447186194,
+      "loss": 1.9733,
+      "step": 9364
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032081602522734986,
+      "loss": 2.1721,
+      "step": 9365
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003207146141939759,
+      "loss": 2.0642,
+      "step": 9366
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003206132116232845,
+      "loss": 2.0363,
+      "step": 9367
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000320511817520062,
+      "loss": 1.9023,
+      "step": 9368
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032041043188909456,
+      "loss": 2.0426,
+      "step": 9369
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032030905473516764,
+      "loss": 2.1521,
+      "step": 9370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003202076860630666,
+      "loss": 2.1201,
+      "step": 9371
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032010632587757624,
+      "loss": 1.9411,
+      "step": 9372
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00032000497418348104,
+      "loss": 1.9327,
+      "step": 9373
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003199036309855649,
+      "loss": 1.9153,
+      "step": 9374
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031980229628861165,
+      "loss": 2.011,
+      "step": 9375
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003197009700974043,
+      "loss": 1.9829,
+      "step": 9376
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031959965241672584,
+      "loss": 2.1569,
+      "step": 9377
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031949834325135863,
+      "loss": 2.1154,
+      "step": 9378
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003193970426060848,
+      "loss": 1.995,
+      "step": 9379
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031929575048568573,
+      "loss": 1.8933,
+      "step": 9380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003191944668949429,
+      "loss": 2.0773,
+      "step": 9381
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031909319183863705,
+      "loss": 1.9248,
+      "step": 9382
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003189919253215485,
+      "loss": 1.9846,
+      "step": 9383
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003188906673484574,
+      "loss": 1.9729,
+      "step": 9384
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003187894179241433,
+      "loss": 2.0248,
+      "step": 9385
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031868817705338527,
+      "loss": 1.9765,
+      "step": 9386
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031858694474096237,
+      "loss": 2.1037,
+      "step": 9387
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003184857209916528,
+      "loss": 1.9694,
+      "step": 9388
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003183845058102347,
+      "loss": 2.0666,
+      "step": 9389
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031828329920148546,
+      "loss": 1.9501,
+      "step": 9390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031818210117018255,
+      "loss": 2.0096,
+      "step": 9391
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003180809117211024,
+      "loss": 2.1272,
+      "step": 9392
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031797973085902177,
+      "loss": 2.0007,
+      "step": 9393
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003178785585887163,
+      "loss": 1.8594,
+      "step": 9394
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003177773949149617,
+      "loss": 2.2245,
+      "step": 9395
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031767623984253307,
+      "loss": 1.8772,
+      "step": 9396
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003175750933762053,
+      "loss": 2.1019,
+      "step": 9397
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003174739555207526,
+      "loss": 2.0722,
+      "step": 9398
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000317372826280949,
+      "loss": 1.9756,
+      "step": 9399
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003172717056615678,
+      "loss": 2.0385,
+      "step": 9400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031717059366738253,
+      "loss": 2.0129,
+      "step": 9401
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003170694903031656,
+      "loss": 2.1676,
+      "step": 9402
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003169683955736894,
+      "loss": 2.099,
+      "step": 9403
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003168673094837258,
+      "loss": 1.9794,
+      "step": 9404
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031676623203804645,
+      "loss": 1.9684,
+      "step": 9405
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003166651632414222,
+      "loss": 2.1037,
+      "step": 9406
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031656410309862396,
+      "loss": 2.095,
+      "step": 9407
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003164630516144218,
+      "loss": 1.9083,
+      "step": 9408
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003163620087935858,
+      "loss": 2.0492,
+      "step": 9409
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031626097464088515,
+      "loss": 2.0108,
+      "step": 9410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031615994916108913,
+      "loss": 1.9953,
+      "step": 9411
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003160589323589662,
+      "loss": 2.0287,
+      "step": 9412
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031595792423928483,
+      "loss": 1.9717,
+      "step": 9413
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003158569248068125,
+      "loss": 2.0681,
+      "step": 9414
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031575593406631694,
+      "loss": 1.8436,
+      "step": 9415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031565495202256485,
+      "loss": 2.0499,
+      "step": 9416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003155539786803231,
+      "loss": 2.0897,
+      "step": 9417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031545301404435757,
+      "loss": 2.3571,
+      "step": 9418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031535205811943426,
+      "loss": 1.9721,
+      "step": 9419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031525111091031837,
+      "loss": 2.0575,
+      "step": 9420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031515017242177504,
+      "loss": 2.154,
+      "step": 9421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003150492426585685,
+      "loss": 2.0527,
+      "step": 9422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031494832162546314,
+      "loss": 1.9874,
+      "step": 9423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031484740932722255,
+      "loss": 2.0223,
+      "step": 9424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031474650576861,
+      "loss": 1.9243,
+      "step": 9425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003146456109543884,
+      "loss": 1.9205,
+      "step": 9426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031454472488932027,
+      "loss": 2.0628,
+      "step": 9427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031444384757816756,
+      "loss": 2.0288,
+      "step": 9428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031434297902569196,
+      "loss": 2.0498,
+      "step": 9429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003142421192366547,
+      "loss": 1.973,
+      "step": 9430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003141412682158167,
+      "loss": 2.1998,
+      "step": 9431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003140404259679381,
+      "loss": 2.0039,
+      "step": 9432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031393959249777914,
+      "loss": 2.0203,
+      "step": 9433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031383876781009923,
+      "loss": 1.9916,
+      "step": 9434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003137379519096577,
+      "loss": 2.0853,
+      "step": 9435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000313637144801213,
+      "loss": 2.0181,
+      "step": 9436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031353634648952376,
+      "loss": 1.9594,
+      "step": 9437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003134355569793477,
+      "loss": 2.1137,
+      "step": 9438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003133347762754425,
+      "loss": 1.9873,
+      "step": 9439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000313234004382565,
+      "loss": 2.0973,
+      "step": 9440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003131332413054721,
+      "loss": 2.0749,
+      "step": 9441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031303248704891984,
+      "loss": 1.8903,
+      "step": 9442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031293174161766427,
+      "loss": 1.9956,
+      "step": 9443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003128310050164607,
+      "loss": 1.9421,
+      "step": 9444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000312730277250064,
+      "loss": 2.0383,
+      "step": 9445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031262955832322894,
+      "loss": 1.9861,
+      "step": 9446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031252884824070957,
+      "loss": 1.9384,
+      "step": 9447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031242814700725973,
+      "loss": 1.899,
+      "step": 9448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003123274546276327,
+      "loss": 2.0292,
+      "step": 9449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031222677110658145,
+      "loss": 2.0466,
+      "step": 9450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031212609644885826,
+      "loss": 2.0913,
+      "step": 9451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003120254306592155,
+      "loss": 2.0009,
+      "step": 9452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031192477374240466,
+      "loss": 1.9285,
+      "step": 9453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003118241257031771,
+      "loss": 2.027,
+      "step": 9454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031172348654628343,
+      "loss": 1.9914,
+      "step": 9455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031162285627647427,
+      "loss": 2.0546,
+      "step": 9456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031152223489849936,
+      "loss": 2.1044,
+      "step": 9457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031142162241710845,
+      "loss": 1.9399,
+      "step": 9458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031132101883705066,
+      "loss": 2.0574,
+      "step": 9459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00031122042416307475,
+      "loss": 1.984,
+      "step": 9460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003111198383999288,
+      "loss": 1.9812,
+      "step": 9461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000311019261552361,
+      "loss": 1.9518,
+      "step": 9462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003109186936251185,
+      "loss": 2.1643,
+      "step": 9463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003108181346229486,
+      "loss": 2.0701,
+      "step": 9464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003107175845505977,
+      "loss": 2.0426,
+      "step": 9465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003106170434128122,
+      "loss": 1.9793,
+      "step": 9466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003105165112143377,
+      "loss": 2.2769,
+      "step": 9467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00031041598795991963,
+      "loss": 2.0138,
+      "step": 9468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003103154736543029,
+      "loss": 2.1048,
+      "step": 9469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00031021496830223213,
+      "loss": 2.1898,
+      "step": 9470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003101144719084512,
+      "loss": 2.0026,
+      "step": 9471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003100139844777039,
+      "loss": 2.0428,
+      "step": 9472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030991350601473336,
+      "loss": 2.1133,
+      "step": 9473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003098130365242826,
+      "loss": 2.0728,
+      "step": 9474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003097125760110938,
+      "loss": 1.7718,
+      "step": 9475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030961212447990904,
+      "loss": 1.9911,
+      "step": 9476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003095116819354698,
+      "loss": 2.1195,
+      "step": 9477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030941124838251735,
+      "loss": 1.9783,
+      "step": 9478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003093108238257921,
+      "loss": 2.0939,
+      "step": 9479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003092104082700346,
+      "loss": 2.0318,
+      "step": 9480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003091100017199845,
+      "loss": 2.062,
+      "step": 9481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003090096041803814,
+      "loss": 2.0747,
+      "step": 9482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003089092156559641,
+      "loss": 1.9851,
+      "step": 9483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030880883615147135,
+      "loss": 2.2097,
+      "step": 9484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003087084656716412,
+      "loss": 2.0014,
+      "step": 9485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003086081042212114,
+      "loss": 1.9274,
+      "step": 9486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003085077518049191,
+      "loss": 2.0189,
+      "step": 9487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003084074084275014,
+      "loss": 2.1625,
+      "step": 9488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003083070740936946,
+      "loss": 1.9988,
+      "step": 9489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030820674880823476,
+      "loss": 1.9069,
+      "step": 9490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030810643257585734,
+      "loss": 1.9939,
+      "step": 9491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003080061254012977,
+      "loss": 1.958,
+      "step": 9492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003079058272892904,
+      "loss": 2.1532,
+      "step": 9493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003078055382445698,
+      "loss": 1.9825,
+      "step": 9494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003077052582718698,
+      "loss": 2.1033,
+      "step": 9495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030760498737592387,
+      "loss": 2.1181,
+      "step": 9496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003075047255614649,
+      "loss": 2.1385,
+      "step": 9497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003074044728332256,
+      "loss": 1.8582,
+      "step": 9498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030730422919593805,
+      "loss": 2.0224,
+      "step": 9499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030720399465433414,
+      "loss": 1.8545,
+      "step": 9500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030710376921314487,
+      "loss": 2.1148,
+      "step": 9501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030700355287710144,
+      "loss": 2.0382,
+      "step": 9502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000306903345650934,
+      "loss": 2.051,
+      "step": 9503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003068031475393729,
+      "loss": 1.985,
+      "step": 9504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030670295854714737,
+      "loss": 2.0852,
+      "step": 9505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030660277867898685,
+      "loss": 2.1892,
+      "step": 9506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030650260793961976,
+      "loss": 2.1238,
+      "step": 9507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003064024463337747,
+      "loss": 2.0942,
+      "step": 9508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030630229386617927,
+      "loss": 2.1451,
+      "step": 9509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003062021505415611,
+      "loss": 1.9025,
+      "step": 9510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030610201636464704,
+      "loss": 1.8199,
+      "step": 9511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030600189134016376,
+      "loss": 2.0949,
+      "step": 9512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003059017754728373,
+      "loss": 2.0296,
+      "step": 9513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003058016687673935,
+      "loss": 1.9966,
+      "step": 9514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030570157122855737,
+      "loss": 2.0937,
+      "step": 9515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030560148286105395,
+      "loss": 2.0065,
+      "step": 9516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003055014036696075,
+      "loss": 2.0127,
+      "step": 9517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003054013336589423,
+      "loss": 1.9232,
+      "step": 9518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003053012728337815,
+      "loss": 2.0017,
+      "step": 9519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003052012211988485,
+      "loss": 1.9131,
+      "step": 9520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003051011787588657,
+      "loss": 2.1032,
+      "step": 9521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030500114551855554,
+      "loss": 2.0839,
+      "step": 9522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003049011214826397,
+      "loss": 2.1269,
+      "step": 9523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030480110665583965,
+      "loss": 2.0378,
+      "step": 9524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003047011010428762,
+      "loss": 1.9764,
+      "step": 9525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030460110464846995,
+      "loss": 2.0393,
+      "step": 9526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003045011174773409,
+      "loss": 2.1214,
+      "step": 9527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030440113953420867,
+      "loss": 2.0845,
+      "step": 9528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030430117082379243,
+      "loss": 2.0518,
+      "step": 9529
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003042012113508111,
+      "loss": 2.1465,
+      "step": 9530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003041012611199827,
+      "loss": 1.9038,
+      "step": 9531
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003040013201360253,
+      "loss": 2.1484,
+      "step": 9532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030390138840365644,
+      "loss": 1.9181,
+      "step": 9533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003038014659275929,
+      "loss": 2.1363,
+      "step": 9534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030370155271255153,
+      "loss": 2.0264,
+      "step": 9535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003036016487632481,
+      "loss": 2.1291,
+      "step": 9536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003035017540843986,
+      "loss": 2.0989,
+      "step": 9537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003034018686807182,
+      "loss": 2.1003,
+      "step": 9538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003033019925569217,
+      "loss": 2.0386,
+      "step": 9539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003032021257177234,
+      "loss": 1.9347,
+      "step": 9540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030310226816783746,
+      "loss": 2.151,
+      "step": 9541
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003030024199119772,
+      "loss": 2.0942,
+      "step": 9542
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003029025809548558,
+      "loss": 2.0193,
+      "step": 9543
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003028027513011857,
+      "loss": 2.0273,
+      "step": 9544
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003027029309556794,
+      "loss": 2.0714,
+      "step": 9545
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030260311992304836,
+      "loss": 1.9731,
+      "step": 9546
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030250331820800414,
+      "loss": 2.0717,
+      "step": 9547
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003024035258152573,
+      "loss": 2.0543,
+      "step": 9548
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003023037427495186,
+      "loss": 2.0712,
+      "step": 9549
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030220396901549775,
+      "loss": 2.0904,
+      "step": 9550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003021042046179046,
+      "loss": 1.9536,
+      "step": 9551
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003020044495614479,
+      "loss": 2.0009,
+      "step": 9552
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003019047038508367,
+      "loss": 2.0354,
+      "step": 9553
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003018049674907788,
+      "loss": 1.9176,
+      "step": 9554
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003017052404859824,
+      "loss": 1.9488,
+      "step": 9555
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030160552284115463,
+      "loss": 2.0405,
+      "step": 9556
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003015058145610024,
+      "loss": 1.9802,
+      "step": 9557
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003014061156502321,
+      "loss": 1.9716,
+      "step": 9558
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030130642611355004,
+      "loss": 1.9637,
+      "step": 9559
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030120674595566143,
+      "loss": 2.0718,
+      "step": 9560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003011070751812717,
+      "loss": 2.096,
+      "step": 9561
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003010074137950853,
+      "loss": 2.1164,
+      "step": 9562
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003009077618018067,
+      "loss": 2.0888,
+      "step": 9563
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030080811920613944,
+      "loss": 2.1186,
+      "step": 9564
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030070848601278715,
+      "loss": 2.1423,
+      "step": 9565
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030060886222645257,
+      "loss": 1.8865,
+      "step": 9566
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030050924785183825,
+      "loss": 2.0371,
+      "step": 9567
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030040964289364616,
+      "loss": 2.1781,
+      "step": 9568
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030031004735657794,
+      "loss": 2.1524,
+      "step": 9569
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003002104612453347,
+      "loss": 1.9484,
+      "step": 9570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003001108845646172,
+      "loss": 2.1078,
+      "step": 9571
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00030001131731912555,
+      "loss": 2.084,
+      "step": 9572
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002999117595135597,
+      "loss": 2.0095,
+      "step": 9573
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002998122111526188,
+      "loss": 2.0064,
+      "step": 9574
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029971267224100206,
+      "loss": 1.9576,
+      "step": 9575
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029961314278340766,
+      "loss": 1.9665,
+      "step": 9576
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002995136227845339,
+      "loss": 1.9673,
+      "step": 9577
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002994141122490781,
+      "loss": 2.12,
+      "step": 9578
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002993146111817375,
+      "loss": 2.1056,
+      "step": 9579
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002992151195872087,
+      "loss": 2.0221,
+      "step": 9580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002991156374701881,
+      "loss": 2.0581,
+      "step": 9581
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002990161648353713,
+      "loss": 2.0529,
+      "step": 9582
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002989167016874538,
+      "loss": 2.0627,
+      "step": 9583
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002988172480311303,
+      "loss": 2.012,
+      "step": 9584
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029871780387109544,
+      "loss": 1.9044,
+      "step": 9585
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000298618369212043,
+      "loss": 1.9607,
+      "step": 9586
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002985189440586668,
+      "loss": 1.9806,
+      "step": 9587
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029841952841565966,
+      "loss": 1.8376,
+      "step": 9588
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029832012228771444,
+      "loss": 2.0803,
+      "step": 9589
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002982207256795232,
+      "loss": 2.0925,
+      "step": 9590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002981213385957777,
+      "loss": 2.0155,
+      "step": 9591
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002980219610411693,
+      "loss": 1.9567,
+      "step": 9592
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029792259302038887,
+      "loss": 1.9812,
+      "step": 9593
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029782323453812667,
+      "loss": 2.0802,
+      "step": 9594
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002977238855990728,
+      "loss": 2.1197,
+      "step": 9595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002976245462079167,
+      "loss": 2.0543,
+      "step": 9596
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029752521636934744,
+      "loss": 2.0757,
+      "step": 9597
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029742589608805346,
+      "loss": 2.0569,
+      "step": 9598
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029732658536872327,
+      "loss": 2.0436,
+      "step": 9599
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002972272842160441,
+      "loss": 2.1507,
+      "step": 9600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002971279926347036,
+      "loss": 2.0437,
+      "step": 9601
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029702871062938825,
+      "loss": 2.1011,
+      "step": 9602
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029692943820478466,
+      "loss": 1.9445,
+      "step": 9603
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029683017536557844,
+      "loss": 2.0085,
+      "step": 9604
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002967309221164553,
+      "loss": 2.1524,
+      "step": 9605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002966316784621,
+      "loss": 2.0688,
+      "step": 9606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002965324444071972,
+      "loss": 2.0247,
+      "step": 9607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002964332199564309,
+      "loss": 2.0159,
+      "step": 9608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029633400511448483,
+      "loss": 2.04,
+      "step": 9609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000296234799886042,
+      "loss": 2.0525,
+      "step": 9610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002961356042757852,
+      "loss": 2.1998,
+      "step": 9611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029603641828839666,
+      "loss": 2.1556,
+      "step": 9612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002959372419285584,
+      "loss": 1.9453,
+      "step": 9613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00029583807520095146,
+      "loss": 2.0164,
+      "step": 9614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000295738918110257,
+      "loss": 1.9255,
+      "step": 9615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029563977066115516,
+      "loss": 2.0132,
+      "step": 9616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029554063285832624,
+      "loss": 2.0446,
+      "step": 9617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002954415047064496,
+      "loss": 2.002,
+      "step": 9618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029534238621020437,
+      "loss": 1.9158,
+      "step": 9619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002952432773742691,
+      "loss": 2.1477,
+      "step": 9620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029514417820332206,
+      "loss": 2.1932,
+      "step": 9621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002950450887020409,
+      "loss": 1.9002,
+      "step": 9622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029494600887510283,
+      "loss": 1.8944,
+      "step": 9623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029484693872718483,
+      "loss": 2.0331,
+      "step": 9624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002947478782629629,
+      "loss": 2.0052,
+      "step": 9625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029464882748711334,
+      "loss": 1.8515,
+      "step": 9626
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029454978640431126,
+      "loss": 2.0277,
+      "step": 9627
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002944507550192318,
+      "loss": 1.9081,
+      "step": 9628
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002943517333365493,
+      "loss": 1.8307,
+      "step": 9629
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000294252721360938,
+      "loss": 2.0717,
+      "step": 9630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029415371909707135,
+      "loss": 1.8611,
+      "step": 9631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002940547265496226,
+      "loss": 1.9396,
+      "step": 9632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002939557437232642,
+      "loss": 2.136,
+      "step": 9633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002938567706226687,
+      "loss": 2.002,
+      "step": 9634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029375780725250764,
+      "loss": 2.0643,
+      "step": 9635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029365885361745236,
+      "loss": 2.2234,
+      "step": 9636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029355990972217366,
+      "loss": 1.9467,
+      "step": 9637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002934609755713421,
+      "loss": 1.8392,
+      "step": 9638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002933620511696273,
+      "loss": 2.0785,
+      "step": 9639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000293263136521699,
+      "loss": 1.9478,
+      "step": 9640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029316423163222604,
+      "loss": 2.0967,
+      "step": 9641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002930653365058771,
+      "loss": 2.0685,
+      "step": 9642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029296645114731994,
+      "loss": 1.9914,
+      "step": 9643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002928675755612226,
+      "loss": 2.0534,
+      "step": 9644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029276870975225187,
+      "loss": 2.015,
+      "step": 9645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029266985372507475,
+      "loss": 2.0903,
+      "step": 9646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002925710074843572,
+      "loss": 2.1222,
+      "step": 9647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029247217103476524,
+      "loss": 2.0844,
+      "step": 9648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029237334438096397,
+      "loss": 2.0442,
+      "step": 9649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002922745275276184,
+      "loss": 2.1327,
+      "step": 9650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029217572047939266,
+      "loss": 2.079,
+      "step": 9651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029207692324095103,
+      "loss": 1.9812,
+      "step": 9652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029197813581695667,
+      "loss": 2.0041,
+      "step": 9653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002918793582120727,
+      "loss": 2.0739,
+      "step": 9654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029178059043096167,
+      "loss": 2.1773,
+      "step": 9655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002916818324782856,
+      "loss": 2.2105,
+      "step": 9656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029158308435870615,
+      "loss": 1.9611,
+      "step": 9657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002914843460768842,
+      "loss": 2.2269,
+      "step": 9658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029138561763748084,
+      "loss": 2.1111,
+      "step": 9659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029128689904515606,
+      "loss": 2.1277,
+      "step": 9660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029118819030456963,
+      "loss": 2.1539,
+      "step": 9661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029108949142038053,
+      "loss": 2.1696,
+      "step": 9662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002909908023972482,
+      "loss": 1.8064,
+      "step": 9663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029089212323983056,
+      "loss": 1.9145,
+      "step": 9664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029079345395278554,
+      "loss": 2.1565,
+      "step": 9665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002906947945407705,
+      "loss": 2.1617,
+      "step": 9666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029059614500844265,
+      "loss": 2.103,
+      "step": 9667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029049750536045836,
+      "loss": 2.1299,
+      "step": 9668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002903988756014735,
+      "loss": 1.9195,
+      "step": 9669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002903002557361436,
+      "loss": 2.1107,
+      "step": 9670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002902016457691241,
+      "loss": 2.0938,
+      "step": 9671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029010304570506934,
+      "loss": 1.9494,
+      "step": 9672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029000445554863355,
+      "loss": 1.9662,
+      "step": 9673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028990587530447015,
+      "loss": 2.0938,
+      "step": 9674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002898073049772328,
+      "loss": 1.9651,
+      "step": 9675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028970874457157406,
+      "loss": 2.0233,
+      "step": 9676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028961019409214604,
+      "loss": 2.0,
+      "step": 9677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028951165354360064,
+      "loss": 1.9806,
+      "step": 9678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028941312293058935,
+      "loss": 1.9217,
+      "step": 9679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028931460225776296,
+      "loss": 2.0447,
+      "step": 9680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002892160915297719,
+      "loss": 1.8866,
+      "step": 9681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002891175907512658,
+      "loss": 2.0285,
+      "step": 9682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002890190999268946,
+      "loss": 1.8765,
+      "step": 9683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028892061906130706,
+      "loss": 2.0352,
+      "step": 9684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028882214815915177,
+      "loss": 2.104,
+      "step": 9685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002887236872250765,
+      "loss": 2.109,
+      "step": 9686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002886252362637292,
+      "loss": 2.1224,
+      "step": 9687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028852679527975687,
+      "loss": 2.0788,
+      "step": 9688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002884283642778062,
+      "loss": 2.1027,
+      "step": 9689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002883299432625229,
+      "loss": 2.0969,
+      "step": 9690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002882315322385534,
+      "loss": 1.8667,
+      "step": 9691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002881331312105426,
+      "loss": 1.9407,
+      "step": 9692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028803474018313516,
+      "loss": 1.9449,
+      "step": 9693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002879363591609752,
+      "loss": 2.1786,
+      "step": 9694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028783798814870705,
+      "loss": 2.0148,
+      "step": 9695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002877396271509737,
+      "loss": 2.1296,
+      "step": 9696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002876412761724181,
+      "loss": 1.9492,
+      "step": 9697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028754293521768244,
+      "loss": 2.1206,
+      "step": 9698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000287444604291409,
+      "loss": 1.9789,
+      "step": 9699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000287346283398239,
+      "loss": 2.2141,
+      "step": 9700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028724797254281356,
+      "loss": 2.0371,
+      "step": 9701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002871496717297728,
+      "loss": 2.0006,
+      "step": 9702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028705138096375726,
+      "loss": 2.1312,
+      "step": 9703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002869531002494062,
+      "loss": 2.0552,
+      "step": 9704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002868548295913588,
+      "loss": 2.0177,
+      "step": 9705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002867565689942534,
+      "loss": 2.1194,
+      "step": 9706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002866583184627285,
+      "loss": 2.127,
+      "step": 9707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002865600780014216,
+      "loss": 2.0355,
+      "step": 9708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028646184761496986,
+      "loss": 2.0946,
+      "step": 9709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002863636273080098,
+      "loss": 2.0829,
+      "step": 9710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000286265417085178,
+      "loss": 1.9231,
+      "step": 9711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002861672169511098,
+      "loss": 2.0743,
+      "step": 9712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000286069026910441,
+      "loss": 1.9823,
+      "step": 9713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028597084696780587,
+      "loss": 2.1064,
+      "step": 9714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002858726771278391,
+      "loss": 1.9873,
+      "step": 9715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002857745173951745,
+      "loss": 1.9523,
+      "step": 9716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002856763677744452,
+      "loss": 2.0088,
+      "step": 9717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028557822827028406,
+      "loss": 2.0558,
+      "step": 9718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028548009888732385,
+      "loss": 1.8787,
+      "step": 9719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002853819796301963,
+      "loss": 1.9198,
+      "step": 9720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028528387050353286,
+      "loss": 1.9315,
+      "step": 9721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028518577151196437,
+      "loss": 1.9625,
+      "step": 9722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002850876826601216,
+      "loss": 1.9457,
+      "step": 9723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028498960395263446,
+      "loss": 2.0615,
+      "step": 9724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028489153539413246,
+      "loss": 1.9472,
+      "step": 9725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028479347698924447,
+      "loss": 1.9475,
+      "step": 9726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002846954287425995,
+      "loss": 2.2086,
+      "step": 9727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002845973906588253,
+      "loss": 2.0668,
+      "step": 9728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002844993627425497,
+      "loss": 2.1331,
+      "step": 9729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028440134499839963,
+      "loss": 2.0246,
+      "step": 9730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000284303337431002,
+      "loss": 2.0261,
+      "step": 9731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002842053400449829,
+      "loss": 2.0863,
+      "step": 9732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000284107352844968,
+      "loss": 2.2742,
+      "step": 9733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002840093758355823,
+      "loss": 1.9821,
+      "step": 9734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028391140902145107,
+      "loss": 1.9038,
+      "step": 9735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028381345240719813,
+      "loss": 2.0227,
+      "step": 9736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002837155059974475,
+      "loss": 2.0428,
+      "step": 9737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028361756979682214,
+      "loss": 2.0203,
+      "step": 9738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002835196438099453,
+      "loss": 2.0184,
+      "step": 9739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028342172804143905,
+      "loss": 1.9564,
+      "step": 9740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028332382249592536,
+      "loss": 2.1766,
+      "step": 9741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002832259271780253,
+      "loss": 2.1248,
+      "step": 9742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002831280420923602,
+      "loss": 1.9389,
+      "step": 9743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028303016724355026,
+      "loss": 2.0727,
+      "step": 9744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002829323026362154,
+      "loss": 1.9943,
+      "step": 9745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028283444827497474,
+      "loss": 2.0162,
+      "step": 9746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002827366041644478,
+      "loss": 2.0067,
+      "step": 9747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002826387703092528,
+      "loss": 2.0563,
+      "step": 9748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028254094671400766,
+      "loss": 1.954,
+      "step": 9749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002824431333833298,
+      "loss": 2.0501,
+      "step": 9750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028234533032183643,
+      "loss": 2.13,
+      "step": 9751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002822475375341441,
+      "loss": 2.0586,
+      "step": 9752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028214975502486873,
+      "loss": 1.8343,
+      "step": 9753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002820519827986257,
+      "loss": 2.2052,
+      "step": 9754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002819542208600305,
+      "loss": 2.0954,
+      "step": 9755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028185646921369744,
+      "loss": 2.1625,
+      "step": 9756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028175872786424084,
+      "loss": 1.8777,
+      "step": 9757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002816609968162739,
+      "loss": 1.9884,
+      "step": 9758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002815632760744101,
+      "loss": 2.0163,
+      "step": 9759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028146556564326216,
+      "loss": 1.9454,
+      "step": 9760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028136786552744196,
+      "loss": 2.142,
+      "step": 9761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002812701757315612,
+      "loss": 1.9262,
+      "step": 9762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002811724962602313,
+      "loss": 2.0751,
+      "step": 9763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00028107482711806287,
+      "loss": 2.0261,
+      "step": 9764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00028097716830966597,
+      "loss": 2.0568,
+      "step": 9765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002808795198396502,
+      "loss": 2.0964,
+      "step": 9766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00028078188171262524,
+      "loss": 2.036,
+      "step": 9767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002806842539331996,
+      "loss": 1.9803,
+      "step": 9768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00028058663650598153,
+      "loss": 1.8938,
+      "step": 9769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00028048902943557857,
+      "loss": 1.9676,
+      "step": 9770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00028039143272659843,
+      "loss": 2.0443,
+      "step": 9771
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002802938463836477,
+      "loss": 2.0439,
+      "step": 9772
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002801962704113327,
+      "loss": 2.0308,
+      "step": 9773
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000280098704814259,
+      "loss": 2.0301,
+      "step": 9774
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002800011495970323,
+      "loss": 2.107,
+      "step": 9775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027990360476425725,
+      "loss": 1.911,
+      "step": 9776
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002798060703205383,
+      "loss": 1.9872,
+      "step": 9777
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002797085462704789,
+      "loss": 2.0978,
+      "step": 9778
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027961103261868296,
+      "loss": 2.0165,
+      "step": 9779
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027951352936975317,
+      "loss": 2.0939,
+      "step": 9780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002794160365282917,
+      "loss": 1.9779,
+      "step": 9781
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002793185540989005,
+      "loss": 2.1625,
+      "step": 9782
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027922108208618113,
+      "loss": 2.0369,
+      "step": 9783
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002791236204947345,
+      "loss": 2.166,
+      "step": 9784
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002790261693291609,
+      "loss": 2.0365,
+      "step": 9785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027892872859406004,
+      "loss": 1.9356,
+      "step": 9786
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002788312982940319,
+      "loss": 1.9877,
+      "step": 9787
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027873387843367495,
+      "loss": 2.0471,
+      "step": 9788
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002786364690175878,
+      "loss": 2.0916,
+      "step": 9789
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027853907005036824,
+      "loss": 2.1307,
+      "step": 9790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027844168153661397,
+      "loss": 2.0571,
+      "step": 9791
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027834430348092196,
+      "loss": 2.1593,
+      "step": 9792
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002782469358878883,
+      "loss": 2.024,
+      "step": 9793
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027814957876210947,
+      "loss": 2.1042,
+      "step": 9794
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027805223210818064,
+      "loss": 2.0578,
+      "step": 9795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002779548959306969,
+      "loss": 2.0191,
+      "step": 9796
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027785757023425246,
+      "loss": 2.1888,
+      "step": 9797
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002777602550234418,
+      "loss": 2.0875,
+      "step": 9798
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027766295030285813,
+      "loss": 2.2197,
+      "step": 9799
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027756565607709463,
+      "loss": 2.1045,
+      "step": 9800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027746837235074334,
+      "loss": 2.0353,
+      "step": 9801
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002773710991283969,
+      "loss": 2.0246,
+      "step": 9802
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027727383641464655,
+      "loss": 1.9848,
+      "step": 9803
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002771765842140833,
+      "loss": 2.0619,
+      "step": 9804
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002770793425312975,
+      "loss": 1.8168,
+      "step": 9805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027698211137087964,
+      "loss": 2.0947,
+      "step": 9806
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027688489073741896,
+      "loss": 1.9315,
+      "step": 9807
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002767876806355045,
+      "loss": 1.9378,
+      "step": 9808
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002766904810697246,
+      "loss": 2.0224,
+      "step": 9809
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027659329204466775,
+      "loss": 2.0485,
+      "step": 9810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027649611356492135,
+      "loss": 1.9144,
+      "step": 9811
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002763989456350723,
+      "loss": 1.9915,
+      "step": 9812
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002763017882597071,
+      "loss": 1.9094,
+      "step": 9813
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027620464144341206,
+      "loss": 2.123,
+      "step": 9814
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027610750519077267,
+      "loss": 2.0638,
+      "step": 9815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002760103795063738,
+      "loss": 2.1363,
+      "step": 9816
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002759132643948,
+      "loss": 2.0795,
+      "step": 9817
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027581615986063565,
+      "loss": 2.0122,
+      "step": 9818
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002757190659084641,
+      "loss": 1.994,
+      "step": 9819
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027562198254286834,
+      "loss": 2.0654,
+      "step": 9820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027552490976843086,
+      "loss": 1.9695,
+      "step": 9821
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027542784758973404,
+      "loss": 2.1179,
+      "step": 9822
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002753307960113593,
+      "loss": 2.132,
+      "step": 9823
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002752337550378876,
+      "loss": 1.9868,
+      "step": 9824
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002751367246738994,
+      "loss": 2.0841,
+      "step": 9825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002750397049239751,
+      "loss": 2.1533,
+      "step": 9826
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027494269579269405,
+      "loss": 2.1243,
+      "step": 9827
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027484569728463527,
+      "loss": 2.0097,
+      "step": 9828
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027474870940437726,
+      "loss": 1.9653,
+      "step": 9829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002746517321564983,
+      "loss": 2.0178,
+      "step": 9830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002745547655455758,
+      "loss": 1.954,
+      "step": 9831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027445780957618683,
+      "loss": 1.8727,
+      "step": 9832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027436086425290776,
+      "loss": 2.2957,
+      "step": 9833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002742639295803149,
+      "loss": 2.0401,
+      "step": 9834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002741670055629838,
+      "loss": 2.1011,
+      "step": 9835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002740700922054893,
+      "loss": 2.145,
+      "step": 9836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027397318951240577,
+      "loss": 1.947,
+      "step": 9837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002738762974883078,
+      "loss": 1.9999,
+      "step": 9838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002737794161377686,
+      "loss": 2.1437,
+      "step": 9839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002736825454653611,
+      "loss": 2.0004,
+      "step": 9840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027358568547565775,
+      "loss": 1.94,
+      "step": 9841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002734888361732309,
+      "loss": 2.0636,
+      "step": 9842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027339199756265195,
+      "loss": 2.0828,
+      "step": 9843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027329516964849177,
+      "loss": 2.0955,
+      "step": 9844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002731983524353208,
+      "loss": 1.8678,
+      "step": 9845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027310154592770933,
+      "loss": 2.0439,
+      "step": 9846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027300475013022663,
+      "loss": 2.0825,
+      "step": 9847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027290796504744183,
+      "loss": 1.9673,
+      "step": 9848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000272811190683923,
+      "loss": 1.9482,
+      "step": 9849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027271442704423875,
+      "loss": 2.1428,
+      "step": 9850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027261767413295617,
+      "loss": 2.1462,
+      "step": 9851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027252093195464226,
+      "loss": 1.8945,
+      "step": 9852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027242420051386333,
+      "loss": 2.0852,
+      "step": 9853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002723274798151857,
+      "loss": 2.0834,
+      "step": 9854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002722307698631745,
+      "loss": 2.0461,
+      "step": 9855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002721340706623949,
+      "loss": 1.9631,
+      "step": 9856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027203738221741086,
+      "loss": 2.0674,
+      "step": 9857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027194070453278685,
+      "loss": 2.1708,
+      "step": 9858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002718440376130861,
+      "loss": 2.1026,
+      "step": 9859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027174738146287135,
+      "loss": 2.0682,
+      "step": 9860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027165073608670506,
+      "loss": 2.0345,
+      "step": 9861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002715541014891493,
+      "loss": 2.1189,
+      "step": 9862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002714574776747652,
+      "loss": 2.0402,
+      "step": 9863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002713608646481138,
+      "loss": 2.024,
+      "step": 9864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002712642624137552,
+      "loss": 2.058,
+      "step": 9865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002711676709762496,
+      "loss": 2.1,
+      "step": 9866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027107109034015615,
+      "loss": 2.0687,
+      "step": 9867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002709745205100337,
+      "loss": 2.0592,
+      "step": 9868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002708779614904404,
+      "loss": 2.0285,
+      "step": 9869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027078141328593444,
+      "loss": 2.125,
+      "step": 9870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002706848759010729,
+      "loss": 2.0771,
+      "step": 9871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027058834934041255,
+      "loss": 2.1008,
+      "step": 9872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002704918336085094,
+      "loss": 1.9892,
+      "step": 9873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002703953287099198,
+      "loss": 2.0743,
+      "step": 9874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002702988346491987,
+      "loss": 1.975,
+      "step": 9875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027020235143090075,
+      "loss": 2.1291,
+      "step": 9876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00027010587905958015,
+      "loss": 2.0727,
+      "step": 9877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000270009417539791,
+      "loss": 2.064,
+      "step": 9878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002699129668760861,
+      "loss": 1.977,
+      "step": 9879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002698165270730184,
+      "loss": 2.0676,
+      "step": 9880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026972009813513976,
+      "loss": 2.0716,
+      "step": 9881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002696236800670022,
+      "loss": 1.9771,
+      "step": 9882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002695272728731567,
+      "loss": 2.1049,
+      "step": 9883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002694308765581538,
+      "loss": 2.0142,
+      "step": 9884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000269334491126544,
+      "loss": 2.0833,
+      "step": 9885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026923811658287653,
+      "loss": 2.0183,
+      "step": 9886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002691417529317008,
+      "loss": 2.1138,
+      "step": 9887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026904540017756495,
+      "loss": 2.0653,
+      "step": 9888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026894905832501755,
+      "loss": 2.0792,
+      "step": 9889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002688527273786059,
+      "loss": 1.9922,
+      "step": 9890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002687564073428771,
+      "loss": 1.9326,
+      "step": 9891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002686600982223775,
+      "loss": 1.9805,
+      "step": 9892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002685638000216534,
+      "loss": 2.1116,
+      "step": 9893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002684675127452502,
+      "loss": 2.0315,
+      "step": 9894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026837123639771287,
+      "loss": 2.0734,
+      "step": 9895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002682749709835857,
+      "loss": 2.02,
+      "step": 9896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026817871650741286,
+      "loss": 2.0185,
+      "step": 9897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026808247297373776,
+      "loss": 1.8943,
+      "step": 9898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002679862403871033,
+      "loss": 2.1067,
+      "step": 9899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002678900187520517,
+      "loss": 2.1085,
+      "step": 9900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000267793808073125,
+      "loss": 2.0446,
+      "step": 9901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002676976083548648,
+      "loss": 1.9432,
+      "step": 9902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026760141960181154,
+      "loss": 1.9495,
+      "step": 9903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026750524181850556,
+      "loss": 1.9775,
+      "step": 9904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002674090750094871,
+      "loss": 2.0545,
+      "step": 9905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002673129191792951,
+      "loss": 1.9264,
+      "step": 9906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002672167743324684,
+      "loss": 2.092,
+      "step": 9907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002671206404735451,
+      "loss": 2.0688,
+      "step": 9908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026702451760706337,
+      "loss": 2.0486,
+      "step": 9909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026692840573756013,
+      "loss": 1.9465,
+      "step": 9910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026683230486957213,
+      "loss": 2.0444,
+      "step": 9911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00026673621500763534,
+      "loss": 2.0644,
+      "step": 9912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002666401361562858,
+      "loss": 2.0393,
+      "step": 9913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026654406832005856,
+      "loss": 2.0213,
+      "step": 9914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002664480115034881,
+      "loss": 2.1379,
+      "step": 9915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002663519657111085,
+      "loss": 1.9663,
+      "step": 9916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026625593094745353,
+      "loss": 2.038,
+      "step": 9917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026615990721705617,
+      "loss": 1.943,
+      "step": 9918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026606389452444896,
+      "loss": 2.0611,
+      "step": 9919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002659678928741637,
+      "loss": 2.0777,
+      "step": 9920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026587190227073235,
+      "loss": 1.8928,
+      "step": 9921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002657759227186856,
+      "loss": 2.2108,
+      "step": 9922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026567995422255384,
+      "loss": 2.1247,
+      "step": 9923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000265583996786867,
+      "loss": 2.1176,
+      "step": 9924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026548805041615473,
+      "loss": 1.8305,
+      "step": 9925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002653921151149459,
+      "loss": 1.9311,
+      "step": 9926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002652961908877686,
+      "loss": 1.9374,
+      "step": 9927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026520027773915074,
+      "loss": 2.1495,
+      "step": 9928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026510437567361987,
+      "loss": 2.0098,
+      "step": 9929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002650084846957027,
+      "loss": 1.9837,
+      "step": 9930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002649126048099254,
+      "loss": 2.1436,
+      "step": 9931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002648167360208137,
+      "loss": 1.8711,
+      "step": 9932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026472087833289294,
+      "loss": 1.9277,
+      "step": 9933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026462503175068785,
+      "loss": 2.1225,
+      "step": 9934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002645291962787226,
+      "loss": 2.0626,
+      "step": 9935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026443337192152055,
+      "loss": 2.1607,
+      "step": 9936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002643375586836053,
+      "loss": 2.07,
+      "step": 9937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026424175656949924,
+      "loss": 2.0091,
+      "step": 9938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002641459655837244,
+      "loss": 1.9604,
+      "step": 9939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002640501857308021,
+      "loss": 2.2064,
+      "step": 9940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002639544170152539,
+      "loss": 2.0413,
+      "step": 9941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002638586594416001,
+      "loss": 2.0245,
+      "step": 9942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002637629130143606,
+      "loss": 2.0077,
+      "step": 9943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002636671777380546,
+      "loss": 2.1312,
+      "step": 9944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026357145361720157,
+      "loss": 1.9795,
+      "step": 9945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026347574065631964,
+      "loss": 2.0112,
+      "step": 9946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002633800388599267,
+      "loss": 2.1803,
+      "step": 9947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002632843482325398,
+      "loss": 2.0935,
+      "step": 9948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026318866877867624,
+      "loss": 1.9859,
+      "step": 9949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002630930005028521,
+      "loss": 2.0416,
+      "step": 9950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002629973434095832,
+      "loss": 1.883,
+      "step": 9951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002629016975033845,
+      "loss": 1.9665,
+      "step": 9952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002628060627887711,
+      "loss": 1.9598,
+      "step": 9953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002627104392702571,
+      "loss": 1.9208,
+      "step": 9954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002626148269523559,
+      "loss": 2.1691,
+      "step": 9955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026251922583958066,
+      "loss": 2.1479,
+      "step": 9956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026242363593644426,
+      "loss": 1.9822,
+      "step": 9957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026232805724745866,
+      "loss": 1.9513,
+      "step": 9958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002622324897771353,
+      "loss": 2.0621,
+      "step": 9959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002621369335299849,
+      "loss": 2.0443,
+      "step": 9960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026204138851051864,
+      "loss": 1.9953,
+      "step": 9961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026194585472324593,
+      "loss": 2.0838,
+      "step": 9962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026185033217267635,
+      "loss": 1.9894,
+      "step": 9963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026175482086331856,
+      "loss": 2.0812,
+      "step": 9964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002616593207996813,
+      "loss": 1.9437,
+      "step": 9965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002615638319862722,
+      "loss": 2.1186,
+      "step": 9966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002614683544275986,
+      "loss": 2.1837,
+      "step": 9967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026137288812816693,
+      "loss": 2.0054,
+      "step": 9968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026127743309248405,
+      "loss": 1.9591,
+      "step": 9969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026118198932505523,
+      "loss": 2.0731,
+      "step": 9970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002610865568303857,
+      "loss": 2.0626,
+      "step": 9971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026099113561297996,
+      "loss": 1.9458,
+      "step": 9972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026089572567734236,
+      "loss": 2.0413,
+      "step": 9973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026080032702797647,
+      "loss": 2.0897,
+      "step": 9974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026070493966938495,
+      "loss": 2.0675,
+      "step": 9975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026060956360607086,
+      "loss": 2.01,
+      "step": 9976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002605141988425358,
+      "loss": 2.0859,
+      "step": 9977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002604188453832814,
+      "loss": 2.0539,
+      "step": 9978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002603235032328082,
+      "loss": 1.9946,
+      "step": 9979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026022817239561703,
+      "loss": 1.9133,
+      "step": 9980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002601328528762075,
+      "loss": 1.9866,
+      "step": 9981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00026003754467907893,
+      "loss": 2.0561,
+      "step": 9982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002599422478087299,
+      "loss": 1.9471,
+      "step": 9983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025984696226965897,
+      "loss": 1.9206,
+      "step": 9984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002597516880663638,
+      "loss": 2.0045,
+      "step": 9985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002596564252033414,
+      "loss": 2.01,
+      "step": 9986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025956117368508823,
+      "loss": 2.0214,
+      "step": 9987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002594659335161008,
+      "loss": 1.9667,
+      "step": 9988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002593707047008744,
+      "loss": 1.9943,
+      "step": 9989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002592754872439042,
+      "loss": 2.1443,
+      "step": 9990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002591802811496844,
+      "loss": 1.9664,
+      "step": 9991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002590850864227092,
+      "loss": 2.1032,
+      "step": 9992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025898990306747204,
+      "loss": 2.1226,
+      "step": 9993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002588947310884657,
+      "loss": 1.9147,
+      "step": 9994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002587995704901823,
+      "loss": 1.9804,
+      "step": 9995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002587044212771141,
+      "loss": 2.0602,
+      "step": 9996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025860928345375215,
+      "loss": 2.0177,
+      "step": 9997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000258514157024587,
+      "loss": 2.0445,
+      "step": 9998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025841904199410895,
+      "loss": 2.027,
+      "step": 9999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002583239383668077,
+      "loss": 1.95,
+      "step": 10000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025822884614717245,
+      "loss": 2.0038,
+      "step": 10001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002581337653396916,
+      "loss": 1.6844,
+      "step": 10002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025803869594885306,
+      "loss": 2.0054,
+      "step": 10003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025794363797914467,
+      "loss": 2.2108,
+      "step": 10004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025784859143505325,
+      "loss": 2.1353,
+      "step": 10005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002577535563210652,
+      "loss": 2.0055,
+      "step": 10006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025765853264166614,
+      "loss": 2.0988,
+      "step": 10007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002575635204013419,
+      "loss": 2.0592,
+      "step": 10008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025746851960457693,
+      "loss": 2.1023,
+      "step": 10009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025737353025585564,
+      "loss": 2.2111,
+      "step": 10010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025727855235966154,
+      "loss": 1.9181,
+      "step": 10011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025718358592047806,
+      "loss": 2.0019,
+      "step": 10012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002570886309427878,
+      "loss": 1.9893,
+      "step": 10013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025699368743107263,
+      "loss": 2.1025,
+      "step": 10014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002568987553898142,
+      "loss": 2.0926,
+      "step": 10015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002568038348234937,
+      "loss": 1.992,
+      "step": 10016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025670892573659147,
+      "loss": 2.125,
+      "step": 10017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002566140281335875,
+      "loss": 2.1393,
+      "step": 10018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002565191420189608,
+      "loss": 1.8875,
+      "step": 10019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002564242673971908,
+      "loss": 1.8912,
+      "step": 10020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025632940427275554,
+      "loss": 1.9765,
+      "step": 10021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002562345526501327,
+      "loss": 1.9536,
+      "step": 10022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002561397125337994,
+      "loss": 2.0136,
+      "step": 10023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002560448839282326,
+      "loss": 2.0789,
+      "step": 10024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002559500668379083,
+      "loss": 2.0714,
+      "step": 10025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002558552612673022,
+      "loss": 2.0951,
+      "step": 10026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025576046722088894,
+      "loss": 1.974,
+      "step": 10027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025566568470314344,
+      "loss": 2.0777,
+      "step": 10028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025557091371853957,
+      "loss": 2.0585,
+      "step": 10029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002554761542715507,
+      "loss": 2.1722,
+      "step": 10030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002553814063666495,
+      "loss": 2.0911,
+      "step": 10031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002552866700083086,
+      "loss": 2.0907,
+      "step": 10032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025519194520099965,
+      "loss": 1.9703,
+      "step": 10033
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002550972319491939,
+      "loss": 2.0375,
+      "step": 10034
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002550025302573618,
+      "loss": 2.0293,
+      "step": 10035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002549078401299738,
+      "loss": 1.9973,
+      "step": 10036
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002548131615714994,
+      "loss": 2.077,
+      "step": 10037
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025471849458640766,
+      "loss": 1.9506,
+      "step": 10038
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025462383917916685,
+      "loss": 1.9726,
+      "step": 10039
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002545291953542453,
+      "loss": 1.9165,
+      "step": 10040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002544345631161102,
+      "loss": 2.0779,
+      "step": 10041
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025433994246922855,
+      "loss": 2.0694,
+      "step": 10042
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002542453334180662,
+      "loss": 2.0218,
+      "step": 10043
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002541507359670896,
+      "loss": 2.1622,
+      "step": 10044
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025405615012076354,
+      "loss": 1.8432,
+      "step": 10045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025396157588355284,
+      "loss": 2.002,
+      "step": 10046
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025386701325992133,
+      "loss": 2.0221,
+      "step": 10047
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025377246225433305,
+      "loss": 2.0003,
+      "step": 10048
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002536779228712508,
+      "loss": 2.0658,
+      "step": 10049
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025358339511513705,
+      "loss": 2.1169,
+      "step": 10050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025348887899045346,
+      "loss": 2.0255,
+      "step": 10051
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000253394374501662,
+      "loss": 2.0103,
+      "step": 10052
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002532998816532231,
+      "loss": 1.9673,
+      "step": 10053
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002532054004495971,
+      "loss": 1.8868,
+      "step": 10054
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002531109308952436,
+      "loss": 1.8563,
+      "step": 10055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025301647299462204,
+      "loss": 2.0034,
+      "step": 10056
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000252922026752191,
+      "loss": 2.0624,
+      "step": 10057
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00025282759217240845,
+      "loss": 1.9794,
+      "step": 10058
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002527331692597319,
+      "loss": 1.9289,
+      "step": 10059
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002526387580186185,
+      "loss": 2.0206,
+      "step": 10060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002525443584535245,
+      "loss": 1.9529,
+      "step": 10061
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000252449970568906,
+      "loss": 1.9909,
+      "step": 10062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025235559436921786,
+      "loss": 1.8033,
+      "step": 10063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025226122985891537,
+      "loss": 2.0098,
+      "step": 10064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025216687704245256,
+      "loss": 2.0311,
+      "step": 10065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002520725359242829,
+      "loss": 2.0337,
+      "step": 10066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025197820650885996,
+      "loss": 2.1279,
+      "step": 10067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025188388880063595,
+      "loss": 2.0547,
+      "step": 10068
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025178958280406297,
+      "loss": 2.0393,
+      "step": 10069
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025169528852359236,
+      "loss": 1.9371,
+      "step": 10070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002516010059636753,
+      "loss": 1.9804,
+      "step": 10071
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000251506735128762,
+      "loss": 2.2101,
+      "step": 10072
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002514124760233023,
+      "loss": 2.133,
+      "step": 10073
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025131822865174516,
+      "loss": 1.9867,
+      "step": 10074
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025122399301853965,
+      "loss": 2.0205,
+      "step": 10075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025112976912813385,
+      "loss": 2.0283,
+      "step": 10076
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025103555698497527,
+      "loss": 1.9865,
+      "step": 10077
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002509413565935107,
+      "loss": 1.9925,
+      "step": 10078
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002508471679581871,
+      "loss": 2.1173,
+      "step": 10079
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002507529910834501,
+      "loss": 2.0815,
+      "step": 10080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025065882597374514,
+      "loss": 1.8589,
+      "step": 10081
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002505646726335168,
+      "loss": 2.0941,
+      "step": 10082
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025047053106720966,
+      "loss": 2.0527,
+      "step": 10083
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025037640127926737,
+      "loss": 1.9722,
+      "step": 10084
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00025028228327413294,
+      "loss": 2.2214,
+      "step": 10085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002501881770562489,
+      "loss": 1.8988,
+      "step": 10086
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002500940826300575,
+      "loss": 2.0929,
+      "step": 10087
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002500000000000001,
+      "loss": 1.974,
+      "step": 10088
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002499059291705176,
+      "loss": 2.1198,
+      "step": 10089
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002498118701460503,
+      "loss": 2.0005,
+      "step": 10090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002497178229310382,
+      "loss": 2.016,
+      "step": 10091
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002496237875299204,
+      "loss": 1.9828,
+      "step": 10092
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002495297639471356,
+      "loss": 2.0334,
+      "step": 10093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024943575218712183,
+      "loss": 1.9061,
+      "step": 10094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024934175225431687,
+      "loss": 2.0267,
+      "step": 10095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002492477641531577,
+      "loss": 1.9404,
+      "step": 10096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002491537878880807,
+      "loss": 2.2071,
+      "step": 10097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002490598234635216,
+      "loss": 1.9334,
+      "step": 10098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002489658708839161,
+      "loss": 1.9583,
+      "step": 10099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024887193015369876,
+      "loss": 1.9374,
+      "step": 10100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024877800127730385,
+      "loss": 2.0824,
+      "step": 10101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002486840842591648,
+      "loss": 2.1079,
+      "step": 10102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002485901791037151,
+      "loss": 2.0779,
+      "step": 10103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024849628581538716,
+      "loss": 2.056,
+      "step": 10104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024840240439861285,
+      "loss": 1.9774,
+      "step": 10105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024830853485782343,
+      "loss": 2.0589,
+      "step": 10106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002482146771974502,
+      "loss": 2.0455,
+      "step": 10107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002481208314219233,
+      "loss": 1.9347,
+      "step": 10108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002480269975356723,
+      "loss": 2.2164,
+      "step": 10109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024793317554312635,
+      "loss": 1.9894,
+      "step": 10110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024783936544871434,
+      "loss": 1.9873,
+      "step": 10111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024774556725686416,
+      "loss": 1.9781,
+      "step": 10112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002476517809720033,
+      "loss": 2.045,
+      "step": 10113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024755800659855855,
+      "loss": 1.9688,
+      "step": 10114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002474642441409566,
+      "loss": 1.9123,
+      "step": 10115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000247370493603623,
+      "loss": 2.0485,
+      "step": 10116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002472767549909832,
+      "loss": 1.9692,
+      "step": 10117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024718302830746146,
+      "loss": 2.0741,
+      "step": 10118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024708931355748237,
+      "loss": 1.9612,
+      "step": 10119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024699561074546935,
+      "loss": 2.0715,
+      "step": 10120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024690191987584525,
+      "loss": 2.0256,
+      "step": 10121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024680824095303243,
+      "loss": 2.0404,
+      "step": 10122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002467145739814531,
+      "loss": 2.0179,
+      "step": 10123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024662091896552837,
+      "loss": 2.2326,
+      "step": 10124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002465272759096789,
+      "loss": 2.0078,
+      "step": 10125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024643364481832477,
+      "loss": 2.2154,
+      "step": 10126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024634002569588585,
+      "loss": 2.1482,
+      "step": 10127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002462464185467811,
+      "loss": 1.9113,
+      "step": 10128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024615282337542895,
+      "loss": 2.002,
+      "step": 10129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002460592401862471,
+      "loss": 1.9402,
+      "step": 10130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002459656689836533,
+      "loss": 1.9887,
+      "step": 10131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002458721097720641,
+      "loss": 1.994,
+      "step": 10132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002457785625558957,
+      "loss": 1.8675,
+      "step": 10133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002456850273395636,
+      "loss": 2.0858,
+      "step": 10134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002455915041274833,
+      "loss": 2.2422,
+      "step": 10135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000245497992924069,
+      "loss": 2.0527,
+      "step": 10136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024540449373373473,
+      "loss": 2.0548,
+      "step": 10137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002453110065608936,
+      "loss": 1.9287,
+      "step": 10138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024521753140995895,
+      "loss": 1.9203,
+      "step": 10139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024512406828534265,
+      "loss": 2.0782,
+      "step": 10140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024503061719145645,
+      "loss": 1.929,
+      "step": 10141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024493717813271136,
+      "loss": 1.9598,
+      "step": 10142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024484375111351813,
+      "loss": 2.0772,
+      "step": 10143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024475033613828665,
+      "loss": 1.9768,
+      "step": 10144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024465693321142637,
+      "loss": 2.034,
+      "step": 10145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024456354233734576,
+      "loss": 1.8753,
+      "step": 10146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002444701635204536,
+      "loss": 2.0569,
+      "step": 10147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024437679676515736,
+      "loss": 1.9413,
+      "step": 10148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002442834420758642,
+      "loss": 1.9046,
+      "step": 10149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024419009945698034,
+      "loss": 2.1431,
+      "step": 10150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024409676891291233,
+      "loss": 1.9925,
+      "step": 10151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002440034504480653,
+      "loss": 1.9102,
+      "step": 10152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024391014406684392,
+      "loss": 1.9564,
+      "step": 10153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024381684977365286,
+      "loss": 1.9928,
+      "step": 10154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024372356757289566,
+      "loss": 2.0101,
+      "step": 10155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024363029746897542,
+      "loss": 1.9046,
+      "step": 10156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024353703946629446,
+      "loss": 2.1323,
+      "step": 10157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002434437935692553,
+      "loss": 2.1509,
+      "step": 10158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024335055978225896,
+      "loss": 2.1879,
+      "step": 10159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024325733810970647,
+      "loss": 1.9729,
+      "step": 10160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024316412855599786,
+      "loss": 2.2241,
+      "step": 10161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002430709311255332,
+      "loss": 2.1141,
+      "step": 10162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002429777458227115,
+      "loss": 2.1234,
+      "step": 10163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000242884572651931,
+      "loss": 2.0334,
+      "step": 10164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024279141161759015,
+      "loss": 2.0722,
+      "step": 10165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024269826272408596,
+      "loss": 2.1834,
+      "step": 10166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024260512597581576,
+      "loss": 2.1295,
+      "step": 10167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024251200137717543,
+      "loss": 2.0552,
+      "step": 10168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024241888893256086,
+      "loss": 1.9159,
+      "step": 10169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024232578864636685,
+      "loss": 2.041,
+      "step": 10170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024223270052298835,
+      "loss": 2.0514,
+      "step": 10171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002421396245668192,
+      "loss": 1.9424,
+      "step": 10172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002420465607822528,
+      "loss": 2.0393,
+      "step": 10173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002419535091736817,
+      "loss": 2.0936,
+      "step": 10174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024186046974549857,
+      "loss": 2.0385,
+      "step": 10175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024176744250209493,
+      "loss": 1.9685,
+      "step": 10176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024167442744786185,
+      "loss": 2.0768,
+      "step": 10177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024158142458718967,
+      "loss": 2.0114,
+      "step": 10178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024148843392446867,
+      "loss": 2.0006,
+      "step": 10179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024139545546408815,
+      "loss": 1.9855,
+      "step": 10180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002413024892104368,
+      "loss": 2.0521,
+      "step": 10181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024120953516790274,
+      "loss": 2.1175,
+      "step": 10182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024111659334087387,
+      "loss": 1.97,
+      "step": 10183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024102366373373717,
+      "loss": 2.0764,
+      "step": 10184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024093074635087914,
+      "loss": 2.0035,
+      "step": 10185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024083784119668545,
+      "loss": 1.9937,
+      "step": 10186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002407449482755418,
+      "loss": 2.0874,
+      "step": 10187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024065206759183285,
+      "loss": 2.0543,
+      "step": 10188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024055919914994268,
+      "loss": 2.0321,
+      "step": 10189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024046634295425473,
+      "loss": 1.9638,
+      "step": 10190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024037349900915246,
+      "loss": 1.9015,
+      "step": 10191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002402806673190181,
+      "loss": 2.1017,
+      "step": 10192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024018784788823344,
+      "loss": 1.9019,
+      "step": 10193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024009504072117978,
+      "loss": 1.946,
+      "step": 10194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00024000224582223796,
+      "loss": 2.0553,
+      "step": 10195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002399094631957881,
+      "loss": 2.0514,
+      "step": 10196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023981669284620972,
+      "loss": 2.134,
+      "step": 10197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023972393477788157,
+      "loss": 1.979,
+      "step": 10198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023963118899518243,
+      "loss": 2.0995,
+      "step": 10199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023953845550248992,
+      "loss": 2.0209,
+      "step": 10200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002394457343041813,
+      "loss": 1.979,
+      "step": 10201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023935302540463305,
+      "loss": 1.9063,
+      "step": 10202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023926032880822157,
+      "loss": 2.1356,
+      "step": 10203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023916764451932214,
+      "loss": 2.0024,
+      "step": 10204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023907497254230982,
+      "loss": 2.0665,
+      "step": 10205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023898231288155858,
+      "loss": 2.0127,
+      "step": 10206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002388896655414426,
+      "loss": 2.0981,
+      "step": 10207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002387970305263349,
+      "loss": 2.1373,
+      "step": 10208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023870440784060803,
+      "loss": 2.1015,
+      "step": 10209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023861179748863382,
+      "loss": 2.0034,
+      "step": 10210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00023851919947478413,
+      "loss": 1.9658,
+      "step": 10211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023842661380342944,
+      "loss": 2.1639,
+      "step": 10212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002383340404789402,
+      "loss": 2.0379,
+      "step": 10213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023824147950568577,
+      "loss": 2.0067,
+      "step": 10214
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023814893088803564,
+      "loss": 2.0411,
+      "step": 10215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023805639463035812,
+      "loss": 1.9902,
+      "step": 10216
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002379638707370212,
+      "loss": 2.0614,
+      "step": 10217
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000237871359212392,
+      "loss": 1.9687,
+      "step": 10218
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002377788600608376,
+      "loss": 1.8139,
+      "step": 10219
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023768637328672404,
+      "loss": 1.9822,
+      "step": 10220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002375938988944169,
+      "loss": 2.1045,
+      "step": 10221
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023750143688828102,
+      "loss": 1.9488,
+      "step": 10222
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023740898727268118,
+      "loss": 2.1946,
+      "step": 10223
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023731655005198094,
+      "loss": 1.8642,
+      "step": 10224
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002372241252305437,
+      "loss": 2.0363,
+      "step": 10225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023713171281273182,
+      "loss": 2.0098,
+      "step": 10226
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002370393128029078,
+      "loss": 2.0796,
+      "step": 10227
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023694692520543294,
+      "loss": 2.069,
+      "step": 10228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023685455002466816,
+      "loss": 2.1094,
+      "step": 10229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023676218726497356,
+      "loss": 2.1973,
+      "step": 10230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023666983693070926,
+      "loss": 2.0939,
+      "step": 10231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023657749902623426,
+      "loss": 2.0187,
+      "step": 10232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023648517355590688,
+      "loss": 2.1288,
+      "step": 10233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023639286052408554,
+      "loss": 1.8523,
+      "step": 10234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023630055993512734,
+      "loss": 1.9585,
+      "step": 10235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023620827179338916,
+      "loss": 2.1522,
+      "step": 10236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000236115996103227,
+      "loss": 2.1364,
+      "step": 10237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002360237328689968,
+      "loss": 2.0076,
+      "step": 10238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023593148209505356,
+      "loss": 2.0992,
+      "step": 10239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023583924378575154,
+      "loss": 1.9792,
+      "step": 10240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023574701794544457,
+      "loss": 2.144,
+      "step": 10241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023565480457848626,
+      "loss": 1.9546,
+      "step": 10242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023556260368922904,
+      "loss": 2.1088,
+      "step": 10243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023547041528202507,
+      "loss": 1.9614,
+      "step": 10244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002353782393612256,
+      "loss": 2.0817,
+      "step": 10245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023528607593118202,
+      "loss": 1.9349,
+      "step": 10246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023519392499624436,
+      "loss": 2.0047,
+      "step": 10247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023510178656076242,
+      "loss": 1.9905,
+      "step": 10248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023500966062908513,
+      "loss": 2.0745,
+      "step": 10249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023491754720556148,
+      "loss": 2.0611,
+      "step": 10250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023482544629453917,
+      "loss": 2.1269,
+      "step": 10251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023473335790036564,
+      "loss": 2.0526,
+      "step": 10252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023464128202738742,
+      "loss": 2.0341,
+      "step": 10253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023454921867995115,
+      "loss": 1.8994,
+      "step": 10254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023445716786240218,
+      "loss": 2.0173,
+      "step": 10255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023436512957908558,
+      "loss": 2.0094,
+      "step": 10256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002342731038343455,
+      "loss": 1.9415,
+      "step": 10257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023418109063252625,
+      "loss": 2.0898,
+      "step": 10258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023408908997797078,
+      "loss": 1.9533,
+      "step": 10259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023399710187502182,
+      "loss": 1.9971,
+      "step": 10260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023390512632802114,
+      "loss": 2.047,
+      "step": 10261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002338131633413106,
+      "loss": 2.1569,
+      "step": 10262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023372121291923093,
+      "loss": 2.018,
+      "step": 10263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023362927506612237,
+      "loss": 1.9849,
+      "step": 10264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023353734978632436,
+      "loss": 2.0628,
+      "step": 10265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023344543708417643,
+      "loss": 2.1103,
+      "step": 10266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023335353696401684,
+      "loss": 2.0539,
+      "step": 10267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023326164943018353,
+      "loss": 2.0972,
+      "step": 10268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002331697744870136,
+      "loss": 1.9558,
+      "step": 10269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002330779121388441,
+      "loss": 2.1144,
+      "step": 10270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023298606239001096,
+      "loss": 2.0351,
+      "step": 10271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023289422524484977,
+      "loss": 2.0236,
+      "step": 10272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002328024007076952,
+      "loss": 2.0692,
+      "step": 10273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023271058878288197,
+      "loss": 1.9462,
+      "step": 10274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023261878947474364,
+      "loss": 1.9842,
+      "step": 10275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002325270027876133,
+      "loss": 2.037,
+      "step": 10276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023243522872582335,
+      "loss": 2.0365,
+      "step": 10277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002323434672937061,
+      "loss": 2.0445,
+      "step": 10278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023225171849559274,
+      "loss": 2.0451,
+      "step": 10279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023215998233581397,
+      "loss": 2.0855,
+      "step": 10280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002320682588186998,
+      "loss": 2.0268,
+      "step": 10281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002319765479485802,
+      "loss": 1.924,
+      "step": 10282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023188484972978384,
+      "loss": 1.9704,
+      "step": 10283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002317931641666392,
+      "loss": 2.0901,
+      "step": 10284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023170149126347379,
+      "loss": 2.258,
+      "step": 10285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002316098310246152,
+      "loss": 1.9253,
+      "step": 10286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023151818345438973,
+      "loss": 1.9321,
+      "step": 10287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023142654855712352,
+      "loss": 2.0225,
+      "step": 10288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002313349263371416,
+      "loss": 2.1166,
+      "step": 10289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002312433167987692,
+      "loss": 2.1961,
+      "step": 10290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023115171994633028,
+      "loss": 1.9764,
+      "step": 10291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023106013578414847,
+      "loss": 2.0272,
+      "step": 10292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002309685643165465,
+      "loss": 1.9823,
+      "step": 10293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002308770055478472,
+      "loss": 2.0811,
+      "step": 10294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002307854594823721,
+      "loss": 1.9799,
+      "step": 10295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023069392612444246,
+      "loss": 2.0274,
+      "step": 10296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023060240547837863,
+      "loss": 2.0245,
+      "step": 10297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023051089754850098,
+      "loss": 2.0603,
+      "step": 10298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023041940233912867,
+      "loss": 2.0206,
+      "step": 10299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002303279198545805,
+      "loss": 1.8545,
+      "step": 10300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023023645009917455,
+      "loss": 1.9505,
+      "step": 10301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023014499307722863,
+      "loss": 1.9647,
+      "step": 10302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023005354879305966,
+      "loss": 2.0105,
+      "step": 10303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000229962117250984,
+      "loss": 1.967,
+      "step": 10304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022987069845531715,
+      "loss": 2.0416,
+      "step": 10305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022977929241037483,
+      "loss": 2.0316,
+      "step": 10306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022968789912047123,
+      "loss": 2.0142,
+      "step": 10307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022959651858992048,
+      "loss": 1.9921,
+      "step": 10308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022950515082303564,
+      "loss": 2.226,
+      "step": 10309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022941379582412996,
+      "loss": 2.0672,
+      "step": 10310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002293224535975154,
+      "loss": 2.0254,
+      "step": 10311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022923112414750346,
+      "loss": 1.9331,
+      "step": 10312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022913980747840497,
+      "loss": 2.1066,
+      "step": 10313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022904850359453067,
+      "loss": 1.9296,
+      "step": 10314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002289572125001901,
+      "loss": 1.9828,
+      "step": 10315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022886593419969237,
+      "loss": 1.9681,
+      "step": 10316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022877466869734587,
+      "loss": 1.9193,
+      "step": 10317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022868341599745895,
+      "loss": 1.914,
+      "step": 10318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022859217610433875,
+      "loss": 2.0847,
+      "step": 10319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022850094902229196,
+      "loss": 1.9101,
+      "step": 10320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022840973475562454,
+      "loss": 2.149,
+      "step": 10321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022831853330864245,
+      "loss": 2.0245,
+      "step": 10322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002282273446856503,
+      "loss": 1.8959,
+      "step": 10323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002281361688909523,
+      "loss": 2.2018,
+      "step": 10324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022804500592885253,
+      "loss": 2.1363,
+      "step": 10325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022795385580365386,
+      "loss": 1.9354,
+      "step": 10326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022786271851965885,
+      "loss": 2.069,
+      "step": 10327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022777159408116914,
+      "loss": 2.1337,
+      "step": 10328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022768048249248646,
+      "loss": 1.9032,
+      "step": 10329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002275893837579112,
+      "loss": 1.9969,
+      "step": 10330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022749829788174346,
+      "loss": 2.014,
+      "step": 10331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022740722486828258,
+      "loss": 1.8899,
+      "step": 10332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022731616472182776,
+      "loss": 2.0206,
+      "step": 10333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000227225117446677,
+      "loss": 2.0868,
+      "step": 10334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022713408304712808,
+      "loss": 1.9242,
+      "step": 10335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022704306152747768,
+      "loss": 2.1325,
+      "step": 10336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002269520528920227,
+      "loss": 2.1321,
+      "step": 10337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022686105714505873,
+      "loss": 1.9262,
+      "step": 10338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022677007429088098,
+      "loss": 1.9905,
+      "step": 10339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002266791043337839,
+      "loss": 2.0197,
+      "step": 10340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022658814727806182,
+      "loss": 1.9507,
+      "step": 10341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022649720312800791,
+      "loss": 2.0408,
+      "step": 10342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022640627188791508,
+      "loss": 2.0144,
+      "step": 10343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022631535356207512,
+      "loss": 1.9756,
+      "step": 10344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022622444815478004,
+      "loss": 2.0725,
+      "step": 10345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022613355567032068,
+      "loss": 2.1146,
+      "step": 10346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022604267611298724,
+      "loss": 2.0089,
+      "step": 10347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022595180948706929,
+      "loss": 2.1431,
+      "step": 10348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022586095579685638,
+      "loss": 1.9748,
+      "step": 10349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002257701150466368,
+      "loss": 2.0539,
+      "step": 10350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002256792872406984,
+      "loss": 1.9536,
+      "step": 10351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002255884723833283,
+      "loss": 2.1033,
+      "step": 10352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022549767047881354,
+      "loss": 2.0209,
+      "step": 10353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022540688153144,
+      "loss": 1.9924,
+      "step": 10354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022531610554549314,
+      "loss": 2.1128,
+      "step": 10355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022522534252525763,
+      "loss": 2.0438,
+      "step": 10356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022513459247501798,
+      "loss": 2.0598,
+      "step": 10357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002250438553990577,
+      "loss": 1.9632,
+      "step": 10358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022495313130165974,
+      "loss": 2.1331,
+      "step": 10359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002248624201871063,
+      "loss": 2.0241,
+      "step": 10360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00022477172205967954,
+      "loss": 2.064,
+      "step": 10361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022468103692366043,
+      "loss": 2.0205,
+      "step": 10362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022459036478332956,
+      "loss": 2.1126,
+      "step": 10363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022449970564296663,
+      "loss": 2.1095,
+      "step": 10364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002244090595068513,
+      "loss": 1.8513,
+      "step": 10365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022431842637926215,
+      "loss": 1.871,
+      "step": 10366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022422780626447724,
+      "loss": 1.8785,
+      "step": 10367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022413719916677388,
+      "loss": 2.025,
+      "step": 10368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022404660509042936,
+      "loss": 2.0751,
+      "step": 10369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002239560240397196,
+      "loss": 1.9438,
+      "step": 10370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022386545601892034,
+      "loss": 2.0422,
+      "step": 10371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002237749010323064,
+      "loss": 1.8983,
+      "step": 10372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022368435908415258,
+      "loss": 2.0384,
+      "step": 10373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022359383017873236,
+      "loss": 1.9195,
+      "step": 10374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022350331432031907,
+      "loss": 2.0904,
+      "step": 10375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000223412811513185,
+      "loss": 2.0633,
+      "step": 10376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022332232176160244,
+      "loss": 2.1075,
+      "step": 10377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022323184506984256,
+      "loss": 1.9895,
+      "step": 10378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002231413814421761,
+      "loss": 2.1077,
+      "step": 10379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022305093088287293,
+      "loss": 1.9157,
+      "step": 10380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002229604933962029,
+      "loss": 2.0044,
+      "step": 10381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002228700689864347,
+      "loss": 2.0239,
+      "step": 10382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022277965765783654,
+      "loss": 1.8914,
+      "step": 10383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022268925941467584,
+      "loss": 2.1432,
+      "step": 10384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022259887426122,
+      "loss": 2.0164,
+      "step": 10385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022250850220173525,
+      "loss": 1.893,
+      "step": 10386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022241814324048737,
+      "loss": 2.0795,
+      "step": 10387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022232779738174126,
+      "loss": 1.8458,
+      "step": 10388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002222374646297618,
+      "loss": 2.0548,
+      "step": 10389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022214714498881283,
+      "loss": 1.881,
+      "step": 10390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022205683846315762,
+      "loss": 1.9989,
+      "step": 10391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022196654505705855,
+      "loss": 1.8901,
+      "step": 10392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022187626477477817,
+      "loss": 2.1702,
+      "step": 10393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022178599762057765,
+      "loss": 2.1314,
+      "step": 10394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022169574359871792,
+      "loss": 2.0545,
+      "step": 10395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022160550271345885,
+      "loss": 1.9695,
+      "step": 10396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002215152749690605,
+      "loss": 2.1127,
+      "step": 10397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022142506036978155,
+      "loss": 2.0378,
+      "step": 10398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022133485891988038,
+      "loss": 2.0084,
+      "step": 10399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022124467062361453,
+      "loss": 1.9514,
+      "step": 10400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022115449548524148,
+      "loss": 2.008,
+      "step": 10401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002210643335090175,
+      "loss": 2.0862,
+      "step": 10402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022097418469919837,
+      "loss": 2.0643,
+      "step": 10403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022088404906003928,
+      "loss": 2.0037,
+      "step": 10404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022079392659579513,
+      "loss": 1.8737,
+      "step": 10405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002207038173107197,
+      "loss": 2.0841,
+      "step": 10406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002206137212090664,
+      "loss": 2.0621,
+      "step": 10407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022052363829508777,
+      "loss": 1.89,
+      "step": 10408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002204335685730363,
+      "loss": 2.1434,
+      "step": 10409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022034351204716328,
+      "loss": 2.0801,
+      "step": 10410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022025346872171965,
+      "loss": 2.1865,
+      "step": 10411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00022016343860095539,
+      "loss": 2.0216,
+      "step": 10412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002200734216891206,
+      "loss": 2.0737,
+      "step": 10413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000219983417990464,
+      "loss": 1.7945,
+      "step": 10414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021989342750923385,
+      "loss": 2.0758,
+      "step": 10415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002198034502496783,
+      "loss": 2.1052,
+      "step": 10416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002197134862160442,
+      "loss": 1.9711,
+      "step": 10417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021962353541257813,
+      "loss": 2.1259,
+      "step": 10418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021953359784352577,
+      "loss": 1.9799,
+      "step": 10419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021944367351313282,
+      "loss": 2.0217,
+      "step": 10420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002193537624256436,
+      "loss": 2.0601,
+      "step": 10421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021926386458530223,
+      "loss": 2.0435,
+      "step": 10422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002191739799963518,
+      "loss": 1.9381,
+      "step": 10423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021908410866303558,
+      "loss": 2.0011,
+      "step": 10424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021899425058959543,
+      "loss": 2.0409,
+      "step": 10425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021890440578027287,
+      "loss": 2.2313,
+      "step": 10426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021881457423930857,
+      "loss": 1.9412,
+      "step": 10427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002187247559709432,
+      "loss": 2.1134,
+      "step": 10428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021863495097941622,
+      "loss": 1.9995,
+      "step": 10429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021854515926896656,
+      "loss": 1.9242,
+      "step": 10430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021845538084383249,
+      "loss": 1.8885,
+      "step": 10431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021836561570825208,
+      "loss": 1.8932,
+      "step": 10432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021827586386646226,
+      "loss": 2.0103,
+      "step": 10433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021818612532269955,
+      "loss": 1.9273,
+      "step": 10434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021809640008119957,
+      "loss": 1.9708,
+      "step": 10435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021800668814619806,
+      "loss": 2.1535,
+      "step": 10436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021791698952192929,
+      "loss": 2.0935,
+      "step": 10437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021782730421262737,
+      "loss": 1.8604,
+      "step": 10438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002177376322225254,
+      "loss": 2.0826,
+      "step": 10439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021764797355585647,
+      "loss": 1.9457,
+      "step": 10440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002175583282168525,
+      "loss": 2.015,
+      "step": 10441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021746869620974498,
+      "loss": 2.0772,
+      "step": 10442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021737907753876452,
+      "loss": 1.9096,
+      "step": 10443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021728947220814172,
+      "loss": 2.0028,
+      "step": 10444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000217199880222106,
+      "loss": 2.0,
+      "step": 10445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021711030158488626,
+      "loss": 2.1928,
+      "step": 10446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021702073630071067,
+      "loss": 2.1366,
+      "step": 10447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021693118437380728,
+      "loss": 1.8364,
+      "step": 10448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021684164580840294,
+      "loss": 1.9367,
+      "step": 10449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021675212060872406,
+      "loss": 1.8736,
+      "step": 10450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002166626087789963,
+      "loss": 2.0304,
+      "step": 10451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021657311032344516,
+      "loss": 2.0038,
+      "step": 10452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021648362524629507,
+      "loss": 2.0262,
+      "step": 10453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021639415355176978,
+      "loss": 2.1041,
+      "step": 10454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021630469524409253,
+      "loss": 2.0891,
+      "step": 10455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021621525032748624,
+      "loss": 2.0115,
+      "step": 10456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021612581880617272,
+      "loss": 1.9695,
+      "step": 10457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021603640068437337,
+      "loss": 2.0507,
+      "step": 10458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021594699596630873,
+      "loss": 2.1378,
+      "step": 10459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002158576046561993,
+      "loss": 2.077,
+      "step": 10460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002157682267582644,
+      "loss": 1.9701,
+      "step": 10461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021567886227672283,
+      "loss": 1.9966,
+      "step": 10462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021558951121579262,
+      "loss": 1.9167,
+      "step": 10463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002155001735796917,
+      "loss": 2.1095,
+      "step": 10464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021541084937263684,
+      "loss": 2.0724,
+      "step": 10465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021532153859884435,
+      "loss": 1.949,
+      "step": 10466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021523224126252972,
+      "loss": 2.0516,
+      "step": 10467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002151429573679084,
+      "loss": 1.9662,
+      "step": 10468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021505368691919453,
+      "loss": 2.029,
+      "step": 10469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021496442992060194,
+      "loss": 1.9439,
+      "step": 10470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021487518637634356,
+      "loss": 1.8429,
+      "step": 10471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021478595629063224,
+      "loss": 2.0461,
+      "step": 10472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021469673966767973,
+      "loss": 1.9687,
+      "step": 10473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021460753651169717,
+      "loss": 1.9593,
+      "step": 10474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021451834682689503,
+      "loss": 2.0893,
+      "step": 10475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021442917061748358,
+      "loss": 2.0322,
+      "step": 10476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021434000788767205,
+      "loss": 2.0385,
+      "step": 10477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021425085864166905,
+      "loss": 2.0999,
+      "step": 10478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021416172288368248,
+      "loss": 1.9928,
+      "step": 10479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021407260061792006,
+      "loss": 2.0439,
+      "step": 10480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021398349184858844,
+      "loss": 2.0558,
+      "step": 10481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002138943965798938,
+      "loss": 2.116,
+      "step": 10482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021380531481604136,
+      "loss": 2.0701,
+      "step": 10483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021371624656123646,
+      "loss": 1.9483,
+      "step": 10484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021362719181968297,
+      "loss": 2.0519,
+      "step": 10485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021353815059558468,
+      "loss": 2.1203,
+      "step": 10486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021344912289314428,
+      "loss": 2.0493,
+      "step": 10487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021336010871656443,
+      "loss": 1.9475,
+      "step": 10488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021327110807004658,
+      "loss": 1.9349,
+      "step": 10489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021318212095779182,
+      "loss": 2.0099,
+      "step": 10490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021309314738400038,
+      "loss": 2.0631,
+      "step": 10491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002130041873528723,
+      "loss": 1.8927,
+      "step": 10492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021291524086860665,
+      "loss": 1.8275,
+      "step": 10493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002128263079354018,
+      "loss": 2.0142,
+      "step": 10494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021273738855745545,
+      "loss": 2.0154,
+      "step": 10495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021264848273896514,
+      "loss": 2.0807,
+      "step": 10496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021255959048412727,
+      "loss": 1.8467,
+      "step": 10497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021247071179713773,
+      "loss": 2.0335,
+      "step": 10498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021238184668219163,
+      "loss": 2.1432,
+      "step": 10499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000212292995143484,
+      "loss": 1.9836,
+      "step": 10500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021220415718520858,
+      "loss": 2.119,
+      "step": 10501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021211533281155888,
+      "loss": 2.1214,
+      "step": 10502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021202652202672724,
+      "loss": 1.9825,
+      "step": 10503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002119377248349062,
+      "loss": 2.1618,
+      "step": 10504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021184894124028704,
+      "loss": 2.0289,
+      "step": 10505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021176017124706025,
+      "loss": 2.0163,
+      "step": 10506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002116714148594165,
+      "loss": 2.1482,
+      "step": 10507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021158267208154497,
+      "loss": 2.103,
+      "step": 10508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021149394291763464,
+      "loss": 2.1167,
+      "step": 10509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00021140522737187346,
+      "loss": 2.0185,
+      "step": 10510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021131652544844937,
+      "loss": 2.0802,
+      "step": 10511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002112278371515492,
+      "loss": 2.0234,
+      "step": 10512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002111391624853592,
+      "loss": 2.0962,
+      "step": 10513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002110505014540648,
+      "loss": 1.905,
+      "step": 10514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021096185406185142,
+      "loss": 1.9539,
+      "step": 10515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021087322031290324,
+      "loss": 2.0577,
+      "step": 10516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021078460021140395,
+      "loss": 2.0499,
+      "step": 10517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021069599376153641,
+      "loss": 1.9312,
+      "step": 10518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002106074009674835,
+      "loss": 2.0479,
+      "step": 10519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021051882183342673,
+      "loss": 1.926,
+      "step": 10520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021043025636354728,
+      "loss": 2.077,
+      "step": 10521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021034170456202551,
+      "loss": 1.9968,
+      "step": 10522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002102531664330416,
+      "loss": 2.0001,
+      "step": 10523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021016464198077452,
+      "loss": 2.0709,
+      "step": 10524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00021007613120940284,
+      "loss": 1.9459,
+      "step": 10525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020998763412310433,
+      "loss": 1.9668,
+      "step": 10526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020989915072605665,
+      "loss": 1.9292,
+      "step": 10527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020981068102243617,
+      "loss": 1.8145,
+      "step": 10528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020972222501641885,
+      "loss": 2.113,
+      "step": 10529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020963378271217993,
+      "loss": 1.7956,
+      "step": 10530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002095453541138943,
+      "loss": 2.0642,
+      "step": 10531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020945693922573598,
+      "loss": 1.9707,
+      "step": 10532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020936853805187828,
+      "loss": 2.1019,
+      "step": 10533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020928015059649375,
+      "loss": 1.993,
+      "step": 10534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020919177686375484,
+      "loss": 1.9771,
+      "step": 10535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020910341685783285,
+      "loss": 1.9708,
+      "step": 10536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020901507058289852,
+      "loss": 2.1012,
+      "step": 10537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002089267380431219,
+      "loss": 2.0689,
+      "step": 10538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020883841924267279,
+      "loss": 1.9259,
+      "step": 10539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020875011418571983,
+      "loss": 2.0459,
+      "step": 10540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020866182287643126,
+      "loss": 2.0255,
+      "step": 10541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020857354531897448,
+      "loss": 1.9881,
+      "step": 10542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020848528151751672,
+      "loss": 1.8965,
+      "step": 10543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020839703147622408,
+      "loss": 2.0235,
+      "step": 10544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002083087951992621,
+      "loss": 2.0227,
+      "step": 10545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020822057269079565,
+      "loss": 2.0819,
+      "step": 10546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002081323639549893,
+      "loss": 1.9555,
+      "step": 10547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002080441689960066,
+      "loss": 1.9398,
+      "step": 10548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002079559878180105,
+      "loss": 2.0674,
+      "step": 10549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020786782042516322,
+      "loss": 1.9553,
+      "step": 10550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002077796668216268,
+      "loss": 1.9947,
+      "step": 10551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020769152701156213,
+      "loss": 2.1216,
+      "step": 10552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020760340099912954,
+      "loss": 1.9289,
+      "step": 10553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020751528878848875,
+      "loss": 1.9952,
+      "step": 10554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002074271903837991,
+      "loss": 2.0142,
+      "step": 10555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020733910578921882,
+      "loss": 1.9738,
+      "step": 10556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020725103500890586,
+      "loss": 2.1185,
+      "step": 10557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020716297804701707,
+      "loss": 2.0714,
+      "step": 10558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002070749349077093,
+      "loss": 2.1137,
+      "step": 10559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020698690559513823,
+      "loss": 1.8894,
+      "step": 10560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020689889011345908,
+      "loss": 2.0788,
+      "step": 10561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020681088846682616,
+      "loss": 2.0648,
+      "step": 10562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002067229006593937,
+      "loss": 2.0641,
+      "step": 10563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002066349266953148,
+      "loss": 2.0342,
+      "step": 10564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000206546966578742,
+      "loss": 2.0439,
+      "step": 10565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020645902031382702,
+      "loss": 1.9471,
+      "step": 10566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020637108790472154,
+      "loss": 2.1004,
+      "step": 10567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020628316935557595,
+      "loss": 2.1126,
+      "step": 10568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020619526467054023,
+      "loss": 2.0591,
+      "step": 10569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020610737385376348,
+      "loss": 2.2687,
+      "step": 10570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002060194969093947,
+      "loss": 2.0565,
+      "step": 10571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020593163384158175,
+      "loss": 1.8297,
+      "step": 10572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020584378465447195,
+      "loss": 1.9971,
+      "step": 10573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020575594935221182,
+      "loss": 2.0454,
+      "step": 10574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020566812793894768,
+      "loss": 2.0518,
+      "step": 10575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002055803204188248,
+      "loss": 2.0086,
+      "step": 10576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020549252679598784,
+      "loss": 2.0133,
+      "step": 10577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020540474707458068,
+      "loss": 2.0728,
+      "step": 10578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020531698125874715,
+      "loss": 1.9761,
+      "step": 10579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020522922935262977,
+      "loss": 2.144,
+      "step": 10580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020514149136037062,
+      "loss": 1.8687,
+      "step": 10581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020505376728611102,
+      "loss": 1.9742,
+      "step": 10582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020496605713399207,
+      "loss": 1.9928,
+      "step": 10583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020487836090815365,
+      "loss": 1.9262,
+      "step": 10584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002047906786127353,
+      "loss": 2.0111,
+      "step": 10585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020470301025187566,
+      "loss": 2.1258,
+      "step": 10586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002046153558297132,
+      "loss": 1.9493,
+      "step": 10587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020452771535038516,
+      "loss": 2.0539,
+      "step": 10588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002044400888180285,
+      "loss": 2.0599,
+      "step": 10589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020435247623677916,
+      "loss": 1.9673,
+      "step": 10590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020426487761077302,
+      "loss": 2.1774,
+      "step": 10591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002041772929441447,
+      "loss": 2.0244,
+      "step": 10592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002040897222410285,
+      "loss": 2.0328,
+      "step": 10593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020400216550555772,
+      "loss": 2.0656,
+      "step": 10594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020391462274186562,
+      "loss": 2.0092,
+      "step": 10595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020382709395408422,
+      "loss": 2.0031,
+      "step": 10596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020373957914634488,
+      "loss": 2.1567,
+      "step": 10597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020365207832277887,
+      "loss": 2.1622,
+      "step": 10598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020356459148751632,
+      "loss": 1.9317,
+      "step": 10599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020347711864468672,
+      "loss": 1.8984,
+      "step": 10600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020338965979841884,
+      "loss": 2.0336,
+      "step": 10601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020330221495284135,
+      "loss": 1.9862,
+      "step": 10602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020321478411208156,
+      "loss": 2.0209,
+      "step": 10603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020312736728026647,
+      "loss": 2.0138,
+      "step": 10604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020303996446152222,
+      "loss": 2.1503,
+      "step": 10605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020295257565997477,
+      "loss": 1.9737,
+      "step": 10606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002028652008797488,
+      "loss": 2.0456,
+      "step": 10607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020277784012496863,
+      "loss": 2.1011,
+      "step": 10608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020269049339975782,
+      "loss": 2.0672,
+      "step": 10609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002026031607082396,
+      "loss": 1.9802,
+      "step": 10610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020251584205453605,
+      "loss": 1.9615,
+      "step": 10611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020242853744276897,
+      "loss": 2.0732,
+      "step": 10612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020234124687705895,
+      "loss": 2.0122,
+      "step": 10613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002022539703615268,
+      "loss": 2.1158,
+      "step": 10614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000202166707900292,
+      "loss": 2.0903,
+      "step": 10615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020207945949747336,
+      "loss": 2.0543,
+      "step": 10616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002019922251571895,
+      "loss": 2.1196,
+      "step": 10617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020190500488355772,
+      "loss": 2.0507,
+      "step": 10618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020181779868069544,
+      "loss": 2.0036,
+      "step": 10619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020173060655271879,
+      "loss": 2.0092,
+      "step": 10620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002016434285037434,
+      "loss": 1.884,
+      "step": 10621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020155626453788417,
+      "loss": 2.0299,
+      "step": 10622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020146911465925571,
+      "loss": 1.9907,
+      "step": 10623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020138197887197161,
+      "loss": 1.9725,
+      "step": 10624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002012948571801448,
+      "loss": 1.9301,
+      "step": 10625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020120774958788752,
+      "loss": 2.0434,
+      "step": 10626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020112065609931168,
+      "loss": 2.0702,
+      "step": 10627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020103357671852828,
+      "loss": 1.9703,
+      "step": 10628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020094651144964755,
+      "loss": 2.0499,
+      "step": 10629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020085946029677897,
+      "loss": 2.0835,
+      "step": 10630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020077242326403205,
+      "loss": 1.9372,
+      "step": 10631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020068540035551486,
+      "loss": 1.976,
+      "step": 10632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002005983915753351,
+      "loss": 2.0546,
+      "step": 10633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020051139692759957,
+      "loss": 1.9591,
+      "step": 10634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020042441641641507,
+      "loss": 1.8401,
+      "step": 10635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020033745004588698,
+      "loss": 1.9359,
+      "step": 10636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002002504978201204,
+      "loss": 2.1387,
+      "step": 10637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00020016355974321947,
+      "loss": 1.8869,
+      "step": 10638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002000766358192882,
+      "loss": 2.0319,
+      "step": 10639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001999897260524295,
+      "loss": 1.9861,
+      "step": 10640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019990283044674562,
+      "loss": 2.0723,
+      "step": 10641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019981594900633808,
+      "loss": 1.9338,
+      "step": 10642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019972908173530824,
+      "loss": 2.1091,
+      "step": 10643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019964222863775633,
+      "loss": 1.9695,
+      "step": 10644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019955538971778186,
+      "loss": 2.1015,
+      "step": 10645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019946856497948386,
+      "loss": 2.0898,
+      "step": 10646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019938175442696078,
+      "loss": 2.1202,
+      "step": 10647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019929495806431024,
+      "loss": 2.0974,
+      "step": 10648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019920817589562923,
+      "loss": 1.997,
+      "step": 10649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019912140792501388,
+      "loss": 2.0469,
+      "step": 10650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019903465415656014,
+      "loss": 2.0882,
+      "step": 10651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001989479145943629,
+      "loss": 2.0286,
+      "step": 10652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019886118924251634,
+      "loss": 2.0556,
+      "step": 10653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019877447810511402,
+      "loss": 2.1977,
+      "step": 10654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019868778118624925,
+      "loss": 2.1635,
+      "step": 10655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019860109849001412,
+      "loss": 2.0608,
+      "step": 10656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019851443002050022,
+      "loss": 1.8007,
+      "step": 10657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019842777578179844,
+      "loss": 2.0906,
+      "step": 10658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001983411357779993,
+      "loss": 2.154,
+      "step": 10659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019825451001319222,
+      "loss": 1.904,
+      "step": 10660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019816789849146622,
+      "loss": 1.9216,
+      "step": 10661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019808130121690937,
+      "loss": 2.0382,
+      "step": 10662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001979947181936096,
+      "loss": 2.1291,
+      "step": 10663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001979081494256536,
+      "loss": 2.0213,
+      "step": 10664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001978215949171277,
+      "loss": 2.1181,
+      "step": 10665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019773505467211722,
+      "loss": 1.9579,
+      "step": 10666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019764852869470745,
+      "loss": 2.0631,
+      "step": 10667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019756201698898246,
+      "loss": 2.0785,
+      "step": 10668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019747551955902576,
+      "loss": 1.9869,
+      "step": 10669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019738903640892009,
+      "loss": 1.9544,
+      "step": 10670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019730256754274794,
+      "loss": 2.0538,
+      "step": 10671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019721611296459076,
+      "loss": 2.0717,
+      "step": 10672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019712967267852916,
+      "loss": 2.0595,
+      "step": 10673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019704324668864365,
+      "loss": 2.0376,
+      "step": 10674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019695683499901362,
+      "loss": 2.0843,
+      "step": 10675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019687043761371792,
+      "loss": 2.1076,
+      "step": 10676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019678405453683444,
+      "loss": 2.0748,
+      "step": 10677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019669768577244108,
+      "loss": 2.0197,
+      "step": 10678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019661133132461438,
+      "loss": 1.9898,
+      "step": 10679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019652499119743062,
+      "loss": 2.0544,
+      "step": 10680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019643866539496496,
+      "loss": 2.077,
+      "step": 10681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001963523539212926,
+      "loss": 2.0404,
+      "step": 10682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019626605678048736,
+      "loss": 1.9673,
+      "step": 10683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019617977397662278,
+      "loss": 2.0384,
+      "step": 10684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019609350551377135,
+      "loss": 1.8587,
+      "step": 10685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019600725139600557,
+      "loss": 1.9314,
+      "step": 10686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019592101162739655,
+      "loss": 1.9284,
+      "step": 10687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019583478621201507,
+      "loss": 1.9184,
+      "step": 10688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019574857515393103,
+      "loss": 2.0716,
+      "step": 10689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019566237845721407,
+      "loss": 2.0442,
+      "step": 10690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001955761961259327,
+      "loss": 1.938,
+      "step": 10691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019549002816415494,
+      "loss": 2.0378,
+      "step": 10692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019540387457594798,
+      "loss": 1.9856,
+      "step": 10693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019531773536537883,
+      "loss": 1.8495,
+      "step": 10694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019523161053651317,
+      "loss": 2.0052,
+      "step": 10695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019514550009341641,
+      "loss": 1.9459,
+      "step": 10696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019505940404015287,
+      "loss": 2.092,
+      "step": 10697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019497332238078697,
+      "loss": 2.1693,
+      "step": 10698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019488725511938165,
+      "loss": 2.154,
+      "step": 10699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019480120225999958,
+      "loss": 2.1247,
+      "step": 10700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019471516380670245,
+      "loss": 2.0302,
+      "step": 10701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019462913976355184,
+      "loss": 1.9731,
+      "step": 10702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019454313013460807,
+      "loss": 2.039,
+      "step": 10703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019445713492393098,
+      "loss": 2.0472,
+      "step": 10704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019437115413557958,
+      "loss": 2.0016,
+      "step": 10705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001942851877736128,
+      "loss": 2.0852,
+      "step": 10706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019419923584208814,
+      "loss": 1.9896,
+      "step": 10707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019411329834506286,
+      "loss": 2.0265,
+      "step": 10708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019402737528659308,
+      "loss": 2.094,
+      "step": 10709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019394146667073503,
+      "loss": 1.935,
+      "step": 10710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019385557250154362,
+      "loss": 2.0594,
+      "step": 10711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019376969278307317,
+      "loss": 1.991,
+      "step": 10712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001936838275193773,
+      "loss": 2.0496,
+      "step": 10713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001935979767145094,
+      "loss": 2.0553,
+      "step": 10714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019351214037252156,
+      "loss": 1.8748,
+      "step": 10715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019342631849746556,
+      "loss": 2.0035,
+      "step": 10716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019334051109339218,
+      "loss": 1.9408,
+      "step": 10717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019325471816435203,
+      "loss": 2.0702,
+      "step": 10718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001931689397143946,
+      "loss": 1.9772,
+      "step": 10719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019308317574756885,
+      "loss": 2.0937,
+      "step": 10720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019299742626792282,
+      "loss": 1.9237,
+      "step": 10721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001929116912795045,
+      "loss": 2.189,
+      "step": 10722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001928259707863605,
+      "loss": 2.0007,
+      "step": 10723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001927402647925371,
+      "loss": 1.9582,
+      "step": 10724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019265457330207959,
+      "loss": 1.991,
+      "step": 10725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019256889631903324,
+      "loss": 2.208,
+      "step": 10726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019248323384744192,
+      "loss": 2.15,
+      "step": 10727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001923975858913492,
+      "loss": 2.0701,
+      "step": 10728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001923119524547976,
+      "loss": 1.9817,
+      "step": 10729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001922263335418296,
+      "loss": 2.0269,
+      "step": 10730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019214072915648645,
+      "loss": 2.0573,
+      "step": 10731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019205513930280887,
+      "loss": 2.0098,
+      "step": 10732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019196956398483673,
+      "loss": 1.9926,
+      "step": 10733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001918840032066097,
+      "loss": 1.9838,
+      "step": 10734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019179845697216624,
+      "loss": 2.166,
+      "step": 10735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019171292528554442,
+      "loss": 2.0293,
+      "step": 10736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019162740815078128,
+      "loss": 1.9546,
+      "step": 10737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019154190557191386,
+      "loss": 1.9503,
+      "step": 10738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019145641755297777,
+      "loss": 1.9489,
+      "step": 10739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019137094409800838,
+      "loss": 1.9593,
+      "step": 10740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019128548521104,
+      "loss": 1.9603,
+      "step": 10741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019120004089610683,
+      "loss": 2.2126,
+      "step": 10742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019111461115724183,
+      "loss": 2.041,
+      "step": 10743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019102919599847757,
+      "loss": 1.975,
+      "step": 10744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001909437954238456,
+      "loss": 2.2094,
+      "step": 10745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019085840943737736,
+      "loss": 1.9444,
+      "step": 10746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019077303804310313,
+      "loss": 2.1843,
+      "step": 10747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019068768124505265,
+      "loss": 1.9032,
+      "step": 10748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001906023390472547,
+      "loss": 2.0915,
+      "step": 10749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019051701145373812,
+      "loss": 2.1673,
+      "step": 10750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019043169846853027,
+      "loss": 2.0158,
+      "step": 10751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001903464000956581,
+      "loss": 2.0016,
+      "step": 10752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001902611163391479,
+      "loss": 2.1578,
+      "step": 10753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001901758472030254,
+      "loss": 2.0422,
+      "step": 10754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001900905926913154,
+      "loss": 2.1021,
+      "step": 10755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019000535280804215,
+      "loss": 2.1172,
+      "step": 10756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018992012755722894,
+      "loss": 2.0127,
+      "step": 10757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018983491694289894,
+      "loss": 2.1309,
+      "step": 10758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018974972096907417,
+      "loss": 2.0162,
+      "step": 10759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018966453963977604,
+      "loss": 1.9547,
+      "step": 10760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018957937295902518,
+      "loss": 2.1134,
+      "step": 10761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018949422093084183,
+      "loss": 2.0736,
+      "step": 10762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001894090835592454,
+      "loss": 2.0571,
+      "step": 10763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018932396084825432,
+      "loss": 2.0405,
+      "step": 10764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001892388528018869,
+      "loss": 2.1647,
+      "step": 10765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018915375942416029,
+      "loss": 1.9747,
+      "step": 10766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018906868071909107,
+      "loss": 1.9489,
+      "step": 10767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018898361669069497,
+      "loss": 1.9939,
+      "step": 10768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001888985673429876,
+      "loss": 1.9905,
+      "step": 10769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018881353267998324,
+      "loss": 1.9311,
+      "step": 10770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001887285127056958,
+      "loss": 2.0272,
+      "step": 10771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018864350742413822,
+      "loss": 2.0067,
+      "step": 10772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018855851683932323,
+      "loss": 1.9968,
+      "step": 10773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018847354095526253,
+      "loss": 2.1858,
+      "step": 10774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018838857977596706,
+      "loss": 1.9932,
+      "step": 10775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018830363330544703,
+      "loss": 1.9102,
+      "step": 10776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018821870154771253,
+      "loss": 2.1399,
+      "step": 10777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001881337845067723,
+      "loss": 2.0639,
+      "step": 10778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018804888218663463,
+      "loss": 2.0149,
+      "step": 10779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001879639945913069,
+      "loss": 2.0156,
+      "step": 10780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018787912172479643,
+      "loss": 2.1161,
+      "step": 10781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018779426359110917,
+      "loss": 2.0601,
+      "step": 10782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018770942019425064,
+      "loss": 1.8133,
+      "step": 10783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018762459153822547,
+      "loss": 1.9695,
+      "step": 10784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018753977762703811,
+      "loss": 1.8945,
+      "step": 10785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001874549784646918,
+      "loss": 2.1154,
+      "step": 10786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018737019405518928,
+      "loss": 2.0642,
+      "step": 10787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018728542440253233,
+      "loss": 1.9611,
+      "step": 10788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018720066951072268,
+      "loss": 1.897,
+      "step": 10789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018711592938376083,
+      "loss": 2.0008,
+      "step": 10790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018703120402564655,
+      "loss": 2.2997,
+      "step": 10791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000186946493440379,
+      "loss": 2.0454,
+      "step": 10792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001868617976319571,
+      "loss": 2.0025,
+      "step": 10793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018677711660437847,
+      "loss": 2.1126,
+      "step": 10794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018669245036164023,
+      "loss": 2.0784,
+      "step": 10795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018660779890773867,
+      "loss": 2.008,
+      "step": 10796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018652316224666986,
+      "loss": 2.1049,
+      "step": 10797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001864385403824287,
+      "loss": 1.9557,
+      "step": 10798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018635393331900953,
+      "loss": 2.0121,
+      "step": 10799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018626934106040578,
+      "loss": 1.9842,
+      "step": 10800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018618476361061082,
+      "loss": 1.9254,
+      "step": 10801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018610020097361662,
+      "loss": 1.9859,
+      "step": 10802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001860156531534149,
+      "loss": 1.9324,
+      "step": 10803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018593112015399615,
+      "loss": 2.0344,
+      "step": 10804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018584660197935106,
+      "loss": 2.094,
+      "step": 10805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001857620986334687,
+      "loss": 2.0308,
+      "step": 10806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018567761012033802,
+      "loss": 2.0708,
+      "step": 10807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018559313644394675,
+      "loss": 1.9119,
+      "step": 10808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018550867760828266,
+      "loss": 2.131,
+      "step": 10809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018542423361733224,
+      "loss": 2.0523,
+      "step": 10810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018533980447508135,
+      "loss": 2.087,
+      "step": 10811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001852553901855152,
+      "loss": 2.0581,
+      "step": 10812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018517099075261855,
+      "loss": 2.0187,
+      "step": 10813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001850866061803752,
+      "loss": 2.0559,
+      "step": 10814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001850022364727682,
+      "loss": 2.0205,
+      "step": 10815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018491788163377986,
+      "loss": 2.0066,
+      "step": 10816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001848335416673923,
+      "loss": 2.1181,
+      "step": 10817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018474921657758635,
+      "loss": 2.1054,
+      "step": 10818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018466490636834232,
+      "loss": 1.9793,
+      "step": 10819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001845806110436397,
+      "loss": 2.1123,
+      "step": 10820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001844963306074578,
+      "loss": 2.1226,
+      "step": 10821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018441206506377466,
+      "loss": 1.9217,
+      "step": 10822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018432781441656777,
+      "loss": 2.083,
+      "step": 10823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018424357866981379,
+      "loss": 2.0909,
+      "step": 10824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018415935782748927,
+      "loss": 2.074,
+      "step": 10825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018407515189356933,
+      "loss": 2.0031,
+      "step": 10826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018399096087202876,
+      "loss": 1.9541,
+      "step": 10827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018390678476684142,
+      "loss": 2.1297,
+      "step": 10828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018382262358198088,
+      "loss": 2.2462,
+      "step": 10829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018373847732141963,
+      "loss": 2.0585,
+      "step": 10830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001836543459891295,
+      "loss": 2.0505,
+      "step": 10831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018357022958908164,
+      "loss": 1.8769,
+      "step": 10832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018348612812524672,
+      "loss": 2.0192,
+      "step": 10833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018340204160159453,
+      "loss": 2.1872,
+      "step": 10834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018331797002209393,
+      "loss": 1.9718,
+      "step": 10835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018323391339071334,
+      "loss": 2.0765,
+      "step": 10836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018314987171142062,
+      "loss": 2.0022,
+      "step": 10837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018306584498818264,
+      "loss": 2.009,
+      "step": 10838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018298183322496553,
+      "loss": 2.2051,
+      "step": 10839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001828978364257349,
+      "loss": 1.9425,
+      "step": 10840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018281385459445565,
+      "loss": 2.0274,
+      "step": 10841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018272988773509197,
+      "loss": 2.0468,
+      "step": 10842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018264593585160722,
+      "loss": 1.9941,
+      "step": 10843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018256199894796387,
+      "loss": 1.9769,
+      "step": 10844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001824780770281244,
+      "loss": 2.0475,
+      "step": 10845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018239417009604985,
+      "loss": 2.1221,
+      "step": 10846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001823102781557009,
+      "loss": 1.888,
+      "step": 10847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018222640121103723,
+      "loss": 2.0795,
+      "step": 10848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001821425392660183,
+      "loss": 2.2082,
+      "step": 10849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001820586923246026,
+      "loss": 2.1625,
+      "step": 10850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018197486039074766,
+      "loss": 1.9926,
+      "step": 10851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018189104346841056,
+      "loss": 2.0175,
+      "step": 10852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018180724156154792,
+      "loss": 1.9852,
+      "step": 10853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018172345467411526,
+      "loss": 2.1166,
+      "step": 10854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018163968281006726,
+      "loss": 1.9068,
+      "step": 10855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018155592597335853,
+      "loss": 1.8931,
+      "step": 10856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018147218416794236,
+      "loss": 1.9476,
+      "step": 10857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018138845739777167,
+      "loss": 2.142,
+      "step": 10858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018130474566679832,
+      "loss": 2.1289,
+      "step": 10859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018122104897897403,
+      "loss": 2.0056,
+      "step": 10860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001811373673382493,
+      "loss": 2.0285,
+      "step": 10861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018105370074857413,
+      "loss": 2.1823,
+      "step": 10862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001809700492138976,
+      "loss": 2.0032,
+      "step": 10863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018088641273816851,
+      "loss": 1.9819,
+      "step": 10864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018080279132533467,
+      "loss": 1.8867,
+      "step": 10865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001807191849793431,
+      "loss": 2.0687,
+      "step": 10866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018063559370414006,
+      "loss": 2.0441,
+      "step": 10867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001805520175036716,
+      "loss": 2.0558,
+      "step": 10868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001804684563818826,
+      "loss": 2.093,
+      "step": 10869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018038491034271726,
+      "loss": 2.0244,
+      "step": 10870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018030137939011894,
+      "loss": 1.9199,
+      "step": 10871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018021786352803089,
+      "loss": 2.1163,
+      "step": 10872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018013436276039513,
+      "loss": 1.8585,
+      "step": 10873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018005087709115297,
+      "loss": 1.8784,
+      "step": 10874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017996740652424505,
+      "loss": 2.0857,
+      "step": 10875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017988395106361172,
+      "loss": 2.0576,
+      "step": 10876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017980051071319199,
+      "loss": 2.0437,
+      "step": 10877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017971708547692456,
+      "loss": 2.0329,
+      "step": 10878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017963367535874704,
+      "loss": 2.0836,
+      "step": 10879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017955028036259696,
+      "loss": 2.1221,
+      "step": 10880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001794669004924106,
+      "loss": 2.1391,
+      "step": 10881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017938353575212363,
+      "loss": 1.9885,
+      "step": 10882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001793001861456709,
+      "loss": 1.9835,
+      "step": 10883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001792168516769871,
+      "loss": 2.1734,
+      "step": 10884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017913353235000557,
+      "loss": 2.0943,
+      "step": 10885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017905022816865925,
+      "loss": 2.0238,
+      "step": 10886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017896693913688,
+      "loss": 2.103,
+      "step": 10887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017888366525859966,
+      "loss": 2.0981,
+      "step": 10888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017880040653774887,
+      "loss": 1.9182,
+      "step": 10889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017871716297825747,
+      "loss": 2.1063,
+      "step": 10890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017863393458405469,
+      "loss": 2.1156,
+      "step": 10891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017855072135906935,
+      "loss": 1.9979,
+      "step": 10892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017846752330722925,
+      "loss": 2.074,
+      "step": 10893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017838434043246143,
+      "loss": 2.0997,
+      "step": 10894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017830117273869223,
+      "loss": 2.0373,
+      "step": 10895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017821802022984762,
+      "loss": 2.1336,
+      "step": 10896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017813488290985247,
+      "loss": 1.9295,
+      "step": 10897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017805176078263096,
+      "loss": 2.0555,
+      "step": 10898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017796865385210665,
+      "loss": 1.9539,
+      "step": 10899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001778855621222025,
+      "loss": 2.0572,
+      "step": 10900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017780248559684064,
+      "loss": 2.0517,
+      "step": 10901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017771942427994243,
+      "loss": 1.9737,
+      "step": 10902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017763637817542831,
+      "loss": 2.0427,
+      "step": 10903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017755334728721873,
+      "loss": 2.2573,
+      "step": 10904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017747033161923258,
+      "loss": 1.9764,
+      "step": 10905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017738733117538853,
+      "loss": 2.0788,
+      "step": 10906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001773043459596042,
+      "loss": 1.8261,
+      "step": 10907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017722137597579697,
+      "loss": 2.0308,
+      "step": 10908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017713842122788308,
+      "loss": 2.0305,
+      "step": 10909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017705548171977826,
+      "loss": 2.1926,
+      "step": 10910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001769725574553971,
+      "loss": 1.9254,
+      "step": 10911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001768896484386543,
+      "loss": 1.9502,
+      "step": 10912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017680675467346313,
+      "loss": 1.9039,
+      "step": 10913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017672387616373642,
+      "loss": 2.0854,
+      "step": 10914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017664101291338597,
+      "loss": 2.1528,
+      "step": 10915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001765581649263235,
+      "loss": 2.0847,
+      "step": 10916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017647533220645944,
+      "loss": 2.0183,
+      "step": 10917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017639251475770374,
+      "loss": 1.9998,
+      "step": 10918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001763097125839654,
+      "loss": 1.907,
+      "step": 10919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001762269256891531,
+      "loss": 1.9466,
+      "step": 10920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017614415407717448,
+      "loss": 2.0367,
+      "step": 10921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001760613977519366,
+      "loss": 2.0253,
+      "step": 10922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017597865671734554,
+      "loss": 2.0987,
+      "step": 10923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001758959309773071,
+      "loss": 1.9938,
+      "step": 10924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017581322053572613,
+      "loss": 2.0257,
+      "step": 10925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017573052539650668,
+      "loss": 2.0277,
+      "step": 10926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017564784556355195,
+      "loss": 2.3275,
+      "step": 10927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017556518104076496,
+      "loss": 2.093,
+      "step": 10928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001754825318320476,
+      "loss": 2.0486,
+      "step": 10929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017539989794130095,
+      "loss": 2.0643,
+      "step": 10930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017531727937242542,
+      "loss": 2.0827,
+      "step": 10931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017523467612932115,
+      "loss": 2.0533,
+      "step": 10932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017515208821588708,
+      "loss": 2.0046,
+      "step": 10933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001750695156360214,
+      "loss": 2.0047,
+      "step": 10934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017498695839362167,
+      "loss": 2.0485,
+      "step": 10935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017490441649258516,
+      "loss": 2.0084,
+      "step": 10936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001748218899368077,
+      "loss": 2.1241,
+      "step": 10937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017473937873018492,
+      "loss": 1.9384,
+      "step": 10938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017465688287661123,
+      "loss": 2.0573,
+      "step": 10939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017457440237998107,
+      "loss": 1.8794,
+      "step": 10940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017449193724418749,
+      "loss": 2.005,
+      "step": 10941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017440948747312303,
+      "loss": 1.901,
+      "step": 10942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001743270530706793,
+      "loss": 2.016,
+      "step": 10943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001742446340407478,
+      "loss": 1.9849,
+      "step": 10944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017416223038721875,
+      "loss": 2.0087,
+      "step": 10945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017407984211398155,
+      "loss": 1.9024,
+      "step": 10946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017399746922492544,
+      "loss": 1.8697,
+      "step": 10947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017391511172393847,
+      "loss": 1.9942,
+      "step": 10948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001738327696149082,
+      "loss": 2.0565,
+      "step": 10949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017375044290172104,
+      "loss": 2.0003,
+      "step": 10950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017366813158826346,
+      "loss": 2.0706,
+      "step": 10951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001735858356784205,
+      "loss": 2.0482,
+      "step": 10952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017350355517607675,
+      "loss": 2.1695,
+      "step": 10953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017342129008511586,
+      "loss": 2.1237,
+      "step": 10954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017333904040942128,
+      "loss": 2.0727,
+      "step": 10955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017325680615287514,
+      "loss": 1.9897,
+      "step": 10956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00017317458731935925,
+      "loss": 2.0582,
+      "step": 10957
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017309238391275422,
+      "loss": 2.081,
+      "step": 10958
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001730101959369406,
+      "loss": 1.9318,
+      "step": 10959
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001729280233957977,
+      "loss": 1.9366,
+      "step": 10960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001728458662932043,
+      "loss": 1.9949,
+      "step": 10961
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017276372463303813,
+      "loss": 2.183,
+      "step": 10962
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017268159841917692,
+      "loss": 2.1167,
+      "step": 10963
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017259948765549694,
+      "loss": 2.0384,
+      "step": 10964
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001725173923458741,
+      "loss": 1.9487,
+      "step": 10965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017243531249418325,
+      "loss": 2.0176,
+      "step": 10966
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017235324810429913,
+      "loss": 2.0376,
+      "step": 10967
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017227119918009515,
+      "loss": 2.0024,
+      "step": 10968
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017218916572544425,
+      "loss": 2.0565,
+      "step": 10969
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017210714774421843,
+      "loss": 2.014,
+      "step": 10970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017202514524028946,
+      "loss": 1.9362,
+      "step": 10971
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017194315821752781,
+      "loss": 2.0311,
+      "step": 10972
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001718611866798036,
+      "loss": 1.9783,
+      "step": 10973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017177923063098577,
+      "loss": 2.2256,
+      "step": 10974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001716972900749432,
+      "loss": 1.9025,
+      "step": 10975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017161536501554358,
+      "loss": 2.16,
+      "step": 10976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001715334554566539,
+      "loss": 2.0288,
+      "step": 10977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001714515614021403,
+      "loss": 1.9922,
+      "step": 10978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017136968285586874,
+      "loss": 2.151,
+      "step": 10979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017128781982170388,
+      "loss": 1.9434,
+      "step": 10980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017120597230350988,
+      "loss": 1.8886,
+      "step": 10981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001711241403051499,
+      "loss": 1.8823,
+      "step": 10982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017104232383048702,
+      "loss": 2.1276,
+      "step": 10983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017096052288338286,
+      "loss": 1.9064,
+      "step": 10984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017087873746769876,
+      "loss": 2.0296,
+      "step": 10985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017079696758729495,
+      "loss": 2.0829,
+      "step": 10986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017071521324603156,
+      "loss": 2.0481,
+      "step": 10987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017063347444776726,
+      "loss": 2.1431,
+      "step": 10988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017055175119636047,
+      "loss": 2.1269,
+      "step": 10989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017047004349566842,
+      "loss": 1.9624,
+      "step": 10990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017038835134954833,
+      "loss": 1.9666,
+      "step": 10991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017030667476185607,
+      "loss": 1.8614,
+      "step": 10992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017022501373644688,
+      "loss": 1.996,
+      "step": 10993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001701433682771753,
+      "loss": 2.076,
+      "step": 10994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017006173838789545,
+      "loss": 2.0449,
+      "step": 10995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016998012407246034,
+      "loss": 2.0831,
+      "step": 10996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001698985253347223,
+      "loss": 1.9947,
+      "step": 10997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016981694217853277,
+      "loss": 2.1437,
+      "step": 10998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001697353746077431,
+      "loss": 1.9759,
+      "step": 10999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001696538226262032,
+      "loss": 2.0399,
+      "step": 11000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001695722862377626,
+      "loss": 2.102,
+      "step": 11001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001694907654462698,
+      "loss": 1.9789,
+      "step": 11002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016940926025557306,
+      "loss": 1.9708,
+      "step": 11003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016932777066951953,
+      "loss": 1.9199,
+      "step": 11004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016924629669195568,
+      "loss": 2.147,
+      "step": 11005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016916483832672702,
+      "loss": 1.9449,
+      "step": 11006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016908339557767903,
+      "loss": 2.094,
+      "step": 11007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016900196844865572,
+      "loss": 2.1184,
+      "step": 11008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016892055694350073,
+      "loss": 2.0629,
+      "step": 11009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016883916106605662,
+      "loss": 2.0179,
+      "step": 11010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001687577808201659,
+      "loss": 1.9596,
+      "step": 11011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001686764162096696,
+      "loss": 2.1316,
+      "step": 11012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016859506723840845,
+      "loss": 1.9754,
+      "step": 11013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000168513733910222,
+      "loss": 2.0095,
+      "step": 11014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001684324162289499,
+      "loss": 2.0785,
+      "step": 11015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001683511141984302,
+      "loss": 1.9007,
+      "step": 11016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016826982782250054,
+      "loss": 2.0714,
+      "step": 11017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016818855710499776,
+      "loss": 1.998,
+      "step": 11018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016810730204975833,
+      "loss": 2.0853,
+      "step": 11019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016802606266061744,
+      "loss": 1.941,
+      "step": 11020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016794483894140984,
+      "loss": 1.8381,
+      "step": 11021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016786363089596929,
+      "loss": 2.0714,
+      "step": 11022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016778243852812936,
+      "loss": 1.9492,
+      "step": 11023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001677012618417223,
+      "loss": 1.9823,
+      "step": 11024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001676201008405799,
+      "loss": 1.9,
+      "step": 11025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016753895552853288,
+      "loss": 2.0875,
+      "step": 11026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016745782590941189,
+      "loss": 1.9734,
+      "step": 11027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016737671198704623,
+      "loss": 1.9425,
+      "step": 11028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016729561376526474,
+      "loss": 2.0328,
+      "step": 11029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001672145312478952,
+      "loss": 1.9728,
+      "step": 11030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001671334644387653,
+      "loss": 1.9585,
+      "step": 11031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001670524133417013,
+      "loss": 2.0563,
+      "step": 11032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016697137796052908,
+      "loss": 2.0421,
+      "step": 11033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001668903582990735,
+      "loss": 2.02,
+      "step": 11034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016680935436115923,
+      "loss": 1.9677,
+      "step": 11035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016672836615060972,
+      "loss": 1.9868,
+      "step": 11036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016664739367124748,
+      "loss": 1.9335,
+      "step": 11037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001665664369268951,
+      "loss": 1.9474,
+      "step": 11038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016648549592137364,
+      "loss": 1.9798,
+      "step": 11039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016640457065850378,
+      "loss": 1.937,
+      "step": 11040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016632366114210513,
+      "loss": 2.2314,
+      "step": 11041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016624276737599726,
+      "loss": 2.0523,
+      "step": 11042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016616188936399823,
+      "loss": 1.915,
+      "step": 11043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001660810271099258,
+      "loss": 1.9338,
+      "step": 11044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016600018061759659,
+      "loss": 1.9595,
+      "step": 11045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016591934989082708,
+      "loss": 1.9027,
+      "step": 11046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016583853493343259,
+      "loss": 1.7546,
+      "step": 11047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001657577357492277,
+      "loss": 2.0174,
+      "step": 11048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001656769523420261,
+      "loss": 2.0012,
+      "step": 11049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016559618471564147,
+      "loss": 1.9576,
+      "step": 11050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016551543287388588,
+      "loss": 2.1734,
+      "step": 11051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016543469682057105,
+      "loss": 2.0331,
+      "step": 11052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016535397655950774,
+      "loss": 2.1042,
+      "step": 11053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016527327209450655,
+      "loss": 2.133,
+      "step": 11054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001651925834293766,
+      "loss": 1.9573,
+      "step": 11055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016511191056792668,
+      "loss": 2.071,
+      "step": 11056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001650312535139646,
+      "loss": 1.9804,
+      "step": 11057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016495061227129776,
+      "loss": 2.0208,
+      "step": 11058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016486998684373262,
+      "loss": 2.0753,
+      "step": 11059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001647893772350747,
+      "loss": 2.0799,
+      "step": 11060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016470878344912893,
+      "loss": 1.8922,
+      "step": 11061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016462820548969982,
+      "loss": 1.9927,
+      "step": 11062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016454764336059057,
+      "loss": 2.0428,
+      "step": 11063
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001644670970656041,
+      "loss": 1.961,
+      "step": 11064
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016438656660854202,
+      "loss": 2.0145,
+      "step": 11065
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000164306051993206,
+      "loss": 1.9745,
+      "step": 11066
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001642255532233963,
+      "loss": 2.06,
+      "step": 11067
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016414507030291248,
+      "loss": 1.9385,
+      "step": 11068
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016406460323555388,
+      "loss": 2.011,
+      "step": 11069
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001639841520251184,
+      "loss": 2.0766,
+      "step": 11070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016390371667540383,
+      "loss": 2.0484,
+      "step": 11071
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016382329719020672,
+      "loss": 1.9907,
+      "step": 11072
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016374289357332306,
+      "loss": 1.9671,
+      "step": 11073
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016366250582854796,
+      "loss": 2.0123,
+      "step": 11074
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016358213395967626,
+      "loss": 1.9201,
+      "step": 11075
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016350177797050147,
+      "loss": 2.1224,
+      "step": 11076
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016342143786481655,
+      "loss": 2.0049,
+      "step": 11077
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016334111364641363,
+      "loss": 2.0693,
+      "step": 11078
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016326080531908454,
+      "loss": 1.9566,
+      "step": 11079
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016318051288661983,
+      "loss": 2.0367,
+      "step": 11080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001631002363528095,
+      "loss": 2.2136,
+      "step": 11081
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016301997572144256,
+      "loss": 2.0153,
+      "step": 11082
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016293973099630794,
+      "loss": 1.9016,
+      "step": 11083
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016285950218119312,
+      "loss": 2.035,
+      "step": 11084
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016277928927988506,
+      "loss": 2.1069,
+      "step": 11085
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001626990922961699,
+      "loss": 2.0626,
+      "step": 11086
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001626189112338335,
+      "loss": 2.1262,
+      "step": 11087
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001625387460966603,
+      "loss": 2.2286,
+      "step": 11088
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016245859688843433,
+      "loss": 2.0456,
+      "step": 11089
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016237846361293867,
+      "loss": 2.013,
+      "step": 11090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016229834627395605,
+      "loss": 2.0373,
+      "step": 11091
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001622182448752682,
+      "loss": 1.8644,
+      "step": 11092
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016213815942065591,
+      "loss": 2.066,
+      "step": 11093
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001620580899138993,
+      "loss": 1.9955,
+      "step": 11094
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016197803635877817,
+      "loss": 1.9775,
+      "step": 11095
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016189799875907112,
+      "loss": 2.0606,
+      "step": 11096
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016181797711855602,
+      "loss": 2.0069,
+      "step": 11097
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016173797144100987,
+      "loss": 1.9388,
+      "step": 11098
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016165798173020963,
+      "loss": 2.0351,
+      "step": 11099
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016157800798993066,
+      "loss": 2.0872,
+      "step": 11100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016149805022394804,
+      "loss": 2.0618,
+      "step": 11101
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016141810843603566,
+      "loss": 1.8762,
+      "step": 11102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001613381826299674,
+      "loss": 2.0771,
+      "step": 11103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001612582728095157,
+      "loss": 2.0856,
+      "step": 11104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001611783789784525,
+      "loss": 2.043,
+      "step": 11105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00016109850114054885,
+      "loss": 2.096,
+      "step": 11106
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016101863929957549,
+      "loss": 2.0365,
+      "step": 11107
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001609387934593019,
+      "loss": 2.0888,
+      "step": 11108
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000160858963623497,
+      "loss": 2.0003,
+      "step": 11109
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016077914979592873,
+      "loss": 1.9443,
+      "step": 11110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001606993519803649,
+      "loss": 1.9119,
+      "step": 11111
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001606195701805719,
+      "loss": 1.8764,
+      "step": 11112
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016053980440031547,
+      "loss": 2.1735,
+      "step": 11113
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016046005464336112,
+      "loss": 2.0427,
+      "step": 11114
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000160380320913473,
+      "loss": 1.9876,
+      "step": 11115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016030060321441475,
+      "loss": 1.912,
+      "step": 11116
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016022090154994907,
+      "loss": 1.875,
+      "step": 11117
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001601412159238384,
+      "loss": 2.0494,
+      "step": 11118
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001600615463398439,
+      "loss": 2.0456,
+      "step": 11119
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001599818928017261,
+      "loss": 2.0177,
+      "step": 11120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015990225531324476,
+      "loss": 2.0486,
+      "step": 11121
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015982263387815927,
+      "loss": 2.0453,
+      "step": 11122
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001597430285002277,
+      "loss": 1.8868,
+      "step": 11123
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015966343918320775,
+      "loss": 2.0197,
+      "step": 11124
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001595838659308559,
+      "loss": 2.0129,
+      "step": 11125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015950430874692861,
+      "loss": 2.0717,
+      "step": 11126
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015942476763518098,
+      "loss": 1.927,
+      "step": 11127
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015934524259936757,
+      "loss": 2.0533,
+      "step": 11128
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015926573364324187,
+      "loss": 2.1003,
+      "step": 11129
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015918624077055732,
+      "loss": 2.0489,
+      "step": 11130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015910676398506592,
+      "loss": 2.1083,
+      "step": 11131
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015902730329051923,
+      "loss": 2.1536,
+      "step": 11132
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001589478586906678,
+      "loss": 1.959,
+      "step": 11133
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015886843018926195,
+      "loss": 2.1202,
+      "step": 11134
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015878901779005063,
+      "loss": 1.9952,
+      "step": 11135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015870962149678236,
+      "loss": 1.8704,
+      "step": 11136
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015863024131320463,
+      "loss": 2.1549,
+      "step": 11137
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001585508772430647,
+      "loss": 2.0062,
+      "step": 11138
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015847152929010866,
+      "loss": 2.029,
+      "step": 11139
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015839219745808174,
+      "loss": 2.0619,
+      "step": 11140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001583128817507286,
+      "loss": 1.8235,
+      "step": 11141
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015823358217179333,
+      "loss": 2.1168,
+      "step": 11142
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015815429872501895,
+      "loss": 2.064,
+      "step": 11143
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015807503141414787,
+      "loss": 1.9769,
+      "step": 11144
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015799578024292137,
+      "loss": 2.0213,
+      "step": 11145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001579165452150807,
+      "loss": 2.203,
+      "step": 11146
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001578373263343658,
+      "loss": 1.8897,
+      "step": 11147
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015775812360451598,
+      "loss": 2.121,
+      "step": 11148
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001576789370292696,
+      "loss": 2.0466,
+      "step": 11149
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015759976661236475,
+      "loss": 1.8764,
+      "step": 11150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001575206123575384,
+      "loss": 1.8636,
+      "step": 11151
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015744147426852669,
+      "loss": 1.9892,
+      "step": 11152
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000157362352349065,
+      "loss": 1.9759,
+      "step": 11153
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015728324660288834,
+      "loss": 2.0928,
+      "step": 11154
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015720415703373064,
+      "loss": 2.0338,
+      "step": 11155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015712508364532508,
+      "loss": 2.0015,
+      "step": 11156
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015704602644140382,
+      "loss": 1.9149,
+      "step": 11157
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015696698542569905,
+      "loss": 2.1385,
+      "step": 11158
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001568879606019414,
+      "loss": 2.0392,
+      "step": 11159
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015680895197386097,
+      "loss": 2.1276,
+      "step": 11160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015672995954518716,
+      "loss": 1.9837,
+      "step": 11161
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015665098331964884,
+      "loss": 1.9883,
+      "step": 11162
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015657202330097365,
+      "loss": 1.9947,
+      "step": 11163
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015649307949288876,
+      "loss": 2.0328,
+      "step": 11164
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015641415189912027,
+      "loss": 2.0352,
+      "step": 11165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015633524052339416,
+      "loss": 1.9576,
+      "step": 11166
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000156256345369435,
+      "loss": 2.0775,
+      "step": 11167
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015617746644096682,
+      "loss": 2.0227,
+      "step": 11168
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015609860374171276,
+      "loss": 1.8134,
+      "step": 11169
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015601975727539563,
+      "loss": 1.9684,
+      "step": 11170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015594092704573697,
+      "loss": 2.0563,
+      "step": 11171
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001558621130564578,
+      "loss": 2.0259,
+      "step": 11172
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001557833153112781,
+      "loss": 2.0454,
+      "step": 11173
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015570453381391768,
+      "loss": 1.8525,
+      "step": 11174
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015562576856809507,
+      "loss": 1.9575,
+      "step": 11175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001555470195775281,
+      "loss": 2.0273,
+      "step": 11176
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015546828684593377,
+      "loss": 2.1335,
+      "step": 11177
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015538957037702884,
+      "loss": 1.8413,
+      "step": 11178
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001553108701745286,
+      "loss": 1.8319,
+      "step": 11179
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015523218624214797,
+      "loss": 1.9542,
+      "step": 11180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001551535185836009,
+      "loss": 2.0062,
+      "step": 11181
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015507486720260083,
+      "loss": 2.1287,
+      "step": 11182
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015499623210286033,
+      "loss": 2.0477,
+      "step": 11183
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015491761328809107,
+      "loss": 1.872,
+      "step": 11184
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015483901076200385,
+      "loss": 1.996,
+      "step": 11185
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015476042452830925,
+      "loss": 1.9682,
+      "step": 11186
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015468185459071654,
+      "loss": 2.0597,
+      "step": 11187
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015460330095293446,
+      "loss": 2.1146,
+      "step": 11188
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015452476361867067,
+      "loss": 2.0533,
+      "step": 11189
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015444624259163271,
+      "loss": 1.9886,
+      "step": 11190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015436773787552672,
+      "loss": 1.8862,
+      "step": 11191
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015428924947405838,
+      "loss": 1.997,
+      "step": 11192
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001542107773909323,
+      "loss": 2.0044,
+      "step": 11193
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015413232162985284,
+      "loss": 2.034,
+      "step": 11194
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015405388219452327,
+      "loss": 2.0434,
+      "step": 11195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015397545908864602,
+      "loss": 2.0066,
+      "step": 11196
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015389705231592265,
+      "loss": 1.8817,
+      "step": 11197
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001538186618800545,
+      "loss": 2.0268,
+      "step": 11198
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001537402877847416,
+      "loss": 2.0798,
+      "step": 11199
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001536619300336834,
+      "loss": 2.0883,
+      "step": 11200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001535835886305784,
+      "loss": 1.9915,
+      "step": 11201
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015350526357912488,
+      "loss": 2.0949,
+      "step": 11202
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015342695488301973,
+      "loss": 2.1373,
+      "step": 11203
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015334866254595908,
+      "loss": 1.8603,
+      "step": 11204
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015327038657163893,
+      "loss": 1.9332,
+      "step": 11205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015319212696375396,
+      "loss": 2.0783,
+      "step": 11206
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001531138837259981,
+      "loss": 2.2023,
+      "step": 11207
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001530356568620645,
+      "loss": 2.0192,
+      "step": 11208
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015295744637564595,
+      "loss": 2.0831,
+      "step": 11209
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015287925227043403,
+      "loss": 1.9732,
+      "step": 11210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015280107455011965,
+      "loss": 1.9326,
+      "step": 11211
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015272291321839283,
+      "loss": 2.0337,
+      "step": 11212
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001526447682789433,
+      "loss": 1.9796,
+      "step": 11213
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015256663973545954,
+      "loss": 1.9324,
+      "step": 11214
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015248852759162934,
+      "loss": 2.0133,
+      "step": 11215
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015241043185113963,
+      "loss": 1.9912,
+      "step": 11216
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015233235251767703,
+      "loss": 1.923,
+      "step": 11217
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015225428959492692,
+      "loss": 2.0498,
+      "step": 11218
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015217624308657408,
+      "loss": 2.0729,
+      "step": 11219
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015209821299630227,
+      "loss": 2.1792,
+      "step": 11220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015202019932779504,
+      "loss": 2.0116,
+      "step": 11221
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015194220208473457,
+      "loss": 1.9545,
+      "step": 11222
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001518642212708027,
+      "loss": 2.0753,
+      "step": 11223
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015178625688967988,
+      "loss": 2.0284,
+      "step": 11224
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015170830894504678,
+      "loss": 2.0608,
+      "step": 11225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015163037744058243,
+      "loss": 2.1326,
+      "step": 11226
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001515524623799654,
+      "loss": 2.0697,
+      "step": 11227
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001514745637668733,
+      "loss": 1.8828,
+      "step": 11228
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015139668160498343,
+      "loss": 2.0512,
+      "step": 11229
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015131881589797186,
+      "loss": 2.0901,
+      "step": 11230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015124096664951404,
+      "loss": 1.9605,
+      "step": 11231
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015116313386328445,
+      "loss": 2.0721,
+      "step": 11232
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015108531754295734,
+      "loss": 2.1163,
+      "step": 11233
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015100751769220566,
+      "loss": 2.039,
+      "step": 11234
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015092973431470163,
+      "loss": 1.8129,
+      "step": 11235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015085196741411683,
+      "loss": 2.0786,
+      "step": 11236
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001507742169941222,
+      "loss": 2.0642,
+      "step": 11237
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001506964830583876,
+      "loss": 2.0837,
+      "step": 11238
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001506187656105823,
+      "loss": 1.9056,
+      "step": 11239
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015054106465437462,
+      "loss": 1.9328,
+      "step": 11240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001504633801934324,
+      "loss": 1.9734,
+      "step": 11241
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015038571223142255,
+      "loss": 2.0454,
+      "step": 11242
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015030806077201102,
+      "loss": 1.9423,
+      "step": 11243
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015023042581886303,
+      "loss": 2.0118,
+      "step": 11244
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015015280737564352,
+      "loss": 1.8487,
+      "step": 11245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000150075205446016,
+      "loss": 1.8786,
+      "step": 11246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014999762003364348,
+      "loss": 2.0758,
+      "step": 11247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014992005114218803,
+      "loss": 2.1157,
+      "step": 11248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001498424987753114,
+      "loss": 2.1752,
+      "step": 11249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014976496293667407,
+      "loss": 1.8189,
+      "step": 11250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014968744362993597,
+      "loss": 2.0803,
+      "step": 11251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001496099408587559,
+      "loss": 2.091,
+      "step": 11252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001495324546267926,
+      "loss": 2.0964,
+      "step": 11253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014945498493770337,
+      "loss": 2.1311,
+      "step": 11254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00014937753179514502,
+      "loss": 2.0353,
+      "step": 11255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014930009520277331,
+      "loss": 1.9368,
+      "step": 11256
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001492226751642438,
+      "loss": 2.0067,
+      "step": 11257
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014914527168321067,
+      "loss": 2.1903,
+      "step": 11258
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014906788476332756,
+      "loss": 2.0505,
+      "step": 11259
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014899051440824717,
+      "loss": 1.9886,
+      "step": 11260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014891316062162185,
+      "loss": 2.0922,
+      "step": 11261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014883582340710278,
+      "loss": 1.9165,
+      "step": 11262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014875850276834036,
+      "loss": 1.9918,
+      "step": 11263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014868119870898423,
+      "loss": 1.8985,
+      "step": 11264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014860391123268352,
+      "loss": 1.9216,
+      "step": 11265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014852664034308637,
+      "loss": 2.0954,
+      "step": 11266
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014844938604384006,
+      "loss": 2.016,
+      "step": 11267
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014837214833859097,
+      "loss": 2.0614,
+      "step": 11268
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001482949272309853,
+      "loss": 1.8932,
+      "step": 11269
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001482177227246679,
+      "loss": 1.9936,
+      "step": 11270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014814053482328294,
+      "loss": 2.1077,
+      "step": 11271
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014806336353047377,
+      "loss": 2.1412,
+      "step": 11272
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014798620884988329,
+      "loss": 1.8589,
+      "step": 11273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014790907078515332,
+      "loss": 1.9911,
+      "step": 11274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014783194933992488,
+      "loss": 2.1536,
+      "step": 11275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014775484451783817,
+      "loss": 1.9063,
+      "step": 11276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014767775632253295,
+      "loss": 1.917,
+      "step": 11277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014760068475764787,
+      "loss": 2.0004,
+      "step": 11278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001475236298268209,
+      "loss": 1.9704,
+      "step": 11279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000147446591533689,
+      "loss": 2.091,
+      "step": 11280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014736956988188894,
+      "loss": 2.1038,
+      "step": 11281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014729256487505606,
+      "loss": 2.078,
+      "step": 11282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014721557651682527,
+      "loss": 2.1078,
+      "step": 11283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014713860481083035,
+      "loss": 1.9993,
+      "step": 11284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014706164976070496,
+      "loss": 1.9718,
+      "step": 11285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001469847113700813,
+      "loss": 2.0202,
+      "step": 11286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001469077896425911,
+      "loss": 2.1324,
+      "step": 11287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014683088458186512,
+      "loss": 2.0741,
+      "step": 11288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014675399619153362,
+      "loss": 2.0075,
+      "step": 11289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014667712447522592,
+      "loss": 1.9514,
+      "step": 11290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014660026943657045,
+      "loss": 2.0288,
+      "step": 11291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001465234310791948,
+      "loss": 2.0069,
+      "step": 11292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014644660940672628,
+      "loss": 1.9102,
+      "step": 11293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001463698044227908,
+      "loss": 2.242,
+      "step": 11294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001462930161310137,
+      "loss": 1.8694,
+      "step": 11295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014621624453501974,
+      "loss": 1.9493,
+      "step": 11296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014613948963843265,
+      "loss": 1.9802,
+      "step": 11297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001460627514448754,
+      "loss": 1.9471,
+      "step": 11298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014598602995797005,
+      "loss": 2.0234,
+      "step": 11299
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014590932518133843,
+      "loss": 2.0778,
+      "step": 11300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001458326371186009,
+      "loss": 1.9983,
+      "step": 11301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014575596577337736,
+      "loss": 2.0918,
+      "step": 11302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014567931114928674,
+      "loss": 1.8494,
+      "step": 11303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001456026732499476,
+      "loss": 1.9188,
+      "step": 11304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001455260520789773,
+      "loss": 2.0469,
+      "step": 11305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001454494476399925,
+      "loss": 2.0247,
+      "step": 11306
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014537285993660899,
+      "loss": 2.0367,
+      "step": 11307
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014529628897244214,
+      "loss": 1.9398,
+      "step": 11308
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014521973475110618,
+      "loss": 2.1039,
+      "step": 11309
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014514319727621455,
+      "loss": 1.9932,
+      "step": 11310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014506667655138,
+      "loss": 1.8625,
+      "step": 11311
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001449901725802147,
+      "loss": 1.8373,
+      "step": 11312
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001449136853663296,
+      "loss": 2.1003,
+      "step": 11313
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014483721491333513,
+      "loss": 1.9928,
+      "step": 11314
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014476076122484077,
+      "loss": 2.0437,
+      "step": 11315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014468432430445556,
+      "loss": 2.0437,
+      "step": 11316
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014460790415578734,
+      "loss": 1.8625,
+      "step": 11317
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001445315007824433,
+      "loss": 1.8601,
+      "step": 11318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014445511418802976,
+      "loss": 2.101,
+      "step": 11319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014437874437615261,
+      "loss": 1.8441,
+      "step": 11320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014430239135041655,
+      "loss": 2.0047,
+      "step": 11321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014422605511442565,
+      "loss": 2.0563,
+      "step": 11322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014414973567178292,
+      "loss": 2.1593,
+      "step": 11323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001440734330260911,
+      "loss": 2.0474,
+      "step": 11324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014399714718095187,
+      "loss": 1.9731,
+      "step": 11325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001439208781399659,
+      "loss": 1.9481,
+      "step": 11326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001438446259067332,
+      "loss": 1.9694,
+      "step": 11327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014376839048485334,
+      "loss": 2.153,
+      "step": 11328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001436921718779247,
+      "loss": 2.1694,
+      "step": 11329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014361597008954498,
+      "loss": 2.1026,
+      "step": 11330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014353978512331083,
+      "loss": 1.9983,
+      "step": 11331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014346361698281874,
+      "loss": 2.1568,
+      "step": 11332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014338746567166377,
+      "loss": 2.0574,
+      "step": 11333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001433113311934406,
+      "loss": 1.9989,
+      "step": 11334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001432352135517427,
+      "loss": 2.0045,
+      "step": 11335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001431591127501633,
+      "loss": 1.8923,
+      "step": 11336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001430830287922944,
+      "loss": 2.1256,
+      "step": 11337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014300696168172733,
+      "loss": 1.9907,
+      "step": 11338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014293091142205246,
+      "loss": 2.0472,
+      "step": 11339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014285487801685986,
+      "loss": 1.9438,
+      "step": 11340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001427788614697384,
+      "loss": 2.0069,
+      "step": 11341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001427028617842761,
+      "loss": 1.9213,
+      "step": 11342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014262687896406028,
+      "loss": 1.8927,
+      "step": 11343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014255091301267775,
+      "loss": 2.1739,
+      "step": 11344
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014247496393371417,
+      "loss": 2.1268,
+      "step": 11345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014239903173075447,
+      "loss": 2.0936,
+      "step": 11346
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014232311640738265,
+      "loss": 1.9325,
+      "step": 11347
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001422472179671825,
+      "loss": 1.792,
+      "step": 11348
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001421713364137363,
+      "loss": 1.8918,
+      "step": 11349
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014209547175062594,
+      "loss": 2.1321,
+      "step": 11350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001420196239814322,
+      "loss": 2.1156,
+      "step": 11351
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014194379310973566,
+      "loss": 1.9611,
+      "step": 11352
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001418679791391154,
+      "loss": 2.0298,
+      "step": 11353
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001417921820731502,
+      "loss": 2.0239,
+      "step": 11354
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014171640191541762,
+      "loss": 2.0062,
+      "step": 11355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001416406386694949,
+      "loss": 2.0117,
+      "step": 11356
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001415648923389582,
+      "loss": 2.0676,
+      "step": 11357
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014148916292738278,
+      "loss": 2.0885,
+      "step": 11358
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014141345043834324,
+      "loss": 2.0462,
+      "step": 11359
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014133775487541362,
+      "loss": 2.0627,
+      "step": 11360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014126207624216676,
+      "loss": 1.7715,
+      "step": 11361
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001411864145421749,
+      "loss": 2.0115,
+      "step": 11362
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014111076977900926,
+      "loss": 1.9367,
+      "step": 11363
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014103514195624078,
+      "loss": 1.9898,
+      "step": 11364
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001409595310774391,
+      "loss": 2.117,
+      "step": 11365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001408839371461732,
+      "loss": 1.9716,
+      "step": 11366
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014080836016601116,
+      "loss": 2.0702,
+      "step": 11367
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014073280014052076,
+      "loss": 2.0058,
+      "step": 11368
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014065725707326833,
+      "loss": 1.9031,
+      "step": 11369
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014058173096781972,
+      "loss": 2.0284,
+      "step": 11370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014050622182773987,
+      "loss": 2.0217,
+      "step": 11371
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001404307296565932,
+      "loss": 2.0327,
+      "step": 11372
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014035525445794294,
+      "loss": 2.003,
+      "step": 11373
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014027979623535176,
+      "loss": 1.966,
+      "step": 11374
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014020435499238133,
+      "loss": 1.9694,
+      "step": 11375
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014012893073259285,
+      "loss": 1.976,
+      "step": 11376
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014005352345954642,
+      "loss": 2.0488,
+      "step": 11377
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013997813317680148,
+      "loss": 2.0918,
+      "step": 11378
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001399027598879164,
+      "loss": 2.0221,
+      "step": 11379
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013982740359644937,
+      "loss": 1.9324,
+      "step": 11380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001397520643059571,
+      "loss": 1.8784,
+      "step": 11381
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013967674201999586,
+      "loss": 1.9892,
+      "step": 11382
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001396014367421209,
+      "loss": 1.9838,
+      "step": 11383
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013952614847588708,
+      "loss": 2.0659,
+      "step": 11384
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000139450877224848,
+      "loss": 1.9652,
+      "step": 11385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001393756229925565,
+      "loss": 2.054,
+      "step": 11386
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001393003857825651,
+      "loss": 2.0034,
+      "step": 11387
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013922516559842496,
+      "loss": 1.9036,
+      "step": 11388
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013914996244368666,
+      "loss": 1.9276,
+      "step": 11389
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001390747763218998,
+      "loss": 1.9503,
+      "step": 11390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001389996072366137,
+      "loss": 2.1418,
+      "step": 11391
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001389244551913763,
+      "loss": 2.106,
+      "step": 11392
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000138849320189735,
+      "loss": 1.9601,
+      "step": 11393
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013877420223523614,
+      "loss": 2.0892,
+      "step": 11394
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013869910133142578,
+      "loss": 2.0073,
+      "step": 11395
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001386240174818487,
+      "loss": 1.9431,
+      "step": 11396
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001385489506900491,
+      "loss": 2.1152,
+      "step": 11397
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013847390095957,
+      "loss": 2.0047,
+      "step": 11398
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013839886829395442,
+      "loss": 2.1313,
+      "step": 11399
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013832385269674374,
+      "loss": 1.9986,
+      "step": 11400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013824885417147904,
+      "loss": 2.0509,
+      "step": 11401
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013817387272170006,
+      "loss": 1.9561,
+      "step": 11402
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013809890835094662,
+      "loss": 1.981,
+      "step": 11403
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013802396106275694,
+      "loss": 2.1067,
+      "step": 11404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013794903086066868,
+      "loss": 1.9581,
+      "step": 11405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013787411774821866,
+      "loss": 2.0146,
+      "step": 11406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013779922172894315,
+      "loss": 1.9483,
+      "step": 11407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013772434280637734,
+      "loss": 2.0314,
+      "step": 11408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013764948098405572,
+      "loss": 2.0063,
+      "step": 11409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001375746362655117,
+      "loss": 2.071,
+      "step": 11410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013749980865427846,
+      "loss": 1.9624,
+      "step": 11411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013742499815388786,
+      "loss": 2.1084,
+      "step": 11412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001373502047678712,
+      "loss": 2.0613,
+      "step": 11413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013727542849975867,
+      "loss": 1.9453,
+      "step": 11414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013720066935308018,
+      "loss": 1.9491,
+      "step": 11415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001371259273313644,
+      "loss": 2.0017,
+      "step": 11416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001370512024381394,
+      "loss": 1.9535,
+      "step": 11417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013697649467693217,
+      "loss": 2.0324,
+      "step": 11418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013690180405126934,
+      "loss": 2.0677,
+      "step": 11419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013682713056467638,
+      "loss": 1.9421,
+      "step": 11420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013675247422067805,
+      "loss": 1.9391,
+      "step": 11421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013667783502279812,
+      "loss": 2.068,
+      "step": 11422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013660321297456003,
+      "loss": 2.1411,
+      "step": 11423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013652860807948603,
+      "loss": 1.9587,
+      "step": 11424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013645402034109756,
+      "loss": 2.0773,
+      "step": 11425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013637944976291523,
+      "loss": 2.0371,
+      "step": 11426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013630489634845923,
+      "loss": 1.9935,
+      "step": 11427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013623036010124845,
+      "loss": 1.9111,
+      "step": 11428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001361558410248013,
+      "loss": 1.9411,
+      "step": 11429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013608133912263498,
+      "loss": 1.9251,
+      "step": 11430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013600685439826648,
+      "loss": 2.0339,
+      "step": 11431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001359323868552116,
+      "loss": 1.9493,
+      "step": 11432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013585793649698524,
+      "loss": 1.8442,
+      "step": 11433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013578350332710154,
+      "loss": 2.0406,
+      "step": 11434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013570908734907423,
+      "loss": 1.9029,
+      "step": 11435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001356346885664157,
+      "loss": 2.1245,
+      "step": 11436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013556030698263783,
+      "loss": 2.0906,
+      "step": 11437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013548594260125141,
+      "loss": 2.1262,
+      "step": 11438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013541159542576687,
+      "loss": 2.0753,
+      "step": 11439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001353372654596935,
+      "loss": 1.8564,
+      "step": 11440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013526295270653983,
+      "loss": 1.8991,
+      "step": 11441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013518865716981337,
+      "loss": 1.9363,
+      "step": 11442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013511437885302142,
+      "loss": 2.0077,
+      "step": 11443
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001350401177596699,
+      "loss": 1.9894,
+      "step": 11444
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001349658738932641,
+      "loss": 2.1524,
+      "step": 11445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013489164725730834,
+      "loss": 2.1787,
+      "step": 11446
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013481743785530664,
+      "loss": 2.0671,
+      "step": 11447
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013474324569076167,
+      "loss": 2.0753,
+      "step": 11448
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001346690707671755,
+      "loss": 2.0475,
+      "step": 11449
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001345949130880491,
+      "loss": 1.9498,
+      "step": 11450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013452077265688333,
+      "loss": 2.0866,
+      "step": 11451
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013444664947717756,
+      "loss": 2.1543,
+      "step": 11452
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001343725435524306,
+      "loss": 1.9778,
+      "step": 11453
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013429845488614024,
+      "loss": 2.0921,
+      "step": 11454
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013422438348180394,
+      "loss": 2.0404,
+      "step": 11455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013415032934291794,
+      "loss": 2.15,
+      "step": 11456
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001340762924729777,
+      "loss": 1.9983,
+      "step": 11457
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013400227287547784,
+      "loss": 2.1278,
+      "step": 11458
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013392827055391248,
+      "loss": 2.2038,
+      "step": 11459
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013385428551177463,
+      "loss": 1.9837,
+      "step": 11460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001337803177525565,
+      "loss": 1.815,
+      "step": 11461
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013370636727974943,
+      "loss": 2.0058,
+      "step": 11462
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013363243409684434,
+      "loss": 2.0352,
+      "step": 11463
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013355851820733084,
+      "loss": 2.0516,
+      "step": 11464
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013348461961469805,
+      "loss": 1.9577,
+      "step": 11465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013341073832243383,
+      "loss": 2.0302,
+      "step": 11466
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013333687433402604,
+      "loss": 2.0009,
+      "step": 11467
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013326302765296095,
+      "loss": 2.1092,
+      "step": 11468
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001331891982827243,
+      "loss": 2.0558,
+      "step": 11469
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013311538622680092,
+      "loss": 2.0898,
+      "step": 11470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013304159148867518,
+      "loss": 2.0006,
+      "step": 11471
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013296781407183019,
+      "loss": 1.9226,
+      "step": 11472
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001328940539797484,
+      "loss": 2.0333,
+      "step": 11473
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013282031121591138,
+      "loss": 1.9365,
+      "step": 11474
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013274658578380023,
+      "loss": 1.9776,
+      "step": 11475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013267287768689473,
+      "loss": 2.0203,
+      "step": 11476
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013259918692867402,
+      "loss": 1.9859,
+      "step": 11477
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013252551351261672,
+      "loss": 2.1199,
+      "step": 11478
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013245185744220029,
+      "loss": 2.0577,
+      "step": 11479
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001323782187209014,
+      "loss": 1.9243,
+      "step": 11480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013230459735219586,
+      "loss": 2.0757,
+      "step": 11481
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013223099333955907,
+      "loss": 1.8984,
+      "step": 11482
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001321574066864651,
+      "loss": 2.0622,
+      "step": 11483
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013208383739638746,
+      "loss": 1.9422,
+      "step": 11484
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013201028547279864,
+      "loss": 2.0144,
+      "step": 11485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001319367509191708,
+      "loss": 2.2468,
+      "step": 11486
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013186323373897468,
+      "loss": 2.1172,
+      "step": 11487
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013178973393568056,
+      "loss": 2.0803,
+      "step": 11488
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013171625151275757,
+      "loss": 1.9819,
+      "step": 11489
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001316427864736746,
+      "loss": 2.1543,
+      "step": 11490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013156933882189925,
+      "loss": 1.9044,
+      "step": 11491
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013149590856089834,
+      "loss": 2.0136,
+      "step": 11492
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013142249569413783,
+      "loss": 2.12,
+      "step": 11493
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013134910022508334,
+      "loss": 1.8855,
+      "step": 11494
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000131275722157199,
+      "loss": 2.1017,
+      "step": 11495
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001312023614939486,
+      "loss": 2.0219,
+      "step": 11496
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013112901823879468,
+      "loss": 1.967,
+      "step": 11497
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013105569239519954,
+      "loss": 1.9673,
+      "step": 11498
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013098238396662415,
+      "loss": 1.847,
+      "step": 11499
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013090909295652886,
+      "loss": 2.0146,
+      "step": 11500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013083581936837303,
+      "loss": 1.9885,
+      "step": 11501
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013076256320561564,
+      "loss": 2.0678,
+      "step": 11502
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013068932447171433,
+      "loss": 1.8957,
+      "step": 11503
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001306161031701263,
+      "loss": 2.1077,
+      "step": 11504
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001305428993043074,
+      "loss": 1.7884,
+      "step": 11505
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013046971287771354,
+      "loss": 2.13,
+      "step": 11506
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013039654389379897,
+      "loss": 2.08,
+      "step": 11507
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013032339235601747,
+      "loss": 1.9818,
+      "step": 11508
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001302502582678219,
+      "loss": 2.01,
+      "step": 11509
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001301771416326646,
+      "loss": 1.9592,
+      "step": 11510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001301040424539966,
+      "loss": 2.0368,
+      "step": 11511
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001300309607352685,
+      "loss": 2.1446,
+      "step": 11512
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012995789647992972,
+      "loss": 2.1031,
+      "step": 11513
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012988484969142934,
+      "loss": 2.001,
+      "step": 11514
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012981182037321516,
+      "loss": 2.0547,
+      "step": 11515
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012973880852873443,
+      "loss": 2.0104,
+      "step": 11516
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001296658141614332,
+      "loss": 2.0656,
+      "step": 11517
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001295928372747574,
+      "loss": 2.0267,
+      "step": 11518
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001295198778721513,
+      "loss": 1.9278,
+      "step": 11519
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012944693595705914,
+      "loss": 2.1318,
+      "step": 11520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012937401153292373,
+      "loss": 2.0629,
+      "step": 11521
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012930110460318716,
+      "loss": 1.9906,
+      "step": 11522
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012922821517129108,
+      "loss": 2.2476,
+      "step": 11523
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001291553432406759,
+      "loss": 2.1433,
+      "step": 11524
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001290824888147813,
+      "loss": 2.1297,
+      "step": 11525
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001290096518970461,
+      "loss": 1.9289,
+      "step": 11526
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012893683249090866,
+      "loss": 2.0184,
+      "step": 11527
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000128864030599806,
+      "loss": 2.2594,
+      "step": 11528
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012879124622717464,
+      "loss": 1.9066,
+      "step": 11529
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012871847937644994,
+      "loss": 2.0316,
+      "step": 11530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000128645730051067,
+      "loss": 2.0011,
+      "step": 11531
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012857299825445956,
+      "loss": 2.1683,
+      "step": 11532
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001285002839900608,
+      "loss": 2.0235,
+      "step": 11533
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012842758726130281,
+      "loss": 2.068,
+      "step": 11534
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012835490807161737,
+      "loss": 1.8748,
+      "step": 11535
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012828224642443486,
+      "loss": 1.9277,
+      "step": 11536
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012820960232318518,
+      "loss": 2.1371,
+      "step": 11537
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012813697577129712,
+      "loss": 1.9897,
+      "step": 11538
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012806436677219917,
+      "loss": 2.0514,
+      "step": 11539
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001279917753293184,
+      "loss": 1.8766,
+      "step": 11540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012791920144608133,
+      "loss": 2.0146,
+      "step": 11541
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012784664512591342,
+      "loss": 1.9464,
+      "step": 11542
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012777410637223991,
+      "loss": 2.1123,
+      "step": 11543
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001277015851884845,
+      "loss": 2.0385,
+      "step": 11544
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001276290815780705,
+      "loss": 2.0394,
+      "step": 11545
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012755659554442,
+      "loss": 1.899,
+      "step": 11546
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012748412709095487,
+      "loss": 2.0002,
+      "step": 11547
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012741167622109558,
+      "loss": 2.0794,
+      "step": 11548
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012733924293826198,
+      "loss": 2.0774,
+      "step": 11549
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012726682724587302,
+      "loss": 1.9115,
+      "step": 11550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012719442914734707,
+      "loss": 2.1305,
+      "step": 11551
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012712204864610144,
+      "loss": 1.9015,
+      "step": 11552
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012704968574555247,
+      "loss": 2.0969,
+      "step": 11553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012697734044911612,
+      "loss": 2.0026,
+      "step": 11554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001269050127602071,
+      "loss": 2.1053,
+      "step": 11555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001268327026822395,
+      "loss": 2.0327,
+      "step": 11556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001267604102186264,
+      "loss": 2.0348,
+      "step": 11557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012668813537278033,
+      "loss": 2.1136,
+      "step": 11558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012661587814811283,
+      "loss": 2.0114,
+      "step": 11559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012654363854803453,
+      "loss": 1.9792,
+      "step": 11560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012647141657595512,
+      "loss": 1.9031,
+      "step": 11561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012639921223528406,
+      "loss": 2.1539,
+      "step": 11562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012632702552942926,
+      "loss": 1.9999,
+      "step": 11563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012625485646179825,
+      "loss": 1.9647,
+      "step": 11564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012618270503579731,
+      "loss": 1.9431,
+      "step": 11565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012611057125483255,
+      "loss": 1.7696,
+      "step": 11566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001260384551223086,
+      "loss": 1.8973,
+      "step": 11567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001259663566416296,
+      "loss": 1.9354,
+      "step": 11568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012589427581619856,
+      "loss": 2.0536,
+      "step": 11569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012582221264941816,
+      "loss": 2.0763,
+      "step": 11570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012575016714468985,
+      "loss": 2.1193,
+      "step": 11571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001256781393054142,
+      "loss": 2.0122,
+      "step": 11572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012560612913499114,
+      "loss": 1.9736,
+      "step": 11573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001255341366368199,
+      "loss": 1.9855,
+      "step": 11574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012546216181429855,
+      "loss": 1.9565,
+      "step": 11575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012539020467082447,
+      "loss": 2.1608,
+      "step": 11576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012531826520979406,
+      "loss": 2.053,
+      "step": 11577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012524634343460335,
+      "loss": 2.015,
+      "step": 11578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012517443934864704,
+      "loss": 2.0247,
+      "step": 11579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012510255295531913,
+      "loss": 1.9967,
+      "step": 11580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012503068425801273,
+      "loss": 1.9286,
+      "step": 11581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012495883326012048,
+      "loss": 2.0265,
+      "step": 11582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012488699996503384,
+      "loss": 2.1558,
+      "step": 11583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012481518437614336,
+      "loss": 1.8958,
+      "step": 11584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012474338649683887,
+      "loss": 1.9949,
+      "step": 11585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012467160633050962,
+      "loss": 1.9894,
+      "step": 11586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012459984388054368,
+      "loss": 1.9421,
+      "step": 11587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012452809915032843,
+      "loss": 1.9839,
+      "step": 11588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012445637214325022,
+      "loss": 1.9742,
+      "step": 11589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000124384662862695,
+      "loss": 2.0235,
+      "step": 11590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012431297131204754,
+      "loss": 2.1631,
+      "step": 11591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012424129749469176,
+      "loss": 1.8118,
+      "step": 11592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012416964141401065,
+      "loss": 1.926,
+      "step": 11593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012409800307338698,
+      "loss": 1.9714,
+      "step": 11594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000124026382476202,
+      "loss": 2.0639,
+      "step": 11595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012395477962583634,
+      "loss": 2.1896,
+      "step": 11596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012388319452566978,
+      "loss": 2.0898,
+      "step": 11597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012381162717908152,
+      "loss": 2.0435,
+      "step": 11598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001237400775894495,
+      "loss": 2.0912,
+      "step": 11599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001236685457601512,
+      "loss": 1.8526,
+      "step": 11600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012359703169456272,
+      "loss": 2.0222,
+      "step": 11601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012352553539606015,
+      "loss": 2.021,
+      "step": 11602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012345405686801808,
+      "loss": 2.036,
+      "step": 11603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012338259611381047,
+      "loss": 1.856,
+      "step": 11604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012331115313681025,
+      "loss": 2.0356,
+      "step": 11605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012323972794039002,
+      "loss": 1.9052,
+      "step": 11606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012316832052792104,
+      "loss": 2.2212,
+      "step": 11607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001230969309027739,
+      "loss": 2.0136,
+      "step": 11608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001230255590683183,
+      "loss": 2.0621,
+      "step": 11609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001229542050279233,
+      "loss": 2.1236,
+      "step": 11610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012288286878495697,
+      "loss": 2.1359,
+      "step": 11611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012281155034278647,
+      "loss": 2.0089,
+      "step": 11612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012274024970477805,
+      "loss": 1.9174,
+      "step": 11613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012266896687429752,
+      "loss": 1.9235,
+      "step": 11614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012259770185470953,
+      "loss": 1.9326,
+      "step": 11615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001225264546493779,
+      "loss": 1.9871,
+      "step": 11616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012245522526166551,
+      "loss": 2.0982,
+      "step": 11617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001223840136949349,
+      "loss": 2.0906,
+      "step": 11618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012231281995254728,
+      "loss": 1.951,
+      "step": 11619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012224164403786304,
+      "loss": 2.0079,
+      "step": 11620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001221704859542418,
+      "loss": 1.8273,
+      "step": 11621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012209934570504266,
+      "loss": 2.1131,
+      "step": 11622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012202822329362345,
+      "loss": 2.0126,
+      "step": 11623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012195711872334131,
+      "loss": 2.0145,
+      "step": 11624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012188603199755239,
+      "loss": 2.0247,
+      "step": 11625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012181496311961243,
+      "loss": 2.1514,
+      "step": 11626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012174391209287595,
+      "loss": 1.9702,
+      "step": 11627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012167287892069662,
+      "loss": 1.9563,
+      "step": 11628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012160186360642733,
+      "loss": 1.9708,
+      "step": 11629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012153086615342045,
+      "loss": 2.0964,
+      "step": 11630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012145988656502699,
+      "loss": 2.1428,
+      "step": 11631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001213889248445974,
+      "loss": 2.047,
+      "step": 11632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012131798099548114,
+      "loss": 2.051,
+      "step": 11633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012124705502102712,
+      "loss": 2.0376,
+      "step": 11634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012117614692458317,
+      "loss": 1.8819,
+      "step": 11635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012110525670949618,
+      "loss": 2.021,
+      "step": 11636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001210343843791123,
+      "loss": 1.9311,
+      "step": 11637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012096352993677712,
+      "loss": 2.0462,
+      "step": 11638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012089269338583503,
+      "loss": 2.0231,
+      "step": 11639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012082187472962958,
+      "loss": 2.081,
+      "step": 11640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012075107397150348,
+      "loss": 2.0907,
+      "step": 11641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000120680291114799,
+      "loss": 1.9089,
+      "step": 11642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012060952616285708,
+      "loss": 2.0269,
+      "step": 11643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001205387791190179,
+      "loss": 2.1431,
+      "step": 11644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012046804998662103,
+      "loss": 1.9917,
+      "step": 11645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012039733876900511,
+      "loss": 2.1251,
+      "step": 11646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012032664546950766,
+      "loss": 2.1041,
+      "step": 11647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001202559700914656,
+      "loss": 1.9476,
+      "step": 11648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012018531263821514,
+      "loss": 2.0861,
+      "step": 11649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012011467311309144,
+      "loss": 2.0378,
+      "step": 11650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012004405151942872,
+      "loss": 2.1223,
+      "step": 11651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011997344786056042,
+      "loss": 1.9611,
+      "step": 11652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011990286213981943,
+      "loss": 1.9274,
+      "step": 11653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011983229436053744,
+      "loss": 1.9678,
+      "step": 11654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011976174452604549,
+      "loss": 1.9137,
+      "step": 11655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011969121263967342,
+      "loss": 1.8599,
+      "step": 11656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011962069870475085,
+      "loss": 1.9933,
+      "step": 11657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011955020272460604,
+      "loss": 1.9786,
+      "step": 11658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011947972470256658,
+      "loss": 1.9961,
+      "step": 11659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011940926464195906,
+      "loss": 2.1289,
+      "step": 11660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011933882254610961,
+      "loss": 2.0198,
+      "step": 11661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011926839841834314,
+      "loss": 2.158,
+      "step": 11662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011919799226198386,
+      "loss": 2.1008,
+      "step": 11663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011912760408035483,
+      "loss": 2.0965,
+      "step": 11664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011905723387677897,
+      "loss": 1.9418,
+      "step": 11665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011898688165457772,
+      "loss": 2.0962,
+      "step": 11666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011891654741707186,
+      "loss": 1.8708,
+      "step": 11667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011884623116758119,
+      "loss": 1.9095,
+      "step": 11668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011877593290942507,
+      "loss": 1.9613,
+      "step": 11669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001187056526459216,
+      "loss": 2.1054,
+      "step": 11670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011863539038038818,
+      "loss": 2.0772,
+      "step": 11671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011856514611614121,
+      "loss": 2.0153,
+      "step": 11672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011849491985649663,
+      "loss": 1.8473,
+      "step": 11673
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001184247116047692,
+      "loss": 2.0774,
+      "step": 11674
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011835452136427288,
+      "loss": 2.0636,
+      "step": 11675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001182843491383207,
+      "loss": 2.0521,
+      "step": 11676
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011821419493022518,
+      "loss": 1.8809,
+      "step": 11677
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011814405874329765,
+      "loss": 1.9399,
+      "step": 11678
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011807394058084863,
+      "loss": 1.9648,
+      "step": 11679
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011800384044618784,
+      "loss": 2.1792,
+      "step": 11680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011793375834262437,
+      "loss": 2.0546,
+      "step": 11681
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011786369427346616,
+      "loss": 2.0132,
+      "step": 11682
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011779364824202033,
+      "loss": 1.981,
+      "step": 11683
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011772362025159311,
+      "loss": 2.014,
+      "step": 11684
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011765361030549026,
+      "loss": 1.9349,
+      "step": 11685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011758361840701631,
+      "loss": 2.2904,
+      "step": 11686
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011751364455947499,
+      "loss": 1.9855,
+      "step": 11687
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011744368876616912,
+      "loss": 1.8466,
+      "step": 11688
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011737375103040099,
+      "loss": 2.0869,
+      "step": 11689
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001173038313554718,
+      "loss": 1.974,
+      "step": 11690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011723392974468178,
+      "loss": 1.9106,
+      "step": 11691
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011716404620133036,
+      "loss": 2.0103,
+      "step": 11692
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011709418072871653,
+      "loss": 1.8511,
+      "step": 11693
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011702433333013796,
+      "loss": 1.8724,
+      "step": 11694
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011695450400889151,
+      "loss": 2.0513,
+      "step": 11695
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011688469276827324,
+      "loss": 1.8762,
+      "step": 11696
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011681489961157865,
+      "loss": 2.0752,
+      "step": 11697
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011674512454210201,
+      "loss": 2.0749,
+      "step": 11698
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011667536756313685,
+      "loss": 2.2013,
+      "step": 11699
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001166056286779757,
+      "loss": 1.8962,
+      "step": 11700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011653590788991069,
+      "loss": 1.8507,
+      "step": 11701
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00011646620520223272,
+      "loss": 1.952,
+      "step": 11702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011639652061823186,
+      "loss": 2.0267,
+      "step": 11703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011632685414119721,
+      "loss": 2.01,
+      "step": 11704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001162572057744175,
+      "loss": 1.9773,
+      "step": 11705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001161875755211802,
+      "loss": 2.1592,
+      "step": 11706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011611796338477198,
+      "loss": 1.9777,
+      "step": 11707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011604836936847851,
+      "loss": 2.1346,
+      "step": 11708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011597879347558516,
+      "loss": 1.9931,
+      "step": 11709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011590923570937584,
+      "loss": 2.1032,
+      "step": 11710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011583969607313388,
+      "loss": 1.9887,
+      "step": 11711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011577017457014161,
+      "loss": 2.1103,
+      "step": 11712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011570067120368083,
+      "loss": 1.9422,
+      "step": 11713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011563118597703209,
+      "loss": 2.0824,
+      "step": 11714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011556171889347533,
+      "loss": 1.9935,
+      "step": 11715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011549226995628936,
+      "loss": 2.0004,
+      "step": 11716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011542283916875268,
+      "loss": 2.0394,
+      "step": 11717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011535342653414232,
+      "loss": 2.0059,
+      "step": 11718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011528403205573485,
+      "loss": 1.9347,
+      "step": 11719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011521465573680562,
+      "loss": 2.0868,
+      "step": 11720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011514529758062964,
+      "loss": 1.9278,
+      "step": 11721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011507595759048067,
+      "loss": 2.0664,
+      "step": 11722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011500663576963171,
+      "loss": 2.0734,
+      "step": 11723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011493733212135477,
+      "loss": 1.9502,
+      "step": 11724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011486804664892136,
+      "loss": 2.0832,
+      "step": 11725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001147987793556019,
+      "loss": 1.8177,
+      "step": 11726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011472953024466588,
+      "loss": 2.1464,
+      "step": 11727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011466029931938182,
+      "loss": 1.9663,
+      "step": 11728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011459108658301798,
+      "loss": 1.9978,
+      "step": 11729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011452189203884117,
+      "loss": 2.066,
+      "step": 11730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001144527156901175,
+      "loss": 2.1561,
+      "step": 11731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011438355754011215,
+      "loss": 2.129,
+      "step": 11732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011431441759208983,
+      "loss": 2.0224,
+      "step": 11733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011424529584931397,
+      "loss": 2.167,
+      "step": 11734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001141761923150471,
+      "loss": 1.9681,
+      "step": 11735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011410710699255133,
+      "loss": 2.0793,
+      "step": 11736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011403803988508759,
+      "loss": 2.027,
+      "step": 11737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001139689909959159,
+      "loss": 1.9354,
+      "step": 11738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011389996032829547,
+      "loss": 2.0511,
+      "step": 11739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011383094788548492,
+      "loss": 2.0407,
+      "step": 11740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011376195367074177,
+      "loss": 2.0985,
+      "step": 11741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011369297768732256,
+      "loss": 2.0729,
+      "step": 11742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001136240199384831,
+      "loss": 2.1507,
+      "step": 11743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011355508042747853,
+      "loss": 2.0277,
+      "step": 11744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011348615915756282,
+      "loss": 2.0394,
+      "step": 11745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011341725613198933,
+      "loss": 1.98,
+      "step": 11746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011334837135401022,
+      "loss": 2.0006,
+      "step": 11747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011327950482687727,
+      "loss": 2.0414,
+      "step": 11748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011321065655384099,
+      "loss": 1.9062,
+      "step": 11749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011314182653815126,
+      "loss": 1.8594,
+      "step": 11750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011307301478305681,
+      "loss": 1.8571,
+      "step": 11751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011300422129180598,
+      "loss": 2.0968,
+      "step": 11752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011293544606764594,
+      "loss": 1.9885,
+      "step": 11753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011286668911382292,
+      "loss": 1.904,
+      "step": 11754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011279795043358232,
+      "loss": 1.9195,
+      "step": 11755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011272923003016905,
+      "loss": 1.9709,
+      "step": 11756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011266052790682669,
+      "loss": 2.0424,
+      "step": 11757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011259184406679818,
+      "loss": 1.9054,
+      "step": 11758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011252317851332538,
+      "loss": 1.9218,
+      "step": 11759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011245453124964983,
+      "loss": 2.1132,
+      "step": 11760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011238590227901157,
+      "loss": 1.9531,
+      "step": 11761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001123172916046501,
+      "loss": 2.152,
+      "step": 11762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011224869922980391,
+      "loss": 2.069,
+      "step": 11763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011218012515771098,
+      "loss": 1.9145,
+      "step": 11764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011211156939160794,
+      "loss": 1.9866,
+      "step": 11765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011204303193473087,
+      "loss": 1.9381,
+      "step": 11766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011197451279031474,
+      "loss": 1.9775,
+      "step": 11767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011190601196159405,
+      "loss": 1.9678,
+      "step": 11768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001118375294518021,
+      "loss": 1.9736,
+      "step": 11769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011176906526417141,
+      "loss": 2.1653,
+      "step": 11770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011170061940193349,
+      "loss": 2.0046,
+      "step": 11771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011163219186831941,
+      "loss": 2.0555,
+      "step": 11772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011156378266655903,
+      "loss": 2.1488,
+      "step": 11773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011149539179988139,
+      "loss": 1.9684,
+      "step": 11774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011142701927151455,
+      "loss": 1.9904,
+      "step": 11775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011135866508468612,
+      "loss": 2.0246,
+      "step": 11776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011129032924262245,
+      "loss": 2.0018,
+      "step": 11777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011122201174854913,
+      "loss": 2.0399,
+      "step": 11778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011115371260569079,
+      "loss": 2.0475,
+      "step": 11779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001110854318172716,
+      "loss": 1.9626,
+      "step": 11780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011101716938651435,
+      "loss": 1.944,
+      "step": 11781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001109489253166413,
+      "loss": 2.0565,
+      "step": 11782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011088069961087349,
+      "loss": 1.9895,
+      "step": 11783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011081249227243162,
+      "loss": 1.9329,
+      "step": 11784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011074430330453517,
+      "loss": 1.9831,
+      "step": 11785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011067613271040272,
+      "loss": 1.8285,
+      "step": 11786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011060798049325199,
+      "loss": 2.0113,
+      "step": 11787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011053984665630024,
+      "loss": 1.9571,
+      "step": 11788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011047173120276333,
+      "loss": 2.2104,
+      "step": 11789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011040363413585646,
+      "loss": 2.005,
+      "step": 11790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011033555545879393,
+      "loss": 1.994,
+      "step": 11791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011026749517478935,
+      "loss": 1.9811,
+      "step": 11792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011019945328705533,
+      "loss": 2.083,
+      "step": 11793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00011013142979880347,
+      "loss": 1.9569,
+      "step": 11794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001100634247132446,
+      "loss": 2.0543,
+      "step": 11795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010999543803358891,
+      "loss": 2.0326,
+      "step": 11796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010992746976304546,
+      "loss": 2.0155,
+      "step": 11797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010985951990482246,
+      "loss": 2.0428,
+      "step": 11798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010979158846212716,
+      "loss": 1.9531,
+      "step": 11799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010972367543816641,
+      "loss": 2.069,
+      "step": 11800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010965578083614564,
+      "loss": 2.0859,
+      "step": 11801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010958790465926972,
+      "loss": 1.9629,
+      "step": 11802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010952004691074235,
+      "loss": 2.1295,
+      "step": 11803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010945220759376689,
+      "loss": 2.0689,
+      "step": 11804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010938438671154538,
+      "loss": 1.9468,
+      "step": 11805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010931658426727914,
+      "loss": 1.9177,
+      "step": 11806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010924880026416839,
+      "loss": 2.0364,
+      "step": 11807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010918103470541297,
+      "loss": 2.0251,
+      "step": 11808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010911328759421152,
+      "loss": 1.9767,
+      "step": 11809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010904555893376178,
+      "loss": 1.9879,
+      "step": 11810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010897784872726057,
+      "loss": 1.877,
+      "step": 11811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001089101569779043,
+      "loss": 1.9743,
+      "step": 11812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000108842483688888,
+      "loss": 2.0867,
+      "step": 11813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010877482886340601,
+      "loss": 1.9598,
+      "step": 11814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010870719250465161,
+      "loss": 2.03,
+      "step": 11815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010863957461581775,
+      "loss": 2.0718,
+      "step": 11816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010857197520009593,
+      "loss": 2.0099,
+      "step": 11817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010850439426067705,
+      "loss": 1.9776,
+      "step": 11818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010843683180075092,
+      "loss": 2.0581,
+      "step": 11819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010836928782350696,
+      "loss": 2.0399,
+      "step": 11820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010830176233213323,
+      "loss": 1.9255,
+      "step": 11821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010823425532981707,
+      "loss": 1.9129,
+      "step": 11822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010816676681974492,
+      "loss": 2.0536,
+      "step": 11823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010809929680510256,
+      "loss": 2.031,
+      "step": 11824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010803184528907461,
+      "loss": 2.0938,
+      "step": 11825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010796441227484482,
+      "loss": 1.9963,
+      "step": 11826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010789699776559653,
+      "loss": 1.9926,
+      "step": 11827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010782960176451162,
+      "loss": 1.9913,
+      "step": 11828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010776222427477134,
+      "loss": 2.0157,
+      "step": 11829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010769486529955597,
+      "loss": 1.9504,
+      "step": 11830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001076275248420453,
+      "loss": 1.9389,
+      "step": 11831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010756020290541774,
+      "loss": 1.9034,
+      "step": 11832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010749289949285113,
+      "loss": 1.9933,
+      "step": 11833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010742561460752215,
+      "loss": 1.9793,
+      "step": 11834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001073583482526071,
+      "loss": 1.9354,
+      "step": 11835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010729110043128093,
+      "loss": 1.897,
+      "step": 11836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010722387114671794,
+      "loss": 1.8839,
+      "step": 11837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010715666040209132,
+      "loss": 2.0493,
+      "step": 11838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010708946820057391,
+      "loss": 1.9453,
+      "step": 11839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010702229454533718,
+      "loss": 1.93,
+      "step": 11840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010695513943955182,
+      "loss": 2.1165,
+      "step": 11841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010688800288638767,
+      "loss": 2.0475,
+      "step": 11842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010682088488901392,
+      "loss": 1.9914,
+      "step": 11843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001067537854505986,
+      "loss": 2.0233,
+      "step": 11844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001066867045743089,
+      "loss": 1.9421,
+      "step": 11845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010661964226331117,
+      "loss": 1.909,
+      "step": 11846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010655259852077104,
+      "loss": 2.0254,
+      "step": 11847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010648557334985309,
+      "loss": 1.8737,
+      "step": 11848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010641856675372109,
+      "loss": 1.9908,
+      "step": 11849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010635157873553764,
+      "loss": 2.1645,
+      "step": 11850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00010628460929846506,
+      "loss": 1.913,
+      "step": 11851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010621765844566438,
+      "loss": 1.9748,
+      "step": 11852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010615072618029575,
+      "loss": 1.9561,
+      "step": 11853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010608381250551847,
+      "loss": 2.0892,
+      "step": 11854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010601691742449127,
+      "loss": 1.8565,
+      "step": 11855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010595004094037159,
+      "loss": 1.9458,
+      "step": 11856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001058831830563161,
+      "loss": 1.9177,
+      "step": 11857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010581634377548066,
+      "loss": 2.039,
+      "step": 11858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010574952310102038,
+      "loss": 2.0388,
+      "step": 11859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010568272103608928,
+      "loss": 2.1341,
+      "step": 11860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010561593758384053,
+      "loss": 2.1004,
+      "step": 11861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010554917274742631,
+      "loss": 1.9061,
+      "step": 11862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001054824265299984,
+      "loss": 1.9437,
+      "step": 11863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010541569893470726,
+      "loss": 1.9122,
+      "step": 11864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001053489899647025,
+      "loss": 2.1138,
+      "step": 11865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001052822996231329,
+      "loss": 2.0838,
+      "step": 11866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010521562791314654,
+      "loss": 2.168,
+      "step": 11867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010514897483789048,
+      "loss": 2.0075,
+      "step": 11868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010508234040051079,
+      "loss": 1.962,
+      "step": 11869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001050157246041527,
+      "loss": 2.0749,
+      "step": 11870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010494912745196089,
+      "loss": 1.8347,
+      "step": 11871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010488254894707877,
+      "loss": 2.0215,
+      "step": 11872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010481598909264895,
+      "loss": 2.043,
+      "step": 11873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010474944789181313,
+      "loss": 2.0085,
+      "step": 11874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001046829253477124,
+      "loss": 2.0568,
+      "step": 11875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010461642146348671,
+      "loss": 2.099,
+      "step": 11876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010454993624227521,
+      "loss": 1.9602,
+      "step": 11877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010448346968721596,
+      "loss": 2.1591,
+      "step": 11878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010441702180144663,
+      "loss": 1.9788,
+      "step": 11879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010435059258810359,
+      "loss": 1.9933,
+      "step": 11880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010428418205032247,
+      "loss": 1.93,
+      "step": 11881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010421779019123784,
+      "loss": 2.0386,
+      "step": 11882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010415141701398378,
+      "loss": 2.1523,
+      "step": 11883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001040850625216932,
+      "loss": 2.0127,
+      "step": 11884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010401872671749813,
+      "loss": 1.9677,
+      "step": 11885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010395240960452961,
+      "loss": 2.0942,
+      "step": 11886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010388611118591834,
+      "loss": 2.0109,
+      "step": 11887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010381983146479352,
+      "loss": 1.9698,
+      "step": 11888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010375357044428374,
+      "loss": 1.9611,
+      "step": 11889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010368732812751652,
+      "loss": 2.1131,
+      "step": 11890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010362110451761897,
+      "loss": 1.9392,
+      "step": 11891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001035548996177168,
+      "loss": 1.914,
+      "step": 11892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010348871343093508,
+      "loss": 1.9475,
+      "step": 11893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001034225459603978,
+      "loss": 2.0183,
+      "step": 11894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010335639720922846,
+      "loss": 2.0993,
+      "step": 11895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010329026718054934,
+      "loss": 2.1368,
+      "step": 11896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001032241558774819,
+      "loss": 2.0162,
+      "step": 11897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010315806330314664,
+      "loss": 2.1279,
+      "step": 11898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001030919894606635,
+      "loss": 2.0276,
+      "step": 11899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010302593435315127,
+      "loss": 2.0673,
+      "step": 11900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001029598979837278,
+      "loss": 2.0247,
+      "step": 11901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010289388035551012,
+      "loss": 1.983,
+      "step": 11902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001028278814716146,
+      "loss": 1.9925,
+      "step": 11903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010276190133515646,
+      "loss": 2.0402,
+      "step": 11904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010269593994925003,
+      "loss": 2.1336,
+      "step": 11905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010262999731700884,
+      "loss": 2.2427,
+      "step": 11906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010256407344154572,
+      "loss": 1.8977,
+      "step": 11907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010249816832597231,
+      "loss": 2.1166,
+      "step": 11908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010243228197339943,
+      "loss": 2.0256,
+      "step": 11909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010236641438693705,
+      "loss": 1.9992,
+      "step": 11910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010230056556969446,
+      "loss": 1.9268,
+      "step": 11911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010223473552477974,
+      "loss": 2.0524,
+      "step": 11912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010216892425530023,
+      "loss": 2.0374,
+      "step": 11913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010210313176436226,
+      "loss": 1.9394,
+      "step": 11914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010203735805507158,
+      "loss": 1.9753,
+      "step": 11915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010197160313053282,
+      "loss": 2.0911,
+      "step": 11916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010190586699384963,
+      "loss": 1.9883,
+      "step": 11917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001018401496481251,
+      "loss": 2.1571,
+      "step": 11918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010177445109646116,
+      "loss": 2.1381,
+      "step": 11919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010170877134195889,
+      "loss": 1.9581,
+      "step": 11920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010164311038771845,
+      "loss": 1.9617,
+      "step": 11921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010157746823683939,
+      "loss": 1.8938,
+      "step": 11922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001015118448924201,
+      "loss": 1.9917,
+      "step": 11923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010144624035755812,
+      "loss": 1.8687,
+      "step": 11924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010138065463535,
+      "loss": 1.9795,
+      "step": 11925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001013150877288918,
+      "loss": 2.0566,
+      "step": 11926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010124953964127831,
+      "loss": 2.0021,
+      "step": 11927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010118401037560355,
+      "loss": 2.1843,
+      "step": 11928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010111849993496047,
+      "loss": 1.9452,
+      "step": 11929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010105300832244169,
+      "loss": 2.1777,
+      "step": 11930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010098753554113832,
+      "loss": 1.948,
+      "step": 11931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001009220815941409,
+      "loss": 1.988,
+      "step": 11932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010085664648453885,
+      "loss": 1.9664,
+      "step": 11933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010079123021542109,
+      "loss": 2.0147,
+      "step": 11934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010072583278987534,
+      "loss": 2.0957,
+      "step": 11935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010066045421098846,
+      "loss": 2.059,
+      "step": 11936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010059509448184633,
+      "loss": 1.954,
+      "step": 11937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010052975360553446,
+      "loss": 1.8628,
+      "step": 11938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010046443158513679,
+      "loss": 1.9297,
+      "step": 11939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010039912842373677,
+      "loss": 1.9957,
+      "step": 11940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001003338441244167,
+      "loss": 2.1555,
+      "step": 11941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010026857869025846,
+      "loss": 1.9336,
+      "step": 11942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010020333212434252,
+      "loss": 1.9788,
+      "step": 11943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010013810442974874,
+      "loss": 2.0308,
+      "step": 11944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010007289560955585,
+      "loss": 2.2502,
+      "step": 11945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00010000770566684209,
+      "loss": 1.874,
+      "step": 11946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.99425346046845e-05,
+      "loss": 2.0383,
+      "step": 11947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.987738242615924e-05,
+      "loss": 2.0559,
+      "step": 11948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.981224913434156e-05,
+      "loss": 1.9549,
+      "step": 11949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.974713473230618e-05,
+      "loss": 1.9995,
+      "step": 11950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.968203922312647e-05,
+      "loss": 2.0581,
+      "step": 11951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.961696260987513e-05,
+      "loss": 2.0939,
+      "step": 11952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.955190489562377e-05,
+      "loss": 2.0081,
+      "step": 11953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.948686608344348e-05,
+      "loss": 2.0225,
+      "step": 11954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.942184617640421e-05,
+      "loss": 2.1747,
+      "step": 11955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.935684517757498e-05,
+      "loss": 1.8974,
+      "step": 11956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.929186309002391e-05,
+      "loss": 1.9416,
+      "step": 11957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.922689991681849e-05,
+      "loss": 2.0121,
+      "step": 11958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.916195566102504e-05,
+      "loss": 1.8784,
+      "step": 11959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.90970303257091e-05,
+      "loss": 1.9686,
+      "step": 11960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903212391393513e-05,
+      "loss": 1.9089,
+      "step": 11961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.896723642876715e-05,
+      "loss": 2.1126,
+      "step": 11962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.890236787326784e-05,
+      "loss": 1.9827,
+      "step": 11963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.883751825049919e-05,
+      "loss": 2.0235,
+      "step": 11964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.877268756352204e-05,
+      "loss": 1.9001,
+      "step": 11965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.870787581539692e-05,
+      "loss": 1.8293,
+      "step": 11966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.864308300918284e-05,
+      "loss": 1.8991,
+      "step": 11967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.857830914793825e-05,
+      "loss": 1.9076,
+      "step": 11968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.851355423472052e-05,
+      "loss": 1.9977,
+      "step": 11969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.84488182725864e-05,
+      "loss": 2.0618,
+      "step": 11970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.838410126459141e-05,
+      "loss": 1.9499,
+      "step": 11971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.831940321379051e-05,
+      "loss": 2.0992,
+      "step": 11972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.82547241232375e-05,
+      "loss": 2.0982,
+      "step": 11973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.819006399598529e-05,
+      "loss": 2.1453,
+      "step": 11974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.812542283508624e-05,
+      "loss": 2.0121,
+      "step": 11975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.80608006435914e-05,
+      "loss": 2.0307,
+      "step": 11976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.799619742455107e-05,
+      "loss": 2.0094,
+      "step": 11977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.793161318101463e-05,
+      "loss": 1.8836,
+      "step": 11978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.786704791603074e-05,
+      "loss": 2.0574,
+      "step": 11979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.780250163264698e-05,
+      "loss": 2.0048,
+      "step": 11980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.773797433391008e-05,
+      "loss": 2.0005,
+      "step": 11981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.767346602286575e-05,
+      "loss": 2.0186,
+      "step": 11982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.760897670255914e-05,
+      "loss": 1.9121,
+      "step": 11983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.754450637603418e-05,
+      "loss": 1.9154,
+      "step": 11984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.748005504633406e-05,
+      "loss": 1.9057,
+      "step": 11985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.74156227165009e-05,
+      "loss": 2.1288,
+      "step": 11986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.735120938957626e-05,
+      "loss": 2.0128,
+      "step": 11987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.728681506860049e-05,
+      "loss": 1.8846,
+      "step": 11988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.722243975661322e-05,
+      "loss": 1.9725,
+      "step": 11989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.715808345665284e-05,
+      "loss": 2.0176,
+      "step": 11990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.709374617175748e-05,
+      "loss": 2.073,
+      "step": 11991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.702942790496389e-05,
+      "loss": 2.0353,
+      "step": 11992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.696512865930785e-05,
+      "loss": 1.9932,
+      "step": 11993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.69008484378247e-05,
+      "loss": 1.9777,
+      "step": 11994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.683658724354849e-05,
+      "loss": 2.0283,
+      "step": 11995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.677234507951255e-05,
+      "loss": 1.9398,
+      "step": 11996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.6708121948749e-05,
+      "loss": 2.0468,
+      "step": 11997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.664391785428978e-05,
+      "loss": 1.9864,
+      "step": 11998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.657973279916515e-05,
+      "loss": 2.0413,
+      "step": 11999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.651556678640489e-05,
+      "loss": 2.0775,
+      "step": 12000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.64514198190376e-05,
+      "loss": 1.9993,
+      "step": 12001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.638729190009148e-05,
+      "loss": 2.0221,
+      "step": 12002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.632318303259336e-05,
+      "loss": 1.9758,
+      "step": 12003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.625909321956933e-05,
+      "loss": 1.9438,
+      "step": 12004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.619502246404443e-05,
+      "loss": 2.0459,
+      "step": 12005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61309707690432e-05,
+      "loss": 1.9549,
+      "step": 12006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.606693813758899e-05,
+      "loss": 2.043,
+      "step": 12007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.600292457270416e-05,
+      "loss": 1.9709,
+      "step": 12008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.593893007741022e-05,
+      "loss": 2.0267,
+      "step": 12009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.587495465472812e-05,
+      "loss": 1.9376,
+      "step": 12010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.581099830767754e-05,
+      "loss": 2.0275,
+      "step": 12011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.574706103927727e-05,
+      "loss": 2.0032,
+      "step": 12012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.568314285254532e-05,
+      "loss": 2.1419,
+      "step": 12013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.56192437504989e-05,
+      "loss": 1.9145,
+      "step": 12014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.555536373615414e-05,
+      "loss": 1.9448,
+      "step": 12015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.549150281252633e-05,
+      "loss": 2.2197,
+      "step": 12016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.542766098262967e-05,
+      "loss": 2.0185,
+      "step": 12017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.536383824947792e-05,
+      "loss": 1.9142,
+      "step": 12018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.530003461608356e-05,
+      "loss": 1.8855,
+      "step": 12019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.523625008545822e-05,
+      "loss": 2.1178,
+      "step": 12020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.517248466061256e-05,
+      "loss": 2.0679,
+      "step": 12021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.510873834455674e-05,
+      "loss": 1.9297,
+      "step": 12022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.504501114029956e-05,
+      "loss": 1.9805,
+      "step": 12023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.498130305084918e-05,
+      "loss": 1.9663,
+      "step": 12024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.491761407921251e-05,
+      "loss": 1.9828,
+      "step": 12025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.485394422839616e-05,
+      "loss": 2.0259,
+      "step": 12026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.47902935014054e-05,
+      "loss": 2.1026,
+      "step": 12027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.472666190124458e-05,
+      "loss": 1.9978,
+      "step": 12028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.466304943091719e-05,
+      "loss": 1.8807,
+      "step": 12029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.459945609342612e-05,
+      "loss": 2.0441,
+      "step": 12030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.453588189177308e-05,
+      "loss": 1.9141,
+      "step": 12031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.44723268289588e-05,
+      "loss": 2.0491,
+      "step": 12032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.440879090798316e-05,
+      "loss": 2.0156,
+      "step": 12033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.434527413184546e-05,
+      "loss": 2.0417,
+      "step": 12034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.428177650354364e-05,
+      "loss": 1.9149,
+      "step": 12035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.421829802607501e-05,
+      "loss": 1.8636,
+      "step": 12036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.41548387024358e-05,
+      "loss": 2.0954,
+      "step": 12037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.40913985356216e-05,
+      "loss": 2.01,
+      "step": 12038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.402797752862685e-05,
+      "loss": 2.0058,
+      "step": 12039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.396457568444517e-05,
+      "loss": 1.9634,
+      "step": 12040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.390119300606914e-05,
+      "loss": 1.9933,
+      "step": 12041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.383782949649079e-05,
+      "loss": 1.976,
+      "step": 12042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.377448515870096e-05,
+      "loss": 2.0481,
+      "step": 12043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.371115999568957e-05,
+      "loss": 2.0758,
+      "step": 12044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.364785401044567e-05,
+      "loss": 2.0805,
+      "step": 12045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.35845672059577e-05,
+      "loss": 1.9902,
+      "step": 12046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.352129958521277e-05,
+      "loss": 2.1683,
+      "step": 12047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.345805115119721e-05,
+      "loss": 1.9464,
+      "step": 12048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.339482190689647e-05,
+      "loss": 2.0114,
+      "step": 12049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.333161185529532e-05,
+      "loss": 1.9126,
+      "step": 12050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.326842099937726e-05,
+      "loss": 1.9562,
+      "step": 12051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.32052493421251e-05,
+      "loss": 1.833,
+      "step": 12052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.314209688652053e-05,
+      "loss": 2.0452,
+      "step": 12053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.30789636355448e-05,
+      "loss": 2.1124,
+      "step": 12054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.301584959217774e-05,
+      "loss": 1.9758,
+      "step": 12055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.295275475939847e-05,
+      "loss": 1.851,
+      "step": 12056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.288967914018514e-05,
+      "loss": 1.9363,
+      "step": 12057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.282662273751536e-05,
+      "loss": 1.9551,
+      "step": 12058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.276358555436531e-05,
+      "loss": 2.0064,
+      "step": 12059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.270056759371053e-05,
+      "loss": 2.0623,
+      "step": 12060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.263756885852548e-05,
+      "loss": 2.0117,
+      "step": 12061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.257458935178414e-05,
+      "loss": 1.7889,
+      "step": 12062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.25116290764591e-05,
+      "loss": 2.0172,
+      "step": 12063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.244868803552225e-05,
+      "loss": 2.0655,
+      "step": 12064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.238576623194445e-05,
+      "loss": 1.9424,
+      "step": 12065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.232286366869602e-05,
+      "loss": 1.9546,
+      "step": 12066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.22599803487459e-05,
+      "loss": 2.0374,
+      "step": 12067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.219711627506238e-05,
+      "loss": 1.9098,
+      "step": 12068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21342714506127e-05,
+      "loss": 1.8992,
+      "step": 12069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.207144587836347e-05,
+      "loss": 1.9847,
+      "step": 12070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.200863956128008e-05,
+      "loss": 2.0597,
+      "step": 12071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.194585250232718e-05,
+      "loss": 2.1482,
+      "step": 12072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.188308470446827e-05,
+      "loss": 2.0238,
+      "step": 12073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.182033617066643e-05,
+      "loss": 2.0641,
+      "step": 12074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.17576069038834e-05,
+      "loss": 2.074,
+      "step": 12075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.169489690708016e-05,
+      "loss": 1.9267,
+      "step": 12076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.163220618321667e-05,
+      "loss": 2.0053,
+      "step": 12077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.15695347352522e-05,
+      "loss": 1.9645,
+      "step": 12078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.150688256614503e-05,
+      "loss": 2.1527,
+      "step": 12079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.144424967885234e-05,
+      "loss": 2.1539,
+      "step": 12080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.138163607633055e-05,
+      "loss": 1.9927,
+      "step": 12081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.131904176153527e-05,
+      "loss": 1.8413,
+      "step": 12082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.12564667374211e-05,
+      "loss": 2.0875,
+      "step": 12083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.119391100694158e-05,
+      "loss": 2.0003,
+      "step": 12084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.113137457304965e-05,
+      "loss": 2.1904,
+      "step": 12085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.106885743869714e-05,
+      "loss": 1.9546,
+      "step": 12086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.100635960683495e-05,
+      "loss": 1.9551,
+      "step": 12087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.094388108041302e-05,
+      "loss": 1.9965,
+      "step": 12088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.088142186238068e-05,
+      "loss": 2.0174,
+      "step": 12089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.081898195568611e-05,
+      "loss": 2.1473,
+      "step": 12090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.075656136327659e-05,
+      "loss": 1.9594,
+      "step": 12091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.069416008809833e-05,
+      "loss": 2.0223,
+      "step": 12092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.063177813309715e-05,
+      "loss": 2.0497,
+      "step": 12093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.056941550121745e-05,
+      "loss": 2.1236,
+      "step": 12094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.050707219540283e-05,
+      "loss": 2.0535,
+      "step": 12095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.044474821859605e-05,
+      "loss": 1.9944,
+      "step": 12096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.038244357373909e-05,
+      "loss": 2.1225,
+      "step": 12097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.032015826377282e-05,
+      "loss": 2.0717,
+      "step": 12098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.025789229163717e-05,
+      "loss": 2.1156,
+      "step": 12099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.019564566027116e-05,
+      "loss": 2.0846,
+      "step": 12100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.013341837261319e-05,
+      "loss": 1.8283,
+      "step": 12101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.007121043160049e-05,
+      "loss": 1.985,
+      "step": 12102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.000902184016929e-05,
+      "loss": 2.0728,
+      "step": 12103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.994685260125501e-05,
+      "loss": 1.9823,
+      "step": 12104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.988470271779243e-05,
+      "loss": 2.2343,
+      "step": 12105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.9822572192715e-05,
+      "loss": 1.9908,
+      "step": 12106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.97604610289554e-05,
+      "loss": 1.8932,
+      "step": 12107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.969836922944535e-05,
+      "loss": 1.9578,
+      "step": 12108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.963629679711598e-05,
+      "loss": 2.0829,
+      "step": 12109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.957424373489709e-05,
+      "loss": 1.9309,
+      "step": 12110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.951221004571775e-05,
+      "loss": 2.1322,
+      "step": 12111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.945019573250596e-05,
+      "loss": 2.1035,
+      "step": 12112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.938820079818916e-05,
+      "loss": 1.9964,
+      "step": 12113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.932622524569356e-05,
+      "loss": 1.913,
+      "step": 12114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.926426907794455e-05,
+      "loss": 2.0146,
+      "step": 12115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.92023322978665e-05,
+      "loss": 1.9584,
+      "step": 12116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.914041490838315e-05,
+      "loss": 2.0661,
+      "step": 12117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.907851691241709e-05,
+      "loss": 1.9898,
+      "step": 12118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.901663831288997e-05,
+      "loss": 1.9357,
+      "step": 12119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.89547791127226e-05,
+      "loss": 1.9712,
+      "step": 12120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.8892939314835e-05,
+      "loss": 2.1315,
+      "step": 12121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.883111892214607e-05,
+      "loss": 1.9001,
+      "step": 12122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.876931793757387e-05,
+      "loss": 2.0387,
+      "step": 12123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.870753636403545e-05,
+      "loss": 1.939,
+      "step": 12124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.86457742044473e-05,
+      "loss": 2.0517,
+      "step": 12125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.858403146172456e-05,
+      "loss": 2.0572,
+      "step": 12126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.85223081387817e-05,
+      "loss": 2.1065,
+      "step": 12127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.846060423853197e-05,
+      "loss": 2.0221,
+      "step": 12128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.83989197638883e-05,
+      "loss": 2.0619,
+      "step": 12129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.833725471776216e-05,
+      "loss": 1.8979,
+      "step": 12130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.827560910306426e-05,
+      "loss": 2.1073,
+      "step": 12131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.821398292270433e-05,
+      "loss": 1.9423,
+      "step": 12132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.815237617959154e-05,
+      "loss": 2.1136,
+      "step": 12133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.809078887663363e-05,
+      "loss": 1.9883,
+      "step": 12134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.802922101673782e-05,
+      "loss": 1.9705,
+      "step": 12135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.796767260281002e-05,
+      "loss": 2.0076,
+      "step": 12136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.790614363775574e-05,
+      "loss": 1.9482,
+      "step": 12137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.784463412447913e-05,
+      "loss": 1.9913,
+      "step": 12138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.778314406588366e-05,
+      "loss": 2.0007,
+      "step": 12139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.77216734648716e-05,
+      "loss": 2.059,
+      "step": 12140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.766022232434479e-05,
+      "loss": 2.1827,
+      "step": 12141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.759879064720372e-05,
+      "loss": 1.9587,
+      "step": 12142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.753737843634818e-05,
+      "loss": 2.1762,
+      "step": 12143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.747598569467669e-05,
+      "loss": 1.9857,
+      "step": 12144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.741461242508753e-05,
+      "loss": 2.1232,
+      "step": 12145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.735325863047744e-05,
+      "loss": 2.012,
+      "step": 12146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.729192431374256e-05,
+      "loss": 2.0677,
+      "step": 12147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.723060947777778e-05,
+      "loss": 1.9492,
+      "step": 12148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.716931412547757e-05,
+      "loss": 1.8686,
+      "step": 12149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.710803825973512e-05,
+      "loss": 2.0248,
+      "step": 12150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.704678188344278e-05,
+      "loss": 2.0608,
+      "step": 12151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.69855449994919e-05,
+      "loss": 2.0369,
+      "step": 12152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.692432761077318e-05,
+      "loss": 1.738,
+      "step": 12153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.686312972017618e-05,
+      "loss": 2.1595,
+      "step": 12154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.680195133058948e-05,
+      "loss": 2.0805,
+      "step": 12155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.67407924449008e-05,
+      "loss": 2.0816,
+      "step": 12156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.667965306599718e-05,
+      "loss": 1.992,
+      "step": 12157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.66185331967645e-05,
+      "loss": 2.2428,
+      "step": 12158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.655743284008766e-05,
+      "loss": 2.0322,
+      "step": 12159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.649635199885065e-05,
+      "loss": 2.0842,
+      "step": 12160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.643529067593686e-05,
+      "loss": 1.9202,
+      "step": 12161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.63742488742284e-05,
+      "loss": 2.0748,
+      "step": 12162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.631322659660662e-05,
+      "loss": 2.136,
+      "step": 12163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.625222384595177e-05,
+      "loss": 2.1683,
+      "step": 12164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.619124062514355e-05,
+      "loss": 2.0799,
+      "step": 12165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.613027693706038e-05,
+      "loss": 1.9418,
+      "step": 12166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.606933278457995e-05,
+      "loss": 2.067,
+      "step": 12167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.600840817057875e-05,
+      "loss": 1.9766,
+      "step": 12168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.594750309793287e-05,
+      "loss": 2.1523,
+      "step": 12169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.588661756951705e-05,
+      "loss": 1.8803,
+      "step": 12170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.582575158820516e-05,
+      "loss": 2.1215,
+      "step": 12171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.57649051568702e-05,
+      "loss": 2.1022,
+      "step": 12172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.570407827838439e-05,
+      "loss": 1.8935,
+      "step": 12173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.564327095561886e-05,
+      "loss": 2.0004,
+      "step": 12174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.558248319144368e-05,
+      "loss": 2.065,
+      "step": 12175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.552171498872846e-05,
+      "loss": 2.1087,
+      "step": 12176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.546096635034145e-05,
+      "loss": 2.1273,
+      "step": 12177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.540023727915013e-05,
+      "loss": 2.1303,
+      "step": 12178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53395277780209e-05,
+      "loss": 1.9205,
+      "step": 12179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527883784981971e-05,
+      "loss": 1.972,
+      "step": 12180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.521816749741107e-05,
+      "loss": 2.0649,
+      "step": 12181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.515751672365884e-05,
+      "loss": 1.909,
+      "step": 12182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.509688553142564e-05,
+      "loss": 1.9931,
+      "step": 12183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.503627392357371e-05,
+      "loss": 2.0462,
+      "step": 12184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.497568190296395e-05,
+      "loss": 2.0708,
+      "step": 12185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.491510947245646e-05,
+      "loss": 2.0314,
+      "step": 12186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.485455663491021e-05,
+      "loss": 1.9658,
+      "step": 12187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.47940233931837e-05,
+      "loss": 2.0309,
+      "step": 12188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.47335097501341e-05,
+      "loss": 1.9927,
+      "step": 12189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.467301570861784e-05,
+      "loss": 2.0367,
+      "step": 12190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.461254127149026e-05,
+      "loss": 1.9624,
+      "step": 12191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.455208644160606e-05,
+      "loss": 1.9732,
+      "step": 12192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.449165122181885e-05,
+      "loss": 1.9818,
+      "step": 12193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.443123561498117e-05,
+      "loss": 1.9742,
+      "step": 12194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.43708396239447e-05,
+      "loss": 1.9715,
+      "step": 12195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.43104632515606e-05,
+      "loss": 2.1706,
+      "step": 12196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.425010650067854e-05,
+      "loss": 2.0446,
+      "step": 12197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.418976937414758e-05,
+      "loss": 1.9299,
+      "step": 12198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.412945187481558e-05,
+      "loss": 2.1927,
+      "step": 12199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.406915400552995e-05,
+      "loss": 1.9521,
+      "step": 12200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.400887576913673e-05,
+      "loss": 2.0818,
+      "step": 12201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.394861716848123e-05,
+      "loss": 1.9331,
+      "step": 12202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.388837820640766e-05,
+      "loss": 1.9402,
+      "step": 12203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.382815888575968e-05,
+      "loss": 2.0827,
+      "step": 12204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.376795920937969e-05,
+      "loss": 2.1172,
+      "step": 12205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.370777918010919e-05,
+      "loss": 1.9941,
+      "step": 12206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.364761880078875e-05,
+      "loss": 2.0335,
+      "step": 12207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.358747807425827e-05,
+      "loss": 2.0167,
+      "step": 12208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.352735700335646e-05,
+      "loss": 1.9764,
+      "step": 12209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.346725559092117e-05,
+      "loss": 1.869,
+      "step": 12210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.340717383978913e-05,
+      "loss": 2.1205,
+      "step": 12211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.334711175279663e-05,
+      "loss": 2.0165,
+      "step": 12212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.328706933277869e-05,
+      "loss": 1.9384,
+      "step": 12213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.322704658256935e-05,
+      "loss": 2.0056,
+      "step": 12214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.316704350500171e-05,
+      "loss": 1.9792,
+      "step": 12215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.31070601029083e-05,
+      "loss": 2.099,
+      "step": 12216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.304709637912044e-05,
+      "loss": 1.7561,
+      "step": 12217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.298715233646841e-05,
+      "loss": 2.0697,
+      "step": 12218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.292722797778174e-05,
+      "loss": 2.0846,
+      "step": 12219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.286732330588909e-05,
+      "loss": 1.9104,
+      "step": 12220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.280743832361809e-05,
+      "loss": 2.0949,
+      "step": 12221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.274757303379538e-05,
+      "loss": 1.963,
+      "step": 12222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.268772743924669e-05,
+      "loss": 1.9669,
+      "step": 12223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.262790154279703e-05,
+      "loss": 2.0353,
+      "step": 12224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.256809534727023e-05,
+      "loss": 2.0265,
+      "step": 12225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.250830885548932e-05,
+      "loss": 2.0003,
+      "step": 12226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.244854207027619e-05,
+      "loss": 2.1126,
+      "step": 12227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.238879499445224e-05,
+      "loss": 2.0726,
+      "step": 12228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.232906763083747e-05,
+      "loss": 1.8922,
+      "step": 12229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.226935998225126e-05,
+      "loss": 2.0349,
+      "step": 12230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.220967205151181e-05,
+      "loss": 2.0394,
+      "step": 12231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.215000384143672e-05,
+      "loss": 2.0514,
+      "step": 12232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.209035535484239e-05,
+      "loss": 1.8893,
+      "step": 12233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.20307265945443e-05,
+      "loss": 1.9215,
+      "step": 12234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.197111756335702e-05,
+      "loss": 1.8225,
+      "step": 12235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.191152826409442e-05,
+      "loss": 1.9383,
+      "step": 12236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.185195869956913e-05,
+      "loss": 2.0041,
+      "step": 12237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.179240887259304e-05,
+      "loss": 2.0813,
+      "step": 12238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.173287878597685e-05,
+      "loss": 1.9023,
+      "step": 12239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.167336844253076e-05,
+      "loss": 2.0405,
+      "step": 12240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.161387784506375e-05,
+      "loss": 2.0661,
+      "step": 12241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.155440699638383e-05,
+      "loss": 1.9459,
+      "step": 12242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.149495589929806e-05,
+      "loss": 2.0003,
+      "step": 12243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.14355245566129e-05,
+      "loss": 2.099,
+      "step": 12244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.137611297113356e-05,
+      "loss": 2.0085,
+      "step": 12245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.131672114566435e-05,
+      "loss": 2.0477,
+      "step": 12246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.125734908300869e-05,
+      "loss": 1.9236,
+      "step": 12247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.119799678596917e-05,
+      "loss": 1.9269,
+      "step": 12248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.113866425734734e-05,
+      "loss": 1.9061,
+      "step": 12249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.107935149994383e-05,
+      "loss": 2.1507,
+      "step": 12250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.10200585165582e-05,
+      "loss": 1.881,
+      "step": 12251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.096078530998934e-05,
+      "loss": 2.0092,
+      "step": 12252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.090153188303518e-05,
+      "loss": 1.9611,
+      "step": 12253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.084229823849242e-05,
+      "loss": 1.9781,
+      "step": 12254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.078308437915705e-05,
+      "loss": 2.0729,
+      "step": 12255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.072389030782428e-05,
+      "loss": 2.1946,
+      "step": 12256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.066471602728804e-05,
+      "loss": 1.9867,
+      "step": 12257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.06055615403416e-05,
+      "loss": 2.0906,
+      "step": 12258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.054642684977692e-05,
+      "loss": 1.9703,
+      "step": 12259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.048731195838571e-05,
+      "loss": 1.9277,
+      "step": 12260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.042821686895808e-05,
+      "loss": 1.9714,
+      "step": 12261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.036914158428343e-05,
+      "loss": 2.0837,
+      "step": 12262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.03100861071503e-05,
+      "loss": 2.0107,
+      "step": 12263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.02510504403463e-05,
+      "loss": 2.0182,
+      "step": 12264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.019203458665802e-05,
+      "loss": 1.9644,
+      "step": 12265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.013303854887099e-05,
+      "loss": 1.8805,
+      "step": 12266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.007406232977027e-05,
+      "loss": 1.9453,
+      "step": 12267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.001510593213946e-05,
+      "loss": 2.0288,
+      "step": 12268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.995616935876149e-05,
+      "loss": 2.1048,
+      "step": 12269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.989725261241815e-05,
+      "loss": 2.2106,
+      "step": 12270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.983835569589077e-05,
+      "loss": 1.888,
+      "step": 12271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.977947861195917e-05,
+      "loss": 2.0959,
+      "step": 12272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.972062136340257e-05,
+      "loss": 2.0045,
+      "step": 12273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.9661783952999e-05,
+      "loss": 2.0568,
+      "step": 12274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.960296638352605e-05,
+      "loss": 2.0078,
+      "step": 12275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.954416865775982e-05,
+      "loss": 1.9341,
+      "step": 12276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.948539077847572e-05,
+      "loss": 1.8631,
+      "step": 12277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.942663274844814e-05,
+      "loss": 1.9398,
+      "step": 12278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.93678945704508e-05,
+      "loss": 1.9544,
+      "step": 12279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.930917624725608e-05,
+      "loss": 2.0141,
+      "step": 12280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.92504777816358e-05,
+      "loss": 1.9545,
+      "step": 12281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.919179917636038e-05,
+      "loss": 2.0124,
+      "step": 12282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.913314043419983e-05,
+      "loss": 2.0448,
+      "step": 12283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.907450155792301e-05,
+      "loss": 1.9777,
+      "step": 12284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.901588255029763e-05,
+      "loss": 2.0103,
+      "step": 12285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.895728341409064e-05,
+      "loss": 2.1296,
+      "step": 12286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.889870415206824e-05,
+      "loss": 1.8859,
+      "step": 12287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.884014476699541e-05,
+      "loss": 1.9623,
+      "step": 12288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.878160526163625e-05,
+      "loss": 2.1093,
+      "step": 12289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.872308563875386e-05,
+      "loss": 1.9486,
+      "step": 12290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.866458590111081e-05,
+      "loss": 1.9769,
+      "step": 12291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.860610605146817e-05,
+      "loss": 2.0225,
+      "step": 12292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.85476460925864e-05,
+      "loss": 2.1189,
+      "step": 12293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.848920602722486e-05,
+      "loss": 1.8602,
+      "step": 12294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.843078585814223e-05,
+      "loss": 2.0421,
+      "step": 12295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.837238558809601e-05,
+      "loss": 2.026,
+      "step": 12296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.831400521984272e-05,
+      "loss": 1.9352,
+      "step": 12297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.825564475613806e-05,
+      "loss": 1.9921,
+      "step": 12298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.819730419973697e-05,
+      "loss": 2.019,
+      "step": 12299
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.813898355339311e-05,
+      "loss": 2.2782,
+      "step": 12300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.808068281985935e-05,
+      "loss": 1.7655,
+      "step": 12301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.802240200188754e-05,
+      "loss": 1.9464,
+      "step": 12302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.796414110222887e-05,
+      "loss": 1.9345,
+      "step": 12303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.79059001236333e-05,
+      "loss": 2.0092,
+      "step": 12304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.784767906884993e-05,
+      "loss": 1.982,
+      "step": 12305
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.77894779406268e-05,
+      "loss": 2.0144,
+      "step": 12306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.773129674171137e-05,
+      "loss": 2.0862,
+      "step": 12307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.767313547484978e-05,
+      "loss": 2.1371,
+      "step": 12308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.761499414278744e-05,
+      "loss": 2.0268,
+      "step": 12309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.755687274826862e-05,
+      "loss": 2.0979,
+      "step": 12310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.749877129403699e-05,
+      "loss": 2.028,
+      "step": 12311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.744068978283497e-05,
+      "loss": 1.9345,
+      "step": 12312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.738262821740411e-05,
+      "loss": 2.0156,
+      "step": 12313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.732458660048496e-05,
+      "loss": 1.9534,
+      "step": 12314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.726656493481754e-05,
+      "loss": 1.7825,
+      "step": 12315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.720856322314035e-05,
+      "loss": 2.0921,
+      "step": 12316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.715058146819126e-05,
+      "loss": 2.1658,
+      "step": 12317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.709261967270697e-05,
+      "loss": 2.0481,
+      "step": 12318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.703467783942375e-05,
+      "loss": 1.8884,
+      "step": 12319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.697675597107644e-05,
+      "loss": 2.1075,
+      "step": 12320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.691885407039906e-05,
+      "loss": 1.9786,
+      "step": 12321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.686097214012455e-05,
+      "loss": 1.998,
+      "step": 12322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.680311018298542e-05,
+      "loss": 1.9283,
+      "step": 12323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.674526820171263e-05,
+      "loss": 2.0249,
+      "step": 12324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.668744619903656e-05,
+      "loss": 1.8811,
+      "step": 12325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.662964417768641e-05,
+      "loss": 1.8908,
+      "step": 12326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.657186214039074e-05,
+      "loss": 1.9665,
+      "step": 12327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.651410008987697e-05,
+      "loss": 1.9657,
+      "step": 12328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.645635802887153e-05,
+      "loss": 2.0259,
+      "step": 12329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.639863596009989e-05,
+      "loss": 2.1119,
+      "step": 12330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.634093388628694e-05,
+      "loss": 1.9617,
+      "step": 12331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.62832518101561e-05,
+      "loss": 1.9589,
+      "step": 12332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.622558973443027e-05,
+      "loss": 1.9405,
+      "step": 12333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.6167947661831e-05,
+      "loss": 2.0988,
+      "step": 12334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.611032559507936e-05,
+      "loss": 1.9974,
+      "step": 12335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.605272353689519e-05,
+      "loss": 2.0041,
+      "step": 12336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.599514148999742e-05,
+      "loss": 1.9791,
+      "step": 12337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.593757945710388e-05,
+      "loss": 1.992,
+      "step": 12338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.588003744093197e-05,
+      "loss": 1.9567,
+      "step": 12339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.582251544419754e-05,
+      "loss": 2.0779,
+      "step": 12340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.576501346961596e-05,
+      "loss": 1.9691,
+      "step": 12341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.570753151990117e-05,
+      "loss": 1.9704,
+      "step": 12342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.565006959776677e-05,
+      "loss": 1.9343,
+      "step": 12343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.55926277059249e-05,
+      "loss": 1.8974,
+      "step": 12344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.553520584708706e-05,
+      "loss": 1.9972,
+      "step": 12345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.547780402396348e-05,
+      "loss": 1.9698,
+      "step": 12346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.542042223926393e-05,
+      "loss": 1.9753,
+      "step": 12347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.536306049569686e-05,
+      "loss": 2.0141,
+      "step": 12348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.530571879596987e-05,
+      "loss": 2.0451,
+      "step": 12349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.524839714278947e-05,
+      "loss": 2.1153,
+      "step": 12350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.51910955388616e-05,
+      "loss": 1.9836,
+      "step": 12351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.513381398689101e-05,
+      "loss": 1.8781,
+      "step": 12352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.507655248958145e-05,
+      "loss": 1.9887,
+      "step": 12353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.501931104963567e-05,
+      "loss": 2.1159,
+      "step": 12354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.496208966975581e-05,
+      "loss": 1.9924,
+      "step": 12355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.490488835264275e-05,
+      "loss": 2.1208,
+      "step": 12356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.484770710099647e-05,
+      "loss": 1.9941,
+      "step": 12357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.479054591751622e-05,
+      "loss": 2.0851,
+      "step": 12358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.47334048049001e-05,
+      "loss": 2.1613,
+      "step": 12359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.467628376584517e-05,
+      "loss": 2.1066,
+      "step": 12360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.46191828030477e-05,
+      "loss": 2.0056,
+      "step": 12361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.45621019192031e-05,
+      "loss": 2.0524,
+      "step": 12362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.450504111700568e-05,
+      "loss": 2.1375,
+      "step": 12363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.444800039914879e-05,
+      "loss": 2.0353,
+      "step": 12364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.439097976832482e-05,
+      "loss": 2.1044,
+      "step": 12365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.433397922722552e-05,
+      "loss": 2.1596,
+      "step": 12366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.427699877854127e-05,
+      "loss": 2.0249,
+      "step": 12367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.422003842496166e-05,
+      "loss": 2.1362,
+      "step": 12368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.416309816917532e-05,
+      "loss": 2.0866,
+      "step": 12369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.410617801387015e-05,
+      "loss": 2.0239,
+      "step": 12370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.404927796173277e-05,
+      "loss": 2.2037,
+      "step": 12371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.399239801544899e-05,
+      "loss": 1.9652,
+      "step": 12372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.393553817770366e-05,
+      "loss": 2.1027,
+      "step": 12373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.387869845118078e-05,
+      "loss": 2.1132,
+      "step": 12374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.382187883856333e-05,
+      "loss": 1.993,
+      "step": 12375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.376507934253323e-05,
+      "loss": 2.0954,
+      "step": 12376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.37082999657715e-05,
+      "loss": 2.0168,
+      "step": 12377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.365154071095842e-05,
+      "loss": 2.0286,
+      "step": 12378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.359480158077314e-05,
+      "loss": 2.024,
+      "step": 12379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.353808257789374e-05,
+      "loss": 1.9074,
+      "step": 12380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.348138370499751e-05,
+      "loss": 1.9256,
+      "step": 12381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.342470496476089e-05,
+      "loss": 2.0474,
+      "step": 12382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.336804635985922e-05,
+      "loss": 2.0422,
+      "step": 12383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.331140789296687e-05,
+      "loss": 2.0041,
+      "step": 12384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.325478956675719e-05,
+      "loss": 1.9468,
+      "step": 12385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.319819138390293e-05,
+      "loss": 1.9383,
+      "step": 12386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.314161334707559e-05,
+      "loss": 1.8814,
+      "step": 12387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.308505545894567e-05,
+      "loss": 1.9793,
+      "step": 12388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.302851772218277e-05,
+      "loss": 1.9385,
+      "step": 12389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.297200013945588e-05,
+      "loss": 1.9549,
+      "step": 12390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.291550271343261e-05,
+      "loss": 2.0823,
+      "step": 12391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.285902544677975e-05,
+      "loss": 2.0922,
+      "step": 12392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.280256834216308e-05,
+      "loss": 1.9352,
+      "step": 12393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.274613140224767e-05,
+      "loss": 1.9339,
+      "step": 12394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.268971462969748e-05,
+      "loss": 1.9363,
+      "step": 12395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.263331802717538e-05,
+      "loss": 1.9973,
+      "step": 12396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.257694159734335e-05,
+      "loss": 2.0322,
+      "step": 12397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.252058534286277e-05,
+      "loss": 1.978,
+      "step": 12398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.246424926639355e-05,
+      "loss": 2.0163,
+      "step": 12399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.240793337059503e-05,
+      "loss": 2.0469,
+      "step": 12400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.23516376581252e-05,
+      "loss": 2.1167,
+      "step": 12401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.229536213164167e-05,
+      "loss": 2.159,
+      "step": 12402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.223910679380063e-05,
+      "loss": 1.9283,
+      "step": 12403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.218287164725746e-05,
+      "loss": 1.9681,
+      "step": 12404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.212665669466651e-05,
+      "loss": 1.9026,
+      "step": 12405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.207046193868139e-05,
+      "loss": 1.8014,
+      "step": 12406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.20142873819546e-05,
+      "loss": 1.9522,
+      "step": 12407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.195813302713761e-05,
+      "loss": 1.9661,
+      "step": 12408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.190199887688103e-05,
+      "loss": 1.9502,
+      "step": 12409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.184588493383471e-05,
+      "loss": 2.0019,
+      "step": 12410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.178979120064721e-05,
+      "loss": 2.1436,
+      "step": 12411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.173371767996633e-05,
+      "loss": 2.0344,
+      "step": 12412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.167766437443874e-05,
+      "loss": 1.9583,
+      "step": 12413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.162163128671051e-05,
+      "loss": 2.0927,
+      "step": 12414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.156561841942638e-05,
+      "loss": 1.8219,
+      "step": 12415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.150962577523035e-05,
+      "loss": 2.1172,
+      "step": 12416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.145365335676524e-05,
+      "loss": 2.0558,
+      "step": 12417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.139770116667333e-05,
+      "loss": 2.0687,
+      "step": 12418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.134176920759555e-05,
+      "loss": 2.0062,
+      "step": 12419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.128585748217209e-05,
+      "loss": 1.9759,
+      "step": 12420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.122996599304188e-05,
+      "loss": 1.9301,
+      "step": 12421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.117409474284347e-05,
+      "loss": 2.1246,
+      "step": 12422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.111824373421383e-05,
+      "loss": 1.8927,
+      "step": 12423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.106241296978943e-05,
+      "loss": 2.0142,
+      "step": 12424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.100660245220559e-05,
+      "loss": 2.1091,
+      "step": 12425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.095081218409655e-05,
+      "loss": 2.065,
+      "step": 12426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0895042168096e-05,
+      "loss": 2.0451,
+      "step": 12427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.083929240683617e-05,
+      "loss": 2.0689,
+      "step": 12428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.078356290294874e-05,
+      "loss": 1.9574,
+      "step": 12429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.07278536590641e-05,
+      "loss": 1.9253,
+      "step": 12430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.067216467781201e-05,
+      "loss": 1.8081,
+      "step": 12431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.061649596182107e-05,
+      "loss": 2.0249,
+      "step": 12432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0560847513719e-05,
+      "loss": 1.9566,
+      "step": 12433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.050521933613235e-05,
+      "loss": 2.0732,
+      "step": 12434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.044961143168716e-05,
+      "loss": 1.9958,
+      "step": 12435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.039402380300814e-05,
+      "loss": 2.006,
+      "step": 12436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.033845645271902e-05,
+      "loss": 2.0638,
+      "step": 12437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.028290938344295e-05,
+      "loss": 2.0037,
+      "step": 12438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.022738259780182e-05,
+      "loss": 2.0562,
+      "step": 12439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.017187609841653e-05,
+      "loss": 2.0893,
+      "step": 12440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.011638988790703e-05,
+      "loss": 1.982,
+      "step": 12441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.006092396889264e-05,
+      "loss": 2.0538,
+      "step": 12442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.000547834399135e-05,
+      "loss": 2.1189,
+      "step": 12443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.995005301582036e-05,
+      "loss": 2.0416,
+      "step": 12444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.989464798699569e-05,
+      "loss": 1.9759,
+      "step": 12445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.983926326013284e-05,
+      "loss": 1.9715,
+      "step": 12446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.978389883784602e-05,
+      "loss": 1.926,
+      "step": 12447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.972855472274852e-05,
+      "loss": 1.9294,
+      "step": 12448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.967323091745265e-05,
+      "loss": 2.0124,
+      "step": 12449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.961792742456996e-05,
+      "loss": 2.0759,
+      "step": 12450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.956264424671088e-05,
+      "loss": 2.0621,
+      "step": 12451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.950738138648482e-05,
+      "loss": 1.9569,
+      "step": 12452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.94521388465002e-05,
+      "loss": 1.9803,
+      "step": 12453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.939691662936492e-05,
+      "loss": 2.0665,
+      "step": 12454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.934171473768542e-05,
+      "loss": 1.9304,
+      "step": 12455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.92865331740673e-05,
+      "loss": 1.846,
+      "step": 12456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.923137194111522e-05,
+      "loss": 2.0515,
+      "step": 12457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.917623104143317e-05,
+      "loss": 2.0424,
+      "step": 12458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.912111047762376e-05,
+      "loss": 2.197,
+      "step": 12459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.906601025228881e-05,
+      "loss": 1.9505,
+      "step": 12460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.901093036802903e-05,
+      "loss": 2.0633,
+      "step": 12461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.895587082744464e-05,
+      "loss": 1.9731,
+      "step": 12462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.890083163313438e-05,
+      "loss": 1.9117,
+      "step": 12463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.884581278769625e-05,
+      "loss": 2.0248,
+      "step": 12464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.879081429372713e-05,
+      "loss": 1.9911,
+      "step": 12465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.87358361538234e-05,
+      "loss": 2.0359,
+      "step": 12466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.86808783705799e-05,
+      "loss": 2.0245,
+      "step": 12467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.862594094659085e-05,
+      "loss": 2.0361,
+      "step": 12468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.85710238844493e-05,
+      "loss": 1.8778,
+      "step": 12469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.851612718674771e-05,
+      "loss": 1.9885,
+      "step": 12470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.846125085607713e-05,
+      "loss": 2.0504,
+      "step": 12471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.840639489502798e-05,
+      "loss": 2.0026,
+      "step": 12472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.835155930618936e-05,
+      "loss": 2.0539,
+      "step": 12473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.829674409214993e-05,
+      "loss": 2.2669,
+      "step": 12474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.824194925549693e-05,
+      "loss": 1.955,
+      "step": 12475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.818717479881681e-05,
+      "loss": 1.9154,
+      "step": 12476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8132420724695e-05,
+      "loss": 2.1422,
+      "step": 12477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.807768703571615e-05,
+      "loss": 1.9061,
+      "step": 12478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.802297373446375e-05,
+      "loss": 1.9599,
+      "step": 12479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.796828082352047e-05,
+      "loss": 2.0884,
+      "step": 12480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.791360830546767e-05,
+      "loss": 2.1154,
+      "step": 12481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78589561828864e-05,
+      "loss": 2.0011,
+      "step": 12482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78043244583561e-05,
+      "loss": 2.147,
+      "step": 12483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.774971313445565e-05,
+      "loss": 1.9943,
+      "step": 12484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.769512221376268e-05,
+      "loss": 2.0676,
+      "step": 12485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.764055169885419e-05,
+      "loss": 1.9246,
+      "step": 12486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.758600159230593e-05,
+      "loss": 1.9933,
+      "step": 12487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.753147189669278e-05,
+      "loss": 2.1656,
+      "step": 12488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.747696261458863e-05,
+      "loss": 2.0961,
+      "step": 12489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.742247374856664e-05,
+      "loss": 1.9026,
+      "step": 12490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.736800530119863e-05,
+      "loss": 1.9076,
+      "step": 12491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.731355727505568e-05,
+      "loss": 1.9136,
+      "step": 12492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.725912967270775e-05,
+      "loss": 1.9858,
+      "step": 12493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.720472249672421e-05,
+      "loss": 2.0047,
+      "step": 12494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.7150335749673e-05,
+      "loss": 1.9826,
+      "step": 12495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.709596943412138e-05,
+      "loss": 1.8972,
+      "step": 12496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.70416235526355e-05,
+      "loss": 1.9858,
+      "step": 12497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.698729810778065e-05,
+      "loss": 2.0808,
+      "step": 12498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.69329931021212e-05,
+      "loss": 2.0334,
+      "step": 12499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.687870853822032e-05,
+      "loss": 2.056,
+      "step": 12500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.682444441864038e-05,
+      "loss": 2.1186,
+      "step": 12501
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.677020074594293e-05,
+      "loss": 2.1743,
+      "step": 12502
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.671597752268827e-05,
+      "loss": 1.8885,
+      "step": 12503
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.666177475143586e-05,
+      "loss": 1.9838,
+      "step": 12504
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.660759243474412e-05,
+      "loss": 2.0598,
+      "step": 12505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.655343057517083e-05,
+      "loss": 2.0234,
+      "step": 12506
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.649928917527237e-05,
+      "loss": 1.9306,
+      "step": 12507
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.644516823760438e-05,
+      "loss": 2.1269,
+      "step": 12508
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.639106776472137e-05,
+      "loss": 2.077,
+      "step": 12509
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.633698775917724e-05,
+      "loss": 1.9727,
+      "step": 12510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.628292822352455e-05,
+      "loss": 2.038,
+      "step": 12511
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.622888916031505e-05,
+      "loss": 2.1247,
+      "step": 12512
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.617487057209942e-05,
+      "loss": 2.0522,
+      "step": 12513
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.612087246142767e-05,
+      "loss": 2.0109,
+      "step": 12514
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.606689483084854e-05,
+      "loss": 2.0676,
+      "step": 12515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.601293768290984e-05,
+      "loss": 1.9206,
+      "step": 12516
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.595900102015839e-05,
+      "loss": 2.0091,
+      "step": 12517
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.590508484514041e-05,
+      "loss": 2.0069,
+      "step": 12518
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.585118916040067e-05,
+      "loss": 1.8782,
+      "step": 12519
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.57973139684832e-05,
+      "loss": 1.9966,
+      "step": 12520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.574345927193092e-05,
+      "loss": 2.0075,
+      "step": 12521
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.568962507328613e-05,
+      "loss": 2.2041,
+      "step": 12522
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.563581137508979e-05,
+      "loss": 1.9729,
+      "step": 12523
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.558201817988197e-05,
+      "loss": 1.9071,
+      "step": 12524
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.552824549020198e-05,
+      "loss": 2.1263,
+      "step": 12525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.547449330858795e-05,
+      "loss": 1.902,
+      "step": 12526
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.542076163757715e-05,
+      "loss": 1.9991,
+      "step": 12527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.536705047970565e-05,
+      "loss": 2.0961,
+      "step": 12528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.531335983750897e-05,
+      "loss": 2.0917,
+      "step": 12529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.525968971352137e-05,
+      "loss": 2.1849,
+      "step": 12530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.52060401102762e-05,
+      "loss": 2.0036,
+      "step": 12531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.515241103030567e-05,
+      "loss": 1.9469,
+      "step": 12532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.509880247614148e-05,
+      "loss": 1.9716,
+      "step": 12533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.504521445031392e-05,
+      "loss": 1.9414,
+      "step": 12534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.499164695535248e-05,
+      "loss": 2.0623,
+      "step": 12535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.493809999378558e-05,
+      "loss": 1.989,
+      "step": 12536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.488457356814103e-05,
+      "loss": 2.0092,
+      "step": 12537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.483106768094516e-05,
+      "loss": 2.0231,
+      "step": 12538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.477758233472369e-05,
+      "loss": 2.0739,
+      "step": 12539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.472411753200108e-05,
+      "loss": 2.1864,
+      "step": 12540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.467067327530118e-05,
+      "loss": 1.8878,
+      "step": 12541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.461724956714665e-05,
+      "loss": 2.1106,
+      "step": 12542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.456384641005919e-05,
+      "loss": 2.0671,
+      "step": 12543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.451046380655939e-05,
+      "loss": 2.2194,
+      "step": 12544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.445710175916736e-05,
+      "loss": 1.9287,
+      "step": 12545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.440376027040167e-05,
+      "loss": 2.1055,
+      "step": 12546
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.435043934278029e-05,
+      "loss": 2.0446,
+      "step": 12547
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.429713897881984e-05,
+      "loss": 2.1539,
+      "step": 12548
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.424385918103654e-05,
+      "loss": 2.2203,
+      "step": 12549
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.41905999519452e-05,
+      "loss": 2.0455,
+      "step": 12550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.413736129405973e-05,
+      "loss": 1.8463,
+      "step": 12551
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.408414320989309e-05,
+      "loss": 2.0506,
+      "step": 12552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.403094570195744e-05,
+      "loss": 2.0123,
+      "step": 12553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.397776877276374e-05,
+      "loss": 1.8259,
+      "step": 12554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.3924612424822e-05,
+      "loss": 2.011,
+      "step": 12555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.387147666064136e-05,
+      "loss": 2.0039,
+      "step": 12556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.381836148273007e-05,
+      "loss": 2.0852,
+      "step": 12557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.376526689359518e-05,
+      "loss": 2.1112,
+      "step": 12558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.371219289574287e-05,
+      "loss": 1.9408,
+      "step": 12559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.365913949167828e-05,
+      "loss": 1.9693,
+      "step": 12560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.360610668390588e-05,
+      "loss": 1.8166,
+      "step": 12561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.355309447492874e-05,
+      "loss": 2.0845,
+      "step": 12562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.350010286724928e-05,
+      "loss": 2.1157,
+      "step": 12563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.34471318633687e-05,
+      "loss": 2.0837,
+      "step": 12564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.339418146578746e-05,
+      "loss": 2.0371,
+      "step": 12565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.334125167700494e-05,
+      "loss": 2.1292,
+      "step": 12566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.328834249951948e-05,
+      "loss": 1.9664,
+      "step": 12567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.323545393582847e-05,
+      "loss": 2.0327,
+      "step": 12568
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.318258598842858e-05,
+      "loss": 1.9649,
+      "step": 12569
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.312973865981514e-05,
+      "loss": 2.0693,
+      "step": 12570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.307691195248272e-05,
+      "loss": 2.0549,
+      "step": 12571
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.302410586892465e-05,
+      "loss": 2.0839,
+      "step": 12572
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.297132041163389e-05,
+      "loss": 1.9753,
+      "step": 12573
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.291855558310178e-05,
+      "loss": 2.05,
+      "step": 12574
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.286581138581898e-05,
+      "loss": 2.0126,
+      "step": 12575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.281308782227507e-05,
+      "loss": 1.9131,
+      "step": 12576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.276038489495888e-05,
+      "loss": 2.0324,
+      "step": 12577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.270770260635806e-05,
+      "loss": 1.9994,
+      "step": 12578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.26550409589593e-05,
+      "loss": 2.1276,
+      "step": 12579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.260239995524825e-05,
+      "loss": 2.0036,
+      "step": 12580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.254977959770992e-05,
+      "loss": 1.9917,
+      "step": 12581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.249717988882791e-05,
+      "loss": 1.9482,
+      "step": 12582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.244460083108516e-05,
+      "loss": 2.0059,
+      "step": 12583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.239204242696344e-05,
+      "loss": 1.9553,
+      "step": 12584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.233950467894372e-05,
+      "loss": 2.1686,
+      "step": 12585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.228698758950585e-05,
+      "loss": 1.9045,
+      "step": 12586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.223449116112878e-05,
+      "loss": 2.0195,
+      "step": 12587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.218201539629032e-05,
+      "loss": 2.0003,
+      "step": 12588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.21295602974677e-05,
+      "loss": 2.1172,
+      "step": 12589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.207712586713682e-05,
+      "loss": 1.9666,
+      "step": 12590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.202471210777261e-05,
+      "loss": 2.0133,
+      "step": 12591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.197231902184914e-05,
+      "loss": 2.1474,
+      "step": 12592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.191994661183959e-05,
+      "loss": 2.064,
+      "step": 12593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.186759488021603e-05,
+      "loss": 1.8847,
+      "step": 12594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.181526382944957e-05,
+      "loss": 1.8409,
+      "step": 12595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.176295346201021e-05,
+      "loss": 1.8704,
+      "step": 12596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.171066378036733e-05,
+      "loss": 2.0274,
+      "step": 12597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.16583947869891e-05,
+      "loss": 2.1103,
+      "step": 12598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.160614648434265e-05,
+      "loss": 1.9381,
+      "step": 12599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.155391887489414e-05,
+      "loss": 1.9245,
+      "step": 12600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.150171196110904e-05,
+      "loss": 1.8758,
+      "step": 12601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.144952574545154e-05,
+      "loss": 1.9479,
+      "step": 12602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.139736023038495e-05,
+      "loss": 1.9661,
+      "step": 12603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.134521541837151e-05,
+      "loss": 1.9669,
+      "step": 12604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.129309131187272e-05,
+      "loss": 1.9757,
+      "step": 12605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.124098791334898e-05,
+      "loss": 1.9635,
+      "step": 12606
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.118890522525955e-05,
+      "loss": 2.0375,
+      "step": 12607
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.113684325006285e-05,
+      "loss": 1.9888,
+      "step": 12608
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.108480199021649e-05,
+      "loss": 1.9471,
+      "step": 12609
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.103278144817681e-05,
+      "loss": 1.8938,
+      "step": 12610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.098078162639931e-05,
+      "loss": 1.9776,
+      "step": 12611
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.092880252733846e-05,
+      "loss": 2.088,
+      "step": 12612
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.087684415344796e-05,
+      "loss": 2.0212,
+      "step": 12613
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.08249065071802e-05,
+      "loss": 2.0358,
+      "step": 12614
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.077298959098676e-05,
+      "loss": 2.1086,
+      "step": 12615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.072109340731835e-05,
+      "loss": 2.0664,
+      "step": 12616
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.066921795862457e-05,
+      "loss": 1.8284,
+      "step": 12617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.061736324735401e-05,
+      "loss": 2.0439,
+      "step": 12618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0565529275954214e-05,
+      "loss": 1.8754,
+      "step": 12619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.051371604687211e-05,
+      "loss": 1.9559,
+      "step": 12620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0461923562553314e-05,
+      "loss": 1.9666,
+      "step": 12621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.041015182544252e-05,
+      "loss": 2.0645,
+      "step": 12622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.035840083798333e-05,
+      "loss": 1.9416,
+      "step": 12623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.030667060261874e-05,
+      "loss": 1.9838,
+      "step": 12624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.025496112179046e-05,
+      "loss": 2.0673,
+      "step": 12625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.020327239793932e-05,
+      "loss": 2.0088,
+      "step": 12626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.0151604433504984e-05,
+      "loss": 1.95,
+      "step": 12627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.009995723092654e-05,
+      "loss": 1.9412,
+      "step": 12628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.004833079264172e-05,
+      "loss": 2.0137,
+      "step": 12629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.999672512108745e-05,
+      "loss": 2.1426,
+      "step": 12630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9945140218699515e-05,
+      "loss": 1.9936,
+      "step": 12631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.989357608791307e-05,
+      "loss": 1.8997,
+      "step": 12632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.98420327311619e-05,
+      "loss": 2.018,
+      "step": 12633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9790510150878994e-05,
+      "loss": 2.0675,
+      "step": 12634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.973900834949625e-05,
+      "loss": 2.0457,
+      "step": 12635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.968752732944488e-05,
+      "loss": 2.0421,
+      "step": 12636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.963606709315483e-05,
+      "loss": 1.9809,
+      "step": 12637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.958462764305505e-05,
+      "loss": 1.921,
+      "step": 12638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9533208981573604e-05,
+      "loss": 2.1761,
+      "step": 12639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.948181111113765e-05,
+      "loss": 1.957,
+      "step": 12640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.943043403417337e-05,
+      "loss": 2.0251,
+      "step": 12641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.937907775310569e-05,
+      "loss": 1.9854,
+      "step": 12642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.93277422703587e-05,
+      "loss": 2.0577,
+      "step": 12643
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.927642758835583e-05,
+      "loss": 2.1812,
+      "step": 12644
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.92251337095191e-05,
+      "loss": 1.9743,
+      "step": 12645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.917386063626967e-05,
+      "loss": 2.029,
+      "step": 12646
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9122608371027675e-05,
+      "loss": 2.0574,
+      "step": 12647
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.907137691621256e-05,
+      "loss": 1.9853,
+      "step": 12648
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.902016627424245e-05,
+      "loss": 1.96,
+      "step": 12649
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.896897644753457e-05,
+      "loss": 1.9569,
+      "step": 12650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.891780743850517e-05,
+      "loss": 2.1183,
+      "step": 12651
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.886665924956969e-05,
+      "loss": 2.0123,
+      "step": 12652
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.881553188314231e-05,
+      "loss": 1.9839,
+      "step": 12653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.876442534163645e-05,
+      "loss": 1.8721,
+      "step": 12654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8713339627464334e-05,
+      "loss": 1.9988,
+      "step": 12655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8662274743037425e-05,
+      "loss": 1.968,
+      "step": 12656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.861123069076612e-05,
+      "loss": 2.0111,
+      "step": 12657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.856020747305979e-05,
+      "loss": 2.0098,
+      "step": 12658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8509205092326767e-05,
+      "loss": 1.9797,
+      "step": 12659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8458223550974575e-05,
+      "loss": 2.0591,
+      "step": 12660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8407262851409683e-05,
+      "loss": 1.9939,
+      "step": 12661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.835632299603749e-05,
+      "loss": 2.1579,
+      "step": 12662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.830540398726236e-05,
+      "loss": 1.8697,
+      "step": 12663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.825450582748804e-05,
+      "loss": 1.8209,
+      "step": 12664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8203628519116925e-05,
+      "loss": 1.9803,
+      "step": 12665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.81527720645505e-05,
+      "loss": 1.9702,
+      "step": 12666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8101936466189274e-05,
+      "loss": 1.908,
+      "step": 12667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.805112172643295e-05,
+      "loss": 2.0302,
+      "step": 12668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.800032784768e-05,
+      "loss": 2.0274,
+      "step": 12669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7949554832328055e-05,
+      "loss": 1.8783,
+      "step": 12670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.789880268277359e-05,
+      "loss": 1.9695,
+      "step": 12671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.784807140141246e-05,
+      "loss": 2.0495,
+      "step": 12672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.779736099063909e-05,
+      "loss": 1.8913,
+      "step": 12673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7746671452847234e-05,
+      "loss": 1.9534,
+      "step": 12674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.769600279042947e-05,
+      "loss": 2.0785,
+      "step": 12675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7645355005777556e-05,
+      "loss": 2.1277,
+      "step": 12676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.759472810128219e-05,
+      "loss": 1.9143,
+      "step": 12677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.754412207933307e-05,
+      "loss": 1.9696,
+      "step": 12678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.749353694231874e-05,
+      "loss": 2.1162,
+      "step": 12679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.744297269262722e-05,
+      "loss": 2.0733,
+      "step": 12680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.73924293326451e-05,
+      "loss": 1.9212,
+      "step": 12681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7341906864758206e-05,
+      "loss": 2.0913,
+      "step": 12682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7291405291351184e-05,
+      "loss": 1.9164,
+      "step": 12683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.724092461480796e-05,
+      "loss": 2.0796,
+      "step": 12684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.719046483751134e-05,
+      "loss": 1.9789,
+      "step": 12685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7140025961843086e-05,
+      "loss": 1.9347,
+      "step": 12686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.708960799018392e-05,
+      "loss": 2.0506,
+      "step": 12687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.703921092491393e-05,
+      "loss": 2.0666,
+      "step": 12688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.698883476841182e-05,
+      "loss": 1.779,
+      "step": 12689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.693847952305553e-05,
+      "loss": 2.0319,
+      "step": 12690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6888145191221765e-05,
+      "loss": 1.9717,
+      "step": 12691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.683783177528667e-05,
+      "loss": 1.9462,
+      "step": 12692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.678753927762509e-05,
+      "loss": 2.0028,
+      "step": 12693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.673726770061083e-05,
+      "loss": 2.0552,
+      "step": 12694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.668701704661683e-05,
+      "loss": 2.0445,
+      "step": 12695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.663678731801519e-05,
+      "loss": 1.7257,
+      "step": 12696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.65865785171768e-05,
+      "loss": 1.9632,
+      "step": 12697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6536390646471594e-05,
+      "loss": 1.8872,
+      "step": 12698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.648622370826845e-05,
+      "loss": 2.0776,
+      "step": 12699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.643607770493564e-05,
+      "loss": 2.2398,
+      "step": 12700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.638595263883994e-05,
+      "loss": 1.9397,
+      "step": 12701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.633584851234752e-05,
+      "loss": 1.976,
+      "step": 12702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6285765327823204e-05,
+      "loss": 1.9805,
+      "step": 12703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6235703087631206e-05,
+      "loss": 1.9725,
+      "step": 12704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.618566179413459e-05,
+      "loss": 2.1751,
+      "step": 12705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6135641449695244e-05,
+      "loss": 2.0033,
+      "step": 12706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.608564205667449e-05,
+      "loss": 2.08,
+      "step": 12707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.603566361743229e-05,
+      "loss": 1.9915,
+      "step": 12708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.598570613432774e-05,
+      "loss": 2.0543,
+      "step": 12709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.59357696097188e-05,
+      "loss": 2.064,
+      "step": 12710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.588585404596286e-05,
+      "loss": 2.0288,
+      "step": 12711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5835959445415975e-05,
+      "loss": 1.9929,
+      "step": 12712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.578608581043315e-05,
+      "loss": 2.0718,
+      "step": 12713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.573623314336857e-05,
+      "loss": 2.0874,
+      "step": 12714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5686401446575555e-05,
+      "loss": 2.1873,
+      "step": 12715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.563659072240618e-05,
+      "loss": 1.9227,
+      "step": 12716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.558680097321156e-05,
+      "loss": 2.0228,
+      "step": 12717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5537032201341874e-05,
+      "loss": 2.1067,
+      "step": 12718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.548728440914652e-05,
+      "loss": 2.0274,
+      "step": 12719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.543755759897351e-05,
+      "loss": 1.9361,
+      "step": 12720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.538785177317018e-05,
+      "loss": 1.8495,
+      "step": 12721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5338166934082556e-05,
+      "loss": 2.072,
+      "step": 12722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5288503084056195e-05,
+      "loss": 1.9062,
+      "step": 12723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5238860225435116e-05,
+      "loss": 2.0187,
+      "step": 12724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.518923836056267e-05,
+      "loss": 2.0627,
+      "step": 12725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5139637491781025e-05,
+      "loss": 2.1212,
+      "step": 12726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.509005762143154e-05,
+      "loss": 2.0436,
+      "step": 12727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.5040498751854576e-05,
+      "loss": 1.89,
+      "step": 12728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4990960885389296e-05,
+      "loss": 2.0194,
+      "step": 12729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.49414440243739e-05,
+      "loss": 1.9394,
+      "step": 12730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4891948171145955e-05,
+      "loss": 2.0295,
+      "step": 12731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4842473328041707e-05,
+      "loss": 2.0118,
+      "step": 12732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.47930194973964e-05,
+      "loss": 2.1924,
+      "step": 12733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4743586681544275e-05,
+      "loss": 1.9605,
+      "step": 12734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4694174882818915e-05,
+      "loss": 1.9583,
+      "step": 12735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.464478410355256e-05,
+      "loss": 1.9323,
+      "step": 12736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.459541434607662e-05,
+      "loss": 2.089,
+      "step": 12737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.454606561272124e-05,
+      "loss": 2.0656,
+      "step": 12738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.449673790581611e-05,
+      "loss": 2.0593,
+      "step": 12739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.444743122768942e-05,
+      "loss": 2.0369,
+      "step": 12740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.439814558066864e-05,
+      "loss": 1.9755,
+      "step": 12741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4348880967080016e-05,
+      "loss": 2.0001,
+      "step": 12742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.4299637389249145e-05,
+      "loss": 1.947,
+      "step": 12743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.425041484950033e-05,
+      "loss": 1.9944,
+      "step": 12744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.420121335015704e-05,
+      "loss": 1.9584,
+      "step": 12745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.415203289354159e-05,
+      "loss": 2.0636,
+      "step": 12746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.410287348197557e-05,
+      "loss": 1.7837,
+      "step": 12747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.405373511777939e-05,
+      "loss": 1.9475,
+      "step": 12748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.4004617803272374e-05,
+      "loss": 2.2063,
+      "step": 12749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.395552154077299e-05,
+      "loss": 2.0428,
+      "step": 12750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.390644633259878e-05,
+      "loss": 2.0058,
+      "step": 12751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3857392181066223e-05,
+      "loss": 2.0433,
+      "step": 12752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3808359088490744e-05,
+      "loss": 2.0378,
+      "step": 12753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.375934705718671e-05,
+      "loss": 2.1033,
+      "step": 12754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.371035608946778e-05,
+      "loss": 1.8958,
+      "step": 12755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3661386187646376e-05,
+      "loss": 1.915,
+      "step": 12756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3612437354033985e-05,
+      "loss": 1.901,
+      "step": 12757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.356350959094097e-05,
+      "loss": 2.1033,
+      "step": 12758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.351460290067706e-05,
+      "loss": 2.0771,
+      "step": 12759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.346571728555066e-05,
+      "loss": 1.8416,
+      "step": 12760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.341685274786934e-05,
+      "loss": 2.0389,
+      "step": 12761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3368009289939404e-05,
+      "loss": 1.8649,
+      "step": 12762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.331918691406668e-05,
+      "loss": 2.1274,
+      "step": 12763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.327038562255554e-05,
+      "loss": 2.1297,
+      "step": 12764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3221605417709486e-05,
+      "loss": 2.0085,
+      "step": 12765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3172846301831056e-05,
+      "loss": 1.8763,
+      "step": 12766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3124108277221914e-05,
+      "loss": 1.951,
+      "step": 12767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.3075391346182554e-05,
+      "loss": 1.876,
+      "step": 12768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.302669551101247e-05,
+      "loss": 2.0731,
+      "step": 12769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2978020774010204e-05,
+      "loss": 2.1113,
+      "step": 12770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2929367137473425e-05,
+      "loss": 1.8582,
+      "step": 12771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.288073460369863e-05,
+      "loss": 2.0163,
+      "step": 12772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2832123174981376e-05,
+      "loss": 1.9871,
+      "step": 12773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.278353285361614e-05,
+      "loss": 1.9201,
+      "step": 12774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.273496364189673e-05,
+      "loss": 1.9328,
+      "step": 12775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.268641554211562e-05,
+      "loss": 1.919,
+      "step": 12776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2637888556564316e-05,
+      "loss": 1.9251,
+      "step": 12777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.258938268753344e-05,
+      "loss": 1.9338,
+      "step": 12778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.254089793731265e-05,
+      "loss": 2.0873,
+      "step": 12779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.249243430819045e-05,
+      "loss": 2.0125,
+      "step": 12780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2443991802454515e-05,
+      "loss": 2.1057,
+      "step": 12781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.239557042239129e-05,
+      "loss": 2.0197,
+      "step": 12782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.234717017028662e-05,
+      "loss": 1.8939,
+      "step": 12783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2298791048424945e-05,
+      "loss": 2.0321,
+      "step": 12784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2250433059089886e-05,
+      "loss": 1.9973,
+      "step": 12785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2202096204563954e-05,
+      "loss": 2.0329,
+      "step": 12786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.215378048712899e-05,
+      "loss": 2.0942,
+      "step": 12787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.21054859090655e-05,
+      "loss": 2.0104,
+      "step": 12788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.205721247265305e-05,
+      "loss": 2.0682,
+      "step": 12789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2008960180170264e-05,
+      "loss": 2.0181,
+      "step": 12790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1960729033894814e-05,
+      "loss": 2.2156,
+      "step": 12791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1912519036103336e-05,
+      "loss": 1.9544,
+      "step": 12792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.186433018907139e-05,
+      "loss": 1.9775,
+      "step": 12793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1816162495073536e-05,
+      "loss": 2.1403,
+      "step": 12794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1768015956383584e-05,
+      "loss": 2.0907,
+      "step": 12795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.17198905752741e-05,
+      "loss": 2.1701,
+      "step": 12796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1671786354016537e-05,
+      "loss": 1.9878,
+      "step": 12797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1623703294881754e-05,
+      "loss": 1.8801,
+      "step": 12798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.157564140013932e-05,
+      "loss": 2.0126,
+      "step": 12799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.152760067205786e-05,
+      "loss": 2.0994,
+      "step": 12800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.147958111290485e-05,
+      "loss": 2.0325,
+      "step": 12801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.143158272494719e-05,
+      "loss": 2.1403,
+      "step": 12802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.138360551045041e-05,
+      "loss": 2.0077,
+      "step": 12803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.133564947167907e-05,
+      "loss": 2.0194,
+      "step": 12804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.128771461089676e-05,
+      "loss": 2.1102,
+      "step": 12805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.123980093036634e-05,
+      "loss": 1.9642,
+      "step": 12806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1191908432349274e-05,
+      "loss": 1.8336,
+      "step": 12807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.1144037119106315e-05,
+      "loss": 2.148,
+      "step": 12808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.109618699289687e-05,
+      "loss": 2.0016,
+      "step": 12809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.104835805597985e-05,
+      "loss": 1.9726,
+      "step": 12810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.100055031061274e-05,
+      "loss": 2.1211,
+      "step": 12811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.095276375905222e-05,
+      "loss": 1.9925,
+      "step": 12812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0904998403553834e-05,
+      "loss": 2.109,
+      "step": 12813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0857254246372387e-05,
+      "loss": 1.987,
+      "step": 12814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.08095312897614e-05,
+      "loss": 1.8514,
+      "step": 12815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.076182953597358e-05,
+      "loss": 1.9622,
+      "step": 12816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0714148987260343e-05,
+      "loss": 1.9317,
+      "step": 12817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0666489645872614e-05,
+      "loss": 2.0567,
+      "step": 12818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.061885151405993e-05,
+      "loss": 2.0565,
+      "step": 12819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0571234594070814e-05,
+      "loss": 1.9535,
+      "step": 12820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.052363888815292e-05,
+      "loss": 1.9295,
+      "step": 12821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0476064398553014e-05,
+      "loss": 1.9802,
+      "step": 12822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0428511127516565e-05,
+      "loss": 1.9883,
+      "step": 12823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0380979077288335e-05,
+      "loss": 1.8517,
+      "step": 12824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.03334682501117e-05,
+      "loss": 1.9504,
+      "step": 12825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.028597864822959e-05,
+      "loss": 1.9835,
+      "step": 12826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0238510273883486e-05,
+      "loss": 1.9585,
+      "step": 12827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.0191063129313986e-05,
+      "loss": 2.0562,
+      "step": 12828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.014363721676063e-05,
+      "loss": 2.1351,
+      "step": 12829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.009623253846224e-05,
+      "loss": 1.9089,
+      "step": 12830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.00488490966563e-05,
+      "loss": 2.0597,
+      "step": 12831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.000148689357936e-05,
+      "loss": 2.03,
+      "step": 12832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9954145931467075e-05,
+      "loss": 1.9644,
+      "step": 12833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.99068262125541e-05,
+      "loss": 1.9119,
+      "step": 12834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.985952773907404e-05,
+      "loss": 2.0037,
+      "step": 12835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9812250513259385e-05,
+      "loss": 1.9374,
+      "step": 12836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9764994537341736e-05,
+      "loss": 2.0796,
+      "step": 12837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.971775981355181e-05,
+      "loss": 2.0855,
+      "step": 12838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.967054634411916e-05,
+      "loss": 2.0649,
+      "step": 12839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.962335413127234e-05,
+      "loss": 2.0332,
+      "step": 12840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.957618317723878e-05,
+      "loss": 2.0983,
+      "step": 12841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.952903348424531e-05,
+      "loss": 2.1405,
+      "step": 12842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9481905054517383e-05,
+      "loss": 1.9917,
+      "step": 12843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9434797890279605e-05,
+      "loss": 2.0017,
+      "step": 12844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.938771199375541e-05,
+      "loss": 2.0301,
+      "step": 12845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.934064736716759e-05,
+      "loss": 1.9677,
+      "step": 12846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.929360401273758e-05,
+      "loss": 1.9186,
+      "step": 12847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.924658193268594e-05,
+      "loss": 1.9868,
+      "step": 12848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.919958112923212e-05,
+      "loss": 1.9581,
+      "step": 12849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9152601604594835e-05,
+      "loss": 2.0037,
+      "step": 12850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.910564336099155e-05,
+      "loss": 1.9808,
+      "step": 12851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.905870640063886e-05,
+      "loss": 2.0567,
+      "step": 12852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.9011790725752124e-05,
+      "loss": 1.8879,
+      "step": 12853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8964896338546064e-05,
+      "loss": 2.0662,
+      "step": 12854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.891802324123412e-05,
+      "loss": 1.953,
+      "step": 12855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.887117143602887e-05,
+      "loss": 1.7998,
+      "step": 12856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.882434092514165e-05,
+      "loss": 1.9585,
+      "step": 12857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8777531710783195e-05,
+      "loss": 2.1031,
+      "step": 12858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.87307437951629e-05,
+      "loss": 2.1798,
+      "step": 12859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.868397718048928e-05,
+      "loss": 1.9458,
+      "step": 12860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.863723186896973e-05,
+      "loss": 2.1065,
+      "step": 12861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.859050786281094e-05,
+      "loss": 1.9652,
+      "step": 12862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.854380516421825e-05,
+      "loss": 2.1255,
+      "step": 12863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8497123775396115e-05,
+      "loss": 1.9663,
+      "step": 12864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8450463698547996e-05,
+      "loss": 2.041,
+      "step": 12865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8403824935876476e-05,
+      "loss": 1.9369,
+      "step": 12866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.835720748958295e-05,
+      "loss": 1.9802,
+      "step": 12867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.831061136186787e-05,
+      "loss": 1.8578,
+      "step": 12868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8264036554930616e-05,
+      "loss": 1.9737,
+      "step": 12869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.821748307096974e-05,
+      "loss": 1.9719,
+      "step": 12870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.81709509121826e-05,
+      "loss": 1.9266,
+      "step": 12871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.812444008076561e-05,
+      "loss": 2.0898,
+      "step": 12872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8077950578914174e-05,
+      "loss": 1.9173,
+      "step": 12873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8031482408822824e-05,
+      "loss": 1.9599,
+      "step": 12874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.79850355726848e-05,
+      "loss": 2.0694,
+      "step": 12875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.793861007269262e-05,
+      "loss": 1.9684,
+      "step": 12876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.78922059110376e-05,
+      "loss": 1.9806,
+      "step": 12877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7845823089910255e-05,
+      "loss": 2.146,
+      "step": 12878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.779946161149984e-05,
+      "loss": 2.1194,
+      "step": 12879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7753121477994766e-05,
+      "loss": 1.9354,
+      "step": 12880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7706802691582284e-05,
+      "loss": 2.0804,
+      "step": 12881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.766050525444893e-05,
+      "loss": 1.9645,
+      "step": 12882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.761422916878e-05,
+      "loss": 2.0182,
+      "step": 12883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.756797443675975e-05,
+      "loss": 1.9459,
+      "step": 12884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.752174106057144e-05,
+      "loss": 2.0665,
+      "step": 12885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.747552904239766e-05,
+      "loss": 1.9313,
+      "step": 12886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.742933838441954e-05,
+      "loss": 2.0917,
+      "step": 12887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.73831690888174e-05,
+      "loss": 2.0505,
+      "step": 12888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7337021157770446e-05,
+      "loss": 1.9834,
+      "step": 12889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.729089459345714e-05,
+      "loss": 1.9691,
+      "step": 12890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.724478939805477e-05,
+      "loss": 1.8959,
+      "step": 12891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7198705573739454e-05,
+      "loss": 2.1461,
+      "step": 12892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.715264312268652e-05,
+      "loss": 1.8837,
+      "step": 12893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.710660204707024e-05,
+      "loss": 2.0884,
+      "step": 12894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.706058234906391e-05,
+      "loss": 1.9528,
+      "step": 12895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.701458403083969e-05,
+      "loss": 2.0425,
+      "step": 12896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.696860709456868e-05,
+      "loss": 1.9501,
+      "step": 12897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.692265154242137e-05,
+      "loss": 1.8198,
+      "step": 12898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6876717376566845e-05,
+      "loss": 2.0977,
+      "step": 12899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.683080459917333e-05,
+      "loss": 2.0082,
+      "step": 12900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.67849132124078e-05,
+      "loss": 2.0208,
+      "step": 12901
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.673904321843675e-05,
+      "loss": 2.0742,
+      "step": 12902
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6693194619425225e-05,
+      "loss": 1.9819,
+      "step": 12903
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.664736741753739e-05,
+      "loss": 2.0432,
+      "step": 12904
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.660156161493623e-05,
+      "loss": 2.0524,
+      "step": 12905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6555777213784126e-05,
+      "loss": 1.9664,
+      "step": 12906
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.651001421624212e-05,
+      "loss": 2.0229,
+      "step": 12907
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6464272624470336e-05,
+      "loss": 1.965,
+      "step": 12908
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.641855244062781e-05,
+      "loss": 2.0324,
+      "step": 12909
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6372853666872715e-05,
+      "loss": 1.9521,
+      "step": 12910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.632717630536221e-05,
+      "loss": 2.038,
+      "step": 12911
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6281520358252174e-05,
+      "loss": 1.9238,
+      "step": 12912
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6235885827697775e-05,
+      "loss": 1.9335,
+      "step": 12913
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.619027271585313e-05,
+      "loss": 2.0726,
+      "step": 12914
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.614468102487124e-05,
+      "loss": 1.9543,
+      "step": 12915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6099110756904096e-05,
+      "loss": 2.0006,
+      "step": 12916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6053561914102706e-05,
+      "loss": 1.8981,
+      "step": 12917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6008034498617176e-05,
+      "loss": 1.9275,
+      "step": 12918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.596252851259647e-05,
+      "loss": 2.0184,
+      "step": 12919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.591704395818852e-05,
+      "loss": 2.0482,
+      "step": 12920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5871580837540285e-05,
+      "loss": 2.1289,
+      "step": 12921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5826139152797817e-05,
+      "loss": 2.0086,
+      "step": 12922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.578071890610608e-05,
+      "loss": 2.0725,
+      "step": 12923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.573532009960896e-05,
+      "loss": 1.9908,
+      "step": 12924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.56899427354493e-05,
+      "loss": 1.8622,
+      "step": 12925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.564458681576922e-05,
+      "loss": 1.8901,
+      "step": 12926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.559925234270951e-05,
+      "loss": 1.8766,
+      "step": 12927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.555393931841001e-05,
+      "loss": 2.1223,
+      "step": 12928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5508647745009626e-05,
+      "loss": 1.8853,
+      "step": 12929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.546337762464631e-05,
+      "loss": 2.0514,
+      "step": 12930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5418128959456853e-05,
+      "loss": 2.0528,
+      "step": 12931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.537290175157705e-05,
+      "loss": 1.9896,
+      "step": 12932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.532769600314174e-05,
+      "loss": 2.0004,
+      "step": 12933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.528251171628489e-05,
+      "loss": 2.0354,
+      "step": 12934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.523734889313913e-05,
+      "loss": 1.8738,
+      "step": 12935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.519220753583636e-05,
+      "loss": 2.1707,
+      "step": 12936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.514708764650721e-05,
+      "loss": 2.1468,
+      "step": 12937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.510198922728159e-05,
+      "loss": 2.0108,
+      "step": 12938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.505691228028819e-05,
+      "loss": 2.0577,
+      "step": 12939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5011856807654794e-05,
+      "loss": 2.0277,
+      "step": 12940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.496682281150799e-05,
+      "loss": 2.0536,
+      "step": 12941
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.492181029397363e-05,
+      "loss": 2.1463,
+      "step": 12942
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.487681925717635e-05,
+      "loss": 2.0669,
+      "step": 12943
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.483184970323989e-05,
+      "loss": 2.0326,
+      "step": 12944
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4786901634286724e-05,
+      "loss": 1.9351,
+      "step": 12945
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4741975052438764e-05,
+      "loss": 2.0031,
+      "step": 12946
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.469706995981648e-05,
+      "loss": 1.976,
+      "step": 12947
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.465218635853957e-05,
+      "loss": 2.0104,
+      "step": 12948
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.460732425072656e-05,
+      "loss": 1.984,
+      "step": 12949
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4562483638495136e-05,
+      "loss": 1.878,
+      "step": 12950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.451766452396183e-05,
+      "loss": 2.0722,
+      "step": 12951
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.447286690924229e-05,
+      "loss": 1.9664,
+      "step": 12952
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4428090796450825e-05,
+      "loss": 1.9423,
+      "step": 12953
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.438333618770129e-05,
+      "loss": 1.9925,
+      "step": 12954
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.433860308510601e-05,
+      "loss": 2.0209,
+      "step": 12955
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.429389149077662e-05,
+      "loss": 1.9759,
+      "step": 12956
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.424920140682337e-05,
+      "loss": 1.767,
+      "step": 12957
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.420453283535597e-05,
+      "loss": 2.0079,
+      "step": 12958
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4159885778482845e-05,
+      "loss": 2.0213,
+      "step": 12959
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.411526023831136e-05,
+      "loss": 1.882,
+      "step": 12960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.407065621694789e-05,
+      "loss": 2.1134,
+      "step": 12961
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.402607371649803e-05,
+      "loss": 2.023,
+      "step": 12962
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.398151273906609e-05,
+      "loss": 2.0816,
+      "step": 12963
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3936973286755444e-05,
+      "loss": 2.0256,
+      "step": 12964
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.389245536166836e-05,
+      "loss": 2.0078,
+      "step": 12965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3847958965906374e-05,
+      "loss": 1.8112,
+      "step": 12966
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3803484101569704e-05,
+      "loss": 2.034,
+      "step": 12967
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3759030770757604e-05,
+      "loss": 1.8846,
+      "step": 12968
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.371459897556856e-05,
+      "loss": 2.0389,
+      "step": 12969
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.367018871809975e-05,
+      "loss": 2.0844,
+      "step": 12970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3625800000447415e-05,
+      "loss": 2.065,
+      "step": 12971
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.358143282470678e-05,
+      "loss": 2.0099,
+      "step": 12972
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.353708719297222e-05,
+      "loss": 1.9904,
+      "step": 12973
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.349276310733685e-05,
+      "loss": 1.9213,
+      "step": 12974
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.344846056989288e-05,
+      "loss": 1.9047,
+      "step": 12975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.340417958273135e-05,
+      "loss": 2.0303,
+      "step": 12976
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.335992014794271e-05,
+      "loss": 2.1572,
+      "step": 12977
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.331568226761595e-05,
+      "loss": 2.001,
+      "step": 12978
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.327146594383918e-05,
+      "loss": 1.9706,
+      "step": 12979
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.322727117869951e-05,
+      "loss": 1.9674,
+      "step": 12980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3183097974283156e-05,
+      "loss": 2.0396,
+      "step": 12981
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3138946332675064e-05,
+      "loss": 1.9626,
+      "step": 12982
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.309481625595935e-05,
+      "loss": 2.0432,
+      "step": 12983
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.305070774621894e-05,
+      "loss": 2.135,
+      "step": 12984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3006620805536075e-05,
+      "loss": 2.0625,
+      "step": 12985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.296255543599165e-05,
+      "loss": 2.012,
+      "step": 12986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.291851163966565e-05,
+      "loss": 1.8404,
+      "step": 12987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.287448941863692e-05,
+      "loss": 1.956,
+      "step": 12988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2830488774983576e-05,
+      "loss": 1.9483,
+      "step": 12989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2786509710782574e-05,
+      "loss": 2.0153,
+      "step": 12990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.274255222810974e-05,
+      "loss": 1.9142,
+      "step": 12991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.269861632903988e-05,
+      "loss": 2.008,
+      "step": 12992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2654702015647086e-05,
+      "loss": 2.0166,
+      "step": 12993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.261080929000405e-05,
+      "loss": 2.0119,
+      "step": 12994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.256693815418272e-05,
+      "loss": 2.1072,
+      "step": 12995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.252308861025372e-05,
+      "loss": 2.0968,
+      "step": 12996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.247926066028712e-05,
+      "loss": 1.9577,
+      "step": 12997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.243545430635154e-05,
+      "loss": 1.9085,
+      "step": 12998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2391669550514764e-05,
+      "loss": 2.1026,
+      "step": 12999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.234790639484343e-05,
+      "loss": 1.9902,
+      "step": 13000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.230416484140348e-05,
+      "loss": 1.9682,
+      "step": 13001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.22604448922595e-05,
+      "loss": 2.0673,
+      "step": 13002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.221674654947516e-05,
+      "loss": 2.0185,
+      "step": 13003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.217306981511304e-05,
+      "loss": 1.9304,
+      "step": 13004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.212941469123499e-05,
+      "loss": 1.9934,
+      "step": 13005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2085781179901474e-05,
+      "loss": 2.0114,
+      "step": 13006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.204216928317217e-05,
+      "loss": 2.0877,
+      "step": 13007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.199857900310555e-05,
+      "loss": 1.9013,
+      "step": 13008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.19550103417593e-05,
+      "loss": 1.9708,
+      "step": 13009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.191146330119e-05,
+      "loss": 1.9909,
+      "step": 13010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1867937883453e-05,
+      "loss": 2.0759,
+      "step": 13011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1824434090602824e-05,
+      "loss": 2.0048,
+      "step": 13012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1780951924693113e-05,
+      "loss": 1.9556,
+      "step": 13013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.173749138777622e-05,
+      "loss": 1.9932,
+      "step": 13014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.169405248190361e-05,
+      "loss": 2.127,
+      "step": 13015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.165063520912554e-05,
+      "loss": 2.1613,
+      "step": 13016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.160723957149165e-05,
+      "loss": 2.0252,
+      "step": 13017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.156386557105024e-05,
+      "loss": 2.0088,
+      "step": 13018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.152051320984856e-05,
+      "loss": 2.1236,
+      "step": 13019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1477182489932985e-05,
+      "loss": 2.0578,
+      "step": 13020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.143387341334892e-05,
+      "loss": 1.9599,
+      "step": 13021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.139058598214057e-05,
+      "loss": 1.9024,
+      "step": 13022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1347320198351194e-05,
+      "loss": 2.0765,
+      "step": 13023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.130407606402303e-05,
+      "loss": 1.8056,
+      "step": 13024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.126085358119736e-05,
+      "loss": 1.7192,
+      "step": 13025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.121765275191436e-05,
+      "loss": 2.0597,
+      "step": 13026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.117447357821319e-05,
+      "loss": 2.0166,
+      "step": 13027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.113131606213194e-05,
+      "loss": 2.008,
+      "step": 13028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1088180205707924e-05,
+      "loss": 1.9869,
+      "step": 13029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1045066010977125e-05,
+      "loss": 1.9822,
+      "step": 13030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.10019734799747e-05,
+      "loss": 1.9067,
+      "step": 13031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.095890261473456e-05,
+      "loss": 1.9496,
+      "step": 13032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0915853417289925e-05,
+      "loss": 1.9999,
+      "step": 13033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0872825889672783e-05,
+      "loss": 1.9405,
+      "step": 13034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0829820033914055e-05,
+      "loss": 2.0216,
+      "step": 13035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.078683585204373e-05,
+      "loss": 2.0586,
+      "step": 13036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0743873346090866e-05,
+      "loss": 1.9553,
+      "step": 13037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0700932518083267e-05,
+      "loss": 2.0154,
+      "step": 13038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.065801337004793e-05,
+      "loss": 2.1981,
+      "step": 13039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.061511590401062e-05,
+      "loss": 2.0274,
+      "step": 13040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0572240121996394e-05,
+      "loss": 2.1734,
+      "step": 13041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.05293860260289e-05,
+      "loss": 1.9923,
+      "step": 13042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0486553618131026e-05,
+      "loss": 2.0229,
+      "step": 13043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.044374290032449e-05,
+      "loss": 1.794,
+      "step": 13044
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.040095387463022e-05,
+      "loss": 1.8911,
+      "step": 13045
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.035818654306783e-05,
+      "loss": 2.0144,
+      "step": 13046
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.03154409076561e-05,
+      "loss": 2.0634,
+      "step": 13047
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0272716970412514e-05,
+      "loss": 1.9716,
+      "step": 13048
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.023001473335408e-05,
+      "loss": 2.0498,
+      "step": 13049
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.018733419849624e-05,
+      "loss": 2.0035,
+      "step": 13050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.01446753678536e-05,
+      "loss": 1.9401,
+      "step": 13051
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.010203824343978e-05,
+      "loss": 1.993,
+      "step": 13052
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.005942282726743e-05,
+      "loss": 2.1835,
+      "step": 13053
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.001682912134802e-05,
+      "loss": 2.126,
+      "step": 13054
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.997425712769209e-05,
+      "loss": 1.9882,
+      "step": 13055
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.99317068483091e-05,
+      "loss": 2.0064,
+      "step": 13056
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.988917828520755e-05,
+      "loss": 1.9224,
+      "step": 13057
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.984667144039495e-05,
+      "loss": 2.0155,
+      "step": 13058
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.980418631587757e-05,
+      "loss": 1.991,
+      "step": 13059
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.976172291366098e-05,
+      "loss": 1.92,
+      "step": 13060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.971928123574947e-05,
+      "loss": 2.06,
+      "step": 13061
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.967686128414633e-05,
+      "loss": 2.0393,
+      "step": 13062
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.963446306085389e-05,
+      "loss": 2.04,
+      "step": 13063
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.959208656787355e-05,
+      "loss": 1.8235,
+      "step": 13064
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9549731807205556e-05,
+      "loss": 1.9669,
+      "step": 13065
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.950739878084908e-05,
+      "loss": 1.9014,
+      "step": 13066
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9465087490802244e-05,
+      "loss": 2.0058,
+      "step": 13067
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9422797939062504e-05,
+      "loss": 2.0586,
+      "step": 13068
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9380530127625824e-05,
+      "loss": 2.1235,
+      "step": 13069
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.933828405848744e-05,
+      "loss": 1.7943,
+      "step": 13070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.929605973364131e-05,
+      "loss": 1.8655,
+      "step": 13071
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9253857155080677e-05,
+      "loss": 2.0746,
+      "step": 13072
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9211676324797606e-05,
+      "loss": 1.9546,
+      "step": 13073
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9169517244783074e-05,
+      "loss": 2.0938,
+      "step": 13074
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9127379917026976e-05,
+      "loss": 2.0501,
+      "step": 13075
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.90852643435185e-05,
+      "loss": 1.9014,
+      "step": 13076
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.904317052624551e-05,
+      "loss": 2.0141,
+      "step": 13077
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.900109846719491e-05,
+      "loss": 2.1929,
+      "step": 13078
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8959048168352496e-05,
+      "loss": 2.0415,
+      "step": 13079
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8917019631703354e-05,
+      "loss": 2.0111,
+      "step": 13080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.887501285923123e-05,
+      "loss": 1.8892,
+      "step": 13081
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.883302785291898e-05,
+      "loss": 2.0346,
+      "step": 13082
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8791064614748185e-05,
+      "loss": 2.0745,
+      "step": 13083
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.874912314669993e-05,
+      "loss": 2.0935,
+      "step": 13084
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8707203450753735e-05,
+      "loss": 2.0654,
+      "step": 13085
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.866530552888836e-05,
+      "loss": 1.9507,
+      "step": 13086
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8623429383081445e-05,
+      "loss": 1.998,
+      "step": 13087
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8581575015309736e-05,
+      "loss": 2.0142,
+      "step": 13088
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.853974242754882e-05,
+      "loss": 1.9446,
+      "step": 13089
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.849793162177323e-05,
+      "loss": 2.0512,
+      "step": 13090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.84561425999565e-05,
+      "loss": 1.8882,
+      "step": 13091
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.841437536407133e-05,
+      "loss": 2.1325,
+      "step": 13092
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.837262991608914e-05,
+      "loss": 1.8873,
+      "step": 13093
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8330906257980356e-05,
+      "loss": 2.1054,
+      "step": 13094
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.828920439171446e-05,
+      "loss": 1.8607,
+      "step": 13095
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.824752431925999e-05,
+      "loss": 1.9444,
+      "step": 13096
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.820586604258425e-05,
+      "loss": 2.0035,
+      "step": 13097
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8164229563653564e-05,
+      "loss": 1.9279,
+      "step": 13098
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.812261488443325e-05,
+      "loss": 2.0905,
+      "step": 13099
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.808102200688779e-05,
+      "loss": 1.9802,
+      "step": 13100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8039450932980334e-05,
+      "loss": 2.0604,
+      "step": 13101
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.799790166467315e-05,
+      "loss": 2.1541,
+      "step": 13102
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7956374203927394e-05,
+      "loss": 1.9953,
+      "step": 13103
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7914868552703375e-05,
+      "loss": 2.0374,
+      "step": 13104
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7873384712960254e-05,
+      "loss": 2.0439,
+      "step": 13105
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.783192268665608e-05,
+      "loss": 1.987,
+      "step": 13106
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.77904824757479e-05,
+      "loss": 2.1273,
+      "step": 13107
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.774906408219197e-05,
+      "loss": 2.0405,
+      "step": 13108
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7707667507943244e-05,
+      "loss": 1.9739,
+      "step": 13109
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.766629275495576e-05,
+      "loss": 1.9359,
+      "step": 13110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7624939825182336e-05,
+      "loss": 1.9867,
+      "step": 13111
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.75836087205752e-05,
+      "loss": 2.1011,
+      "step": 13112
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.754229944308513e-05,
+      "loss": 1.9404,
+      "step": 13113
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.750101199466199e-05,
+      "loss": 1.9163,
+      "step": 13114
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.745974637725463e-05,
+      "loss": 1.9704,
+      "step": 13115
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.741850259281099e-05,
+      "loss": 1.9314,
+      "step": 13116
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7377280643277834e-05,
+      "loss": 2.0357,
+      "step": 13117
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.733608053060089e-05,
+      "loss": 1.9346,
+      "step": 13118
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7294902256724885e-05,
+      "loss": 2.0275,
+      "step": 13119
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.725374582359359e-05,
+      "loss": 2.0236,
+      "step": 13120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7212611233149675e-05,
+      "loss": 2.0216,
+      "step": 13121
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.717149848733481e-05,
+      "loss": 2.0079,
+      "step": 13122
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7130407588089444e-05,
+      "loss": 1.9553,
+      "step": 13123
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7089338537353414e-05,
+      "loss": 2.0079,
+      "step": 13124
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.704829133706511e-05,
+      "loss": 1.9238,
+      "step": 13125
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.70072659891621e-05,
+      "loss": 1.9131,
+      "step": 13126
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.696626249558077e-05,
+      "loss": 2.0426,
+      "step": 13127
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6925280858256746e-05,
+      "loss": 1.8562,
+      "step": 13128
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.688432107912443e-05,
+      "loss": 2.1051,
+      "step": 13129
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.684338316011715e-05,
+      "loss": 2.0254,
+      "step": 13130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.680246710316726e-05,
+      "loss": 2.0251,
+      "step": 13131
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6761572910206156e-05,
+      "loss": 1.8249,
+      "step": 13132
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.672070058316412e-05,
+      "loss": 2.0428,
+      "step": 13133
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6679850123970405e-05,
+      "loss": 2.0234,
+      "step": 13134
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.663902153455312e-05,
+      "loss": 1.9635,
+      "step": 13135
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.659821481683972e-05,
+      "loss": 2.0242,
+      "step": 13136
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6557429972756295e-05,
+      "loss": 1.9147,
+      "step": 13137
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.65166670042279e-05,
+      "loss": 1.8756,
+      "step": 13138
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6475925913178607e-05,
+      "loss": 2.0715,
+      "step": 13139
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.643520670153166e-05,
+      "loss": 1.977,
+      "step": 13140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6394509371208974e-05,
+      "loss": 1.9719,
+      "step": 13141
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.635383392413161e-05,
+      "loss": 2.0633,
+      "step": 13142
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6313180362219425e-05,
+      "loss": 2.0255,
+      "step": 13143
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.627254868739155e-05,
+      "loss": 1.8107,
+      "step": 13144
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.623193890156579e-05,
+      "loss": 2.1141,
+      "step": 13145
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.619135100665893e-05,
+      "loss": 2.0099,
+      "step": 13146
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6150785004587014e-05,
+      "loss": 1.9685,
+      "step": 13147
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6110240897264726e-05,
+      "loss": 2.059,
+      "step": 13148
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.606971868660591e-05,
+      "loss": 1.9258,
+      "step": 13149
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.602921837452311e-05,
+      "loss": 2.1947,
+      "step": 13150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.598873996292834e-05,
+      "loss": 1.9785,
+      "step": 13151
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.594828345373208e-05,
+      "loss": 1.8851,
+      "step": 13152
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.590784884884396e-05,
+      "loss": 1.8636,
+      "step": 13153
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5867436150172626e-05,
+      "loss": 2.0042,
+      "step": 13154
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.582704535962572e-05,
+      "loss": 1.9622,
+      "step": 13155
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5786676479109714e-05,
+      "loss": 2.2636,
+      "step": 13156
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.574632951053009e-05,
+      "loss": 2.1176,
+      "step": 13157
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5706004455791306e-05,
+      "loss": 1.9583,
+      "step": 13158
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.566570131679692e-05,
+      "loss": 2.1205,
+      "step": 13159
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.562542009544922e-05,
+      "loss": 2.0347,
+      "step": 13160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5585160793649594e-05,
+      "loss": 2.1333,
+      "step": 13161
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5544923413298345e-05,
+      "loss": 2.0749,
+      "step": 13162
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.550470795629485e-05,
+      "loss": 2.0039,
+      "step": 13163
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5464514424537295e-05,
+      "loss": 1.9678,
+      "step": 13164
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5424342819923015e-05,
+      "loss": 2.0853,
+      "step": 13165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.538419314434799e-05,
+      "loss": 2.0054,
+      "step": 13166
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5344065399707635e-05,
+      "loss": 2.0506,
+      "step": 13167
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5303959587895896e-05,
+      "loss": 1.9467,
+      "step": 13168
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.526387571080597e-05,
+      "loss": 2.0584,
+      "step": 13169
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.522381377032974e-05,
+      "loss": 2.0761,
+      "step": 13170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5183773768358406e-05,
+      "loss": 1.9373,
+      "step": 13171
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.514375570678191e-05,
+      "loss": 1.9863,
+      "step": 13172
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5103759587489125e-05,
+      "loss": 2.1296,
+      "step": 13173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5063785412367935e-05,
+      "loss": 1.963,
+      "step": 13174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5023833183305375e-05,
+      "loss": 1.9454,
+      "step": 13175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.498390290218711e-05,
+      "loss": 1.8893,
+      "step": 13176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4943994570898085e-05,
+      "loss": 1.9538,
+      "step": 13177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.490410819132189e-05,
+      "loss": 1.7982,
+      "step": 13178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.486424376534142e-05,
+      "loss": 2.0909,
+      "step": 13179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4824401294838326e-05,
+      "loss": 2.0718,
+      "step": 13180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.478458078169328e-05,
+      "loss": 1.9305,
+      "step": 13181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.474478222778571e-05,
+      "loss": 1.934,
+      "step": 13182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.470500563499446e-05,
+      "loss": 2.0471,
+      "step": 13183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.466525100519696e-05,
+      "loss": 2.0704,
+      "step": 13184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.462551834026978e-05,
+      "loss": 1.9577,
+      "step": 13185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.458580764208824e-05,
+      "loss": 2.0845,
+      "step": 13186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4546118912526904e-05,
+      "loss": 2.006,
+      "step": 13187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.450645215345921e-05,
+      "loss": 1.8436,
+      "step": 13188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4466807366757446e-05,
+      "loss": 1.8554,
+      "step": 13189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4427184554292836e-05,
+      "loss": 2.0166,
+      "step": 13190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.438758371793588e-05,
+      "loss": 1.9101,
+      "step": 13191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.434800485955575e-05,
+      "loss": 2.0535,
+      "step": 13192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.430844798102062e-05,
+      "loss": 2.0358,
+      "step": 13193
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4268913084197605e-05,
+      "loss": 2.0157,
+      "step": 13194
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4229400170952995e-05,
+      "loss": 1.922,
+      "step": 13195
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4189909243151854e-05,
+      "loss": 1.9592,
+      "step": 13196
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.415044030265818e-05,
+      "loss": 2.0968,
+      "step": 13197
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.411099335133494e-05,
+      "loss": 1.9333,
+      "step": 13198
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.407156839104425e-05,
+      "loss": 2.1023,
+      "step": 13199
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4032165423647066e-05,
+      "loss": 1.9151,
+      "step": 13200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3992784451003235e-05,
+      "loss": 2.0303,
+      "step": 13201
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.39534254749716e-05,
+      "loss": 1.7945,
+      "step": 13202
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3914088497410064e-05,
+      "loss": 2.1049,
+      "step": 13203
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.387477352017537e-05,
+      "loss": 2.002,
+      "step": 13204
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3835480545123364e-05,
+      "loss": 2.1075,
+      "step": 13205
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3796209574108616e-05,
+      "loss": 1.9851,
+      "step": 13206
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3756960608984976e-05,
+      "loss": 1.9947,
+      "step": 13207
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3717733651604965e-05,
+      "loss": 1.9206,
+      "step": 13208
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.367852870382021e-05,
+      "loss": 2.0404,
+      "step": 13209
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.363934576748123e-05,
+      "loss": 1.9376,
+      "step": 13210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.360018484443772e-05,
+      "loss": 1.9333,
+      "step": 13211
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.356104593653803e-05,
+      "loss": 1.9226,
+      "step": 13212
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.352192904562962e-05,
+      "loss": 1.9484,
+      "step": 13213
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.348283417355885e-05,
+      "loss": 2.0056,
+      "step": 13214
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.34437613221712e-05,
+      "loss": 1.9512,
+      "step": 13215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3404710493310955e-05,
+      "loss": 2.1014,
+      "step": 13216
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.336568168882137e-05,
+      "loss": 1.8765,
+      "step": 13217
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.332667491054464e-05,
+      "loss": 1.9379,
+      "step": 13218
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.328769016032212e-05,
+      "loss": 1.8447,
+      "step": 13219
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.324872743999397e-05,
+      "loss": 1.915,
+      "step": 13220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3209786751399184e-05,
+      "loss": 1.9409,
+      "step": 13221
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.317086809637587e-05,
+      "loss": 1.9981,
+      "step": 13222
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.313197147676128e-05,
+      "loss": 1.9828,
+      "step": 13223
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3093096894391216e-05,
+      "loss": 1.9167,
+      "step": 13224
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.305424435110071e-05,
+      "loss": 2.0024,
+      "step": 13225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.301541384872364e-05,
+      "loss": 1.9449,
+      "step": 13226
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2976605389093027e-05,
+      "loss": 1.9925,
+      "step": 13227
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2937818974040636e-05,
+      "loss": 2.0547,
+      "step": 13228
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.289905460539727e-05,
+      "loss": 2.0539,
+      "step": 13229
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.286031228499259e-05,
+      "loss": 2.064,
+      "step": 13230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.282159201465556e-05,
+      "loss": 1.8428,
+      "step": 13231
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2782893796213744e-05,
+      "loss": 1.9239,
+      "step": 13232
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2744217631493824e-05,
+      "loss": 1.9382,
+      "step": 13233
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.270556352232123e-05,
+      "loss": 1.9036,
+      "step": 13234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2666931470520746e-05,
+      "loss": 2.0052,
+      "step": 13235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2628321477915835e-05,
+      "loss": 1.9116,
+      "step": 13236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.258973354632888e-05,
+      "loss": 1.9229,
+      "step": 13237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.255116767758143e-05,
+      "loss": 1.9804,
+      "step": 13238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.251262387349385e-05,
+      "loss": 1.9955,
+      "step": 13239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.247410213588553e-05,
+      "loss": 1.9552,
+      "step": 13240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.243560246657468e-05,
+      "loss": 2.0077,
+      "step": 13241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2397124867378666e-05,
+      "loss": 2.0248,
+      "step": 13242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.235866934011372e-05,
+      "loss": 2.1572,
+      "step": 13243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2320235886594995e-05,
+      "loss": 2.0131,
+      "step": 13244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2281824508636534e-05,
+      "loss": 2.0219,
+      "step": 13245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2243435208051674e-05,
+      "loss": 2.0299,
+      "step": 13246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2205067986652347e-05,
+      "loss": 1.9909,
+      "step": 13247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.216672284624961e-05,
+      "loss": 2.1005,
+      "step": 13248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.212839978865328e-05,
+      "loss": 2.1055,
+      "step": 13249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.20900988156726e-05,
+      "loss": 1.9641,
+      "step": 13250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.205181992911521e-05,
+      "loss": 1.9618,
+      "step": 13251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.201356313078807e-05,
+      "loss": 2.061,
+      "step": 13252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.19753284224969e-05,
+      "loss": 2.0139,
+      "step": 13253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.193711580604663e-05,
+      "loss": 2.056,
+      "step": 13254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.189892528324084e-05,
+      "loss": 2.0785,
+      "step": 13255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.18607568558823e-05,
+      "loss": 1.9007,
+      "step": 13256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.182261052577251e-05,
+      "loss": 2.0234,
+      "step": 13257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1784486294712255e-05,
+      "loss": 2.0316,
+      "step": 13258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.174638416450093e-05,
+      "loss": 1.9933,
+      "step": 13259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.170830413693715e-05,
+      "loss": 1.9765,
+      "step": 13260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1670246213818254e-05,
+      "loss": 2.0019,
+      "step": 13261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.163221039694081e-05,
+      "loss": 1.9617,
+      "step": 13262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.159419668810015e-05,
+      "loss": 1.9191,
+      "step": 13263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1556205089090575e-05,
+      "loss": 2.0626,
+      "step": 13264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1518235601705306e-05,
+      "loss": 2.0039,
+      "step": 13265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1480288227736685e-05,
+      "loss": 1.9682,
+      "step": 13266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.144236296897601e-05,
+      "loss": 2.0765,
+      "step": 13267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.140445982721324e-05,
+      "loss": 1.9164,
+      "step": 13268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.136657880423754e-05,
+      "loss": 2.044,
+      "step": 13269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.132871990183711e-05,
+      "loss": 2.0503,
+      "step": 13270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1290883121798895e-05,
+      "loss": 1.9882,
+      "step": 13271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.125306846590886e-05,
+      "loss": 1.914,
+      "step": 13272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1215275935951916e-05,
+      "loss": 2.0487,
+      "step": 13273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.117750553371207e-05,
+      "loss": 2.078,
+      "step": 13274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.113975726097207e-05,
+      "loss": 2.0338,
+      "step": 13275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.110203111951376e-05,
+      "loss": 2.1134,
+      "step": 13276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.106432711111779e-05,
+      "loss": 1.9801,
+      "step": 13277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.10266452375641e-05,
+      "loss": 1.9278,
+      "step": 13278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.098898550063123e-05,
+      "loss": 1.9247,
+      "step": 13279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.09513479020968e-05,
+      "loss": 2.0516,
+      "step": 13280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0913732443737355e-05,
+      "loss": 2.0639,
+      "step": 13281
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.087613912732856e-05,
+      "loss": 1.8181,
+      "step": 13282
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.083856795464479e-05,
+      "loss": 1.9466,
+      "step": 13283
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.080101892745957e-05,
+      "loss": 2.072,
+      "step": 13284
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0763492047545195e-05,
+      "loss": 1.8954,
+      "step": 13285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.072598731667314e-05,
+      "loss": 2.0009,
+      "step": 13286
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.068850473661361e-05,
+      "loss": 2.0123,
+      "step": 13287
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.065104430913601e-05,
+      "loss": 2.0365,
+      "step": 13288
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0613606036008324e-05,
+      "loss": 1.9133,
+      "step": 13289
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.057618991899797e-05,
+      "loss": 2.0632,
+      "step": 13290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.053879595987097e-05,
+      "loss": 1.9624,
+      "step": 13291
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0501424160392422e-05,
+      "loss": 1.9694,
+      "step": 13292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0464074522326246e-05,
+      "loss": 1.8827,
+      "step": 13293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0426747047435644e-05,
+      "loss": 2.0109,
+      "step": 13294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0389441737482425e-05,
+      "loss": 1.9475,
+      "step": 13295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.035215859422752e-05,
+      "loss": 2.1696,
+      "step": 13296
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0314897619430626e-05,
+      "loss": 1.9649,
+      "step": 13297
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.027765881485084e-05,
+      "loss": 1.9314,
+      "step": 13298
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0240442182245753e-05,
+      "loss": 1.969,
+      "step": 13299
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0203247723372074e-05,
+      "loss": 1.936,
+      "step": 13300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0166075439985397e-05,
+      "loss": 2.1226,
+      "step": 13301
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0128925333840486e-05,
+      "loss": 1.8605,
+      "step": 13302
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.009179740669088e-05,
+      "loss": 1.9728,
+      "step": 13303
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0054691660289068e-05,
+      "loss": 2.0627,
+      "step": 13304
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0017608096386428e-05,
+      "loss": 2.1787,
+      "step": 13305
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9980546716733613e-05,
+      "loss": 2.0069,
+      "step": 13306
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9943507523079838e-05,
+      "loss": 1.8324,
+      "step": 13307
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9906490517173478e-05,
+      "loss": 2.0998,
+      "step": 13308
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9869495700761805e-05,
+      "loss": 1.9663,
+      "step": 13309
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9832523075591144e-05,
+      "loss": 2.2158,
+      "step": 13310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.97955726434066e-05,
+      "loss": 2.0172,
+      "step": 13311
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9758644405952328e-05,
+      "loss": 2.099,
+      "step": 13312
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9721738364971384e-05,
+      "loss": 2.0774,
+      "step": 13313
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9684854522205985e-05,
+      "loss": 1.9665,
+      "step": 13314
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9647992879397012e-05,
+      "loss": 1.9441,
+      "step": 13315
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9611153438284412e-05,
+      "loss": 2.031,
+      "step": 13316
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.957433620060701e-05,
+      "loss": 1.873,
+      "step": 13317
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9537541168102865e-05,
+      "loss": 2.0877,
+      "step": 13318
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.95007683425087e-05,
+      "loss": 2.035,
+      "step": 13319
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.946401772556023e-05,
+      "loss": 1.9471,
+      "step": 13320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.942728931899219e-05,
+      "loss": 1.8115,
+      "step": 13321
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9390583124538295e-05,
+      "loss": 1.957,
+      "step": 13322
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9353899143931105e-05,
+      "loss": 2.1681,
+      "step": 13323
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.931723737890224e-05,
+      "loss": 1.989,
+      "step": 13324
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9280597831182088e-05,
+      "loss": 2.1054,
+      "step": 13325
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.924398050250027e-05,
+      "loss": 1.9709,
+      "step": 13326
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9207385394585118e-05,
+      "loss": 2.18,
+      "step": 13327
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9170812509164147e-05,
+      "loss": 2.0247,
+      "step": 13328
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9134261847963474e-05,
+      "loss": 1.9925,
+      "step": 13329
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9097733412708548e-05,
+      "loss": 1.9757,
+      "step": 13330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9061227205123498e-05,
+      "loss": 1.9258,
+      "step": 13331
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.902474322693155e-05,
+      "loss": 1.9585,
+      "step": 13332
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8988281479854772e-05,
+      "loss": 1.9826,
+      "step": 13333
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8951841965614346e-05,
+      "loss": 2.0574,
+      "step": 13334
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8915424685930225e-05,
+      "loss": 1.9233,
+      "step": 13335
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.887902964252137e-05,
+      "loss": 1.974,
+      "step": 13336
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8842656837105684e-05,
+      "loss": 2.0443,
+      "step": 13337
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8806306271400183e-05,
+      "loss": 1.9465,
+      "step": 13338
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8769977947120608e-05,
+      "loss": 1.967,
+      "step": 13339
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8733671865981802e-05,
+      "loss": 2.0373,
+      "step": 13340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.869738802969729e-05,
+      "loss": 1.7558,
+      "step": 13341
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8661126439980033e-05,
+      "loss": 2.0946,
+      "step": 13342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8624887098541498e-05,
+      "loss": 2.0387,
+      "step": 13343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.858867000709231e-05,
+      "loss": 1.8027,
+      "step": 13344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.855247516734194e-05,
+      "loss": 1.8571,
+      "step": 13345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8516302580998966e-05,
+      "loss": 2.1241,
+      "step": 13346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8480152249770796e-05,
+      "loss": 2.0098,
+      "step": 13347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8444024175363738e-05,
+      "loss": 2.0286,
+      "step": 13348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8407918359483143e-05,
+      "loss": 2.0842,
+      "step": 13349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.837183480383332e-05,
+      "loss": 2.0534,
+      "step": 13350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8335773510117512e-05,
+      "loss": 1.8261,
+      "step": 13351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8299734480037863e-05,
+      "loss": 2.0693,
+      "step": 13352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8263717715295456e-05,
+      "loss": 2.0718,
+      "step": 13353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8227723217590483e-05,
+      "loss": 2.1145,
+      "step": 13354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8191750988621867e-05,
+      "loss": 1.9815,
+      "step": 13355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8155801030087635e-05,
+      "loss": 2.0083,
+      "step": 13356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.81198733436846e-05,
+      "loss": 1.8994,
+      "step": 13357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8083967931108788e-05,
+      "loss": 1.9511,
+      "step": 13358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8048084794054905e-05,
+      "loss": 2.0634,
+      "step": 13359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8012223934216817e-05,
+      "loss": 1.8968,
+      "step": 13360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.797638535328706e-05,
+      "loss": 1.9965,
+      "step": 13361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.79405690529575e-05,
+      "loss": 2.0668,
+      "step": 13362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.790477503491867e-05,
+      "loss": 1.9928,
+      "step": 13363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7869003300860118e-05,
+      "loss": 1.9863,
+      "step": 13364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7833253852470263e-05,
+      "loss": 1.9069,
+      "step": 13365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.779752669143676e-05,
+      "loss": 1.8224,
+      "step": 13366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7761821819445863e-05,
+      "loss": 2.0862,
+      "step": 13367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7726139238183014e-05,
+      "loss": 1.9416,
+      "step": 13368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7690478949332408e-05,
+      "loss": 1.7981,
+      "step": 13369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7654840954577377e-05,
+      "loss": 2.0313,
+      "step": 13370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7619225255600123e-05,
+      "loss": 2.0012,
+      "step": 13371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7583631854081748e-05,
+      "loss": 2.0269,
+      "step": 13372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7548060751702298e-05,
+      "loss": 1.9824,
+      "step": 13373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7512511950140928e-05,
+      "loss": 2.0391,
+      "step": 13374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.747698545107552e-05,
+      "loss": 2.0858,
+      "step": 13375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7441481256183122e-05,
+      "loss": 2.0763,
+      "step": 13376
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.740599936713939e-05,
+      "loss": 1.8396,
+      "step": 13377
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.737053978561943e-05,
+      "loss": 1.8938,
+      "step": 13378
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733510251329685e-05,
+      "loss": 1.9564,
+      "step": 13379
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.729968755184442e-05,
+      "loss": 2.1401,
+      "step": 13380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.726429490293375e-05,
+      "loss": 2.1167,
+      "step": 13381
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7228924568235504e-05,
+      "loss": 2.0033,
+      "step": 13382
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7193576549419286e-05,
+      "loss": 2.1258,
+      "step": 13383
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7158250848153542e-05,
+      "loss": 1.9254,
+      "step": 13384
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7122947466105662e-05,
+      "loss": 2.0068,
+      "step": 13385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7087666404942202e-05,
+      "loss": 2.0179,
+      "step": 13386
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7052407666328437e-05,
+      "loss": 2.1024,
+      "step": 13387
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7017171251928652e-05,
+      "loss": 2.0576,
+      "step": 13388
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.698195716340601e-05,
+      "loss": 2.0113,
+      "step": 13389
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6946765402422858e-05,
+      "loss": 2.1261,
+      "step": 13390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6911595970640246e-05,
+      "loss": 1.9611,
+      "step": 13391
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6876448869718296e-05,
+      "loss": 2.0384,
+      "step": 13392
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6841324101315846e-05,
+      "loss": 2.1674,
+      "step": 13393
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6806221667091123e-05,
+      "loss": 2.1144,
+      "step": 13394
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.677114156870092e-05,
+      "loss": 1.9825,
+      "step": 13395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.673608380780107e-05,
+      "loss": 1.9716,
+      "step": 13396
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6701048386046367e-05,
+      "loss": 1.9713,
+      "step": 13397
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6666035305090708e-05,
+      "loss": 1.9725,
+      "step": 13398
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.663104456658666e-05,
+      "loss": 2.1302,
+      "step": 13399
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.65960761721859e-05,
+      "loss": 1.9868,
+      "step": 13400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6561130123539e-05,
+      "loss": 1.9537,
+      "step": 13401
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.652620642229553e-05,
+      "loss": 2.0325,
+      "step": 13402
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6491305070104e-05,
+      "loss": 1.8295,
+      "step": 13403
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.645642606861176e-05,
+      "loss": 1.8668,
+      "step": 13404
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6421569419465108e-05,
+      "loss": 1.9844,
+      "step": 13405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6386735124309558e-05,
+      "loss": 2.09,
+      "step": 13406
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.635192318478924e-05,
+      "loss": 1.9544,
+      "step": 13407
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.631713360254734e-05,
+      "loss": 2.2196,
+      "step": 13408
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6282366379226098e-05,
+      "loss": 1.9909,
+      "step": 13409
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.624762151646659e-05,
+      "loss": 1.895,
+      "step": 13410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6212899015908788e-05,
+      "loss": 1.9341,
+      "step": 13411
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.617819887919165e-05,
+      "loss": 1.8641,
+      "step": 13412
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.614352110795326e-05,
+      "loss": 2.0691,
+      "step": 13413
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6108865703830363e-05,
+      "loss": 1.921,
+      "step": 13414
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6074232668458764e-05,
+      "loss": 2.0653,
+      "step": 13415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6039622003473263e-05,
+      "loss": 2.0178,
+      "step": 13416
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6005033710507553e-05,
+      "loss": 2.0152,
+      "step": 13417
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.597046779119433e-05,
+      "loss": 2.0899,
+      "step": 13418
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.593592424716512e-05,
+      "loss": 1.9501,
+      "step": 13419
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.59014030800504e-05,
+      "loss": 1.9982,
+      "step": 13420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5866904291479808e-05,
+      "loss": 1.9444,
+      "step": 13421
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.583242788308171e-05,
+      "loss": 1.9295,
+      "step": 13422
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5797973856483415e-05,
+      "loss": 1.9197,
+      "step": 13423
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.576354221331123e-05,
+      "loss": 2.1719,
+      "step": 13424
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.572913295519047e-05,
+      "loss": 2.1657,
+      "step": 13425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.569474608374528e-05,
+      "loss": 2.0474,
+      "step": 13426
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.566038160059886e-05,
+      "loss": 2.1314,
+      "step": 13427
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5626039507373188e-05,
+      "loss": 1.9571,
+      "step": 13428
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5591719805689418e-05,
+      "loss": 1.9021,
+      "step": 13429
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.555742249716747e-05,
+      "loss": 2.0603,
+      "step": 13430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.552314758342622e-05,
+      "loss": 2.0262,
+      "step": 13431
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5488895066083483e-05,
+      "loss": 2.0582,
+      "step": 13432
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.545466494675619e-05,
+      "loss": 2.0414,
+      "step": 13433
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5420457227060047e-05,
+      "loss": 2.0722,
+      "step": 13434
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.538627190860965e-05,
+      "loss": 1.8935,
+      "step": 13435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.535210899301865e-05,
+      "loss": 1.8623,
+      "step": 13436
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5317968481899656e-05,
+      "loss": 1.9591,
+      "step": 13437
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5283850376864203e-05,
+      "loss": 2.0806,
+      "step": 13438
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5249754679522672e-05,
+      "loss": 2.0031,
+      "step": 13439
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5215681391484447e-05,
+      "loss": 1.8483,
+      "step": 13440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5181630514357957e-05,
+      "loss": 2.133,
+      "step": 13441
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5147602049750472e-05,
+      "loss": 2.0726,
+      "step": 13442
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.511359599926816e-05,
+      "loss": 2.1189,
+      "step": 13443
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5079612364516115e-05,
+      "loss": 1.9508,
+      "step": 13444
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5045651147098614e-05,
+      "loss": 1.9752,
+      "step": 13445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5011712348618598e-05,
+      "loss": 2.0751,
+      "step": 13446
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4977795970678063e-05,
+      "loss": 1.8397,
+      "step": 13447
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.494390201487795e-05,
+      "loss": 2.0472,
+      "step": 13448
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4910030482818147e-05,
+      "loss": 1.9768,
+      "step": 13449
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4876181376097484e-05,
+      "loss": 1.9481,
+      "step": 13450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4842354696313685e-05,
+      "loss": 1.8925,
+      "step": 13451
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.480855044506336e-05,
+      "loss": 1.8948,
+      "step": 13452
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4774768623942346e-05,
+      "loss": 1.9219,
+      "step": 13453
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4741009234545087e-05,
+      "loss": 1.9281,
+      "step": 13454
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4707272278465144e-05,
+      "loss": 2.0066,
+      "step": 13455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.467355775729485e-05,
+      "loss": 1.9908,
+      "step": 13456
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4639865672625884e-05,
+      "loss": 2.0541,
+      "step": 13457
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4606196026048354e-05,
+      "loss": 1.9264,
+      "step": 13458
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4572548819151662e-05,
+      "loss": 1.9719,
+      "step": 13459
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4538924053523926e-05,
+      "loss": 2.0541,
+      "step": 13460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.450532173075243e-05,
+      "loss": 2.0464,
+      "step": 13461
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4471741852423235e-05,
+      "loss": 1.9826,
+      "step": 13462
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4438184420121358e-05,
+      "loss": 1.9855,
+      "step": 13463
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4404649435430803e-05,
+      "loss": 1.9888,
+      "step": 13464
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4371136899934532e-05,
+      "loss": 2.1551,
+      "step": 13465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4337646815214386e-05,
+      "loss": 1.9291,
+      "step": 13466
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4304179182851218e-05,
+      "loss": 2.0221,
+      "step": 13467
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4270734004424644e-05,
+      "loss": 1.9756,
+      "step": 13468
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.423731128151352e-05,
+      "loss": 2.0173,
+      "step": 13469
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4203911015695413e-05,
+      "loss": 2.0667,
+      "step": 13470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4170533208546842e-05,
+      "loss": 1.9771,
+      "step": 13471
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4137177861643324e-05,
+      "loss": 1.9888,
+      "step": 13472
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.410384497655943e-05,
+      "loss": 2.0324,
+      "step": 13473
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4070534554868406e-05,
+      "loss": 1.8965,
+      "step": 13474
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.403724659814266e-05,
+      "loss": 2.1129,
+      "step": 13475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4003981107953377e-05,
+      "loss": 2.1542,
+      "step": 13476
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3970738085870856e-05,
+      "loss": 1.8762,
+      "step": 13477
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3937517533464236e-05,
+      "loss": 2.0937,
+      "step": 13478
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3904319452301536e-05,
+      "loss": 2.0202,
+      "step": 13479
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.387114384394978e-05,
+      "loss": 1.9197,
+      "step": 13480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3837990709975055e-05,
+      "loss": 1.8908,
+      "step": 13481
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3804860051942158e-05,
+      "loss": 2.0225,
+      "step": 13482
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3771751871415005e-05,
+      "loss": 1.8251,
+      "step": 13483
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3738666169956236e-05,
+      "loss": 1.9841,
+      "step": 13484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3705602949127713e-05,
+      "loss": 1.9925,
+      "step": 13485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.367256221049008e-05,
+      "loss": 2.1244,
+      "step": 13486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3639543955602917e-05,
+      "loss": 2.0159,
+      "step": 13487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.36065481860247e-05,
+      "loss": 1.9829,
+      "step": 13488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3573574903312967e-05,
+      "loss": 1.9557,
+      "step": 13489
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3540624109024188e-05,
+      "loss": 2.1532,
+      "step": 13490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3507695804713624e-05,
+      "loss": 2.058,
+      "step": 13491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3474789991935474e-05,
+      "loss": 1.9485,
+      "step": 13492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.344190667224322e-05,
+      "loss": 2.0877,
+      "step": 13493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.340904584718889e-05,
+      "loss": 1.9341,
+      "step": 13494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3376207518323588e-05,
+      "loss": 2.0574,
+      "step": 13495
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.334339168719729e-05,
+      "loss": 1.9276,
+      "step": 13496
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.331059835535915e-05,
+      "loss": 1.9307,
+      "step": 13497
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3277827524356976e-05,
+      "loss": 1.9835,
+      "step": 13498
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3245079195737596e-05,
+      "loss": 1.9182,
+      "step": 13499
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3212353371046936e-05,
+      "loss": 1.9987,
+      "step": 13500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.317965005182965e-05,
+      "loss": 1.9145,
+      "step": 13501
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.314696923962939e-05,
+      "loss": 2.0625,
+      "step": 13502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3114310935988758e-05,
+      "loss": 1.9473,
+      "step": 13503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3081675142449354e-05,
+      "loss": 2.0046,
+      "step": 13504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3049061860551667e-05,
+      "loss": 1.8942,
+      "step": 13505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3016471091835078e-05,
+      "loss": 2.1366,
+      "step": 13506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.298390283783791e-05,
+      "loss": 1.9463,
+      "step": 13507
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2951357100097547e-05,
+      "loss": 2.0261,
+      "step": 13508
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2918833880150203e-05,
+      "loss": 2.0688,
+      "step": 13509
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2886333179531037e-05,
+      "loss": 1.9671,
+      "step": 13510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2853854999774047e-05,
+      "loss": 1.8381,
+      "step": 13511
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.282139934241251e-05,
+      "loss": 1.8505,
+      "step": 13512
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2788966208978246e-05,
+      "loss": 2.1258,
+      "step": 13513
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2756555601002204e-05,
+      "loss": 1.9199,
+      "step": 13514
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.272416752001416e-05,
+      "loss": 1.98,
+      "step": 13515
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2691801967543057e-05,
+      "loss": 2.0095,
+      "step": 13516
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.265945894511656e-05,
+      "loss": 1.9953,
+      "step": 13517
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2627138454261343e-05,
+      "loss": 1.9845,
+      "step": 13518
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.25948404965029e-05,
+      "loss": 1.9562,
+      "step": 13519
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.256256507336596e-05,
+      "loss": 1.9423,
+      "step": 13520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2530312186373913e-05,
+      "loss": 1.984,
+      "step": 13521
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.24980818370491e-05,
+      "loss": 2.1104,
+      "step": 13522
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.246587402691286e-05,
+      "loss": 1.9323,
+      "step": 13523
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2433688757485636e-05,
+      "loss": 1.9236,
+      "step": 13524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2401526030286558e-05,
+      "loss": 1.9587,
+      "step": 13525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2369385846833734e-05,
+      "loss": 2.1937,
+      "step": 13526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2337268208644235e-05,
+      "loss": 2.0029,
+      "step": 13527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2305173117234233e-05,
+      "loss": 2.026,
+      "step": 13528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2273100574118578e-05,
+      "loss": 2.0611,
+      "step": 13529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2241050580811216e-05,
+      "loss": 2.148,
+      "step": 13530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.220902313882489e-05,
+      "loss": 1.9457,
+      "step": 13531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2177018249671442e-05,
+      "loss": 2.0349,
+      "step": 13532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2145035914861666e-05,
+      "loss": 2.0423,
+      "step": 13533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2113076135905018e-05,
+      "loss": 1.97,
+      "step": 13534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.208113891431013e-05,
+      "loss": 2.0533,
+      "step": 13535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.204922425158462e-05,
+      "loss": 2.1363,
+      "step": 13536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2017332149234846e-05,
+      "loss": 1.9761,
+      "step": 13537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.198546260876616e-05,
+      "loss": 2.1834,
+      "step": 13538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.195361563168291e-05,
+      "loss": 1.8483,
+      "step": 13539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1921791219488398e-05,
+      "loss": 2.0468,
+      "step": 13540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1889989373684748e-05,
+      "loss": 1.8172,
+      "step": 13541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.18582100957731e-05,
+      "loss": 1.8658,
+      "step": 13542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1826453387253475e-05,
+      "loss": 1.9148,
+      "step": 13543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1794719249624894e-05,
+      "loss": 1.9946,
+      "step": 13544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.176300768438533e-05,
+      "loss": 2.0304,
+      "step": 13545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1731318693031575e-05,
+      "loss": 1.9797,
+      "step": 13546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.169965227705939e-05,
+      "loss": 1.8176,
+      "step": 13547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.166800843796357e-05,
+      "loss": 1.9452,
+      "step": 13548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1636387177237815e-05,
+      "loss": 2.0859,
+      "step": 13549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.160478849637465e-05,
+      "loss": 2.1288,
+      "step": 13550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.157321239686555e-05,
+      "loss": 1.9362,
+      "step": 13551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.15416588802011e-05,
+      "loss": 1.8798,
+      "step": 13552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.151012794787066e-05,
+      "loss": 1.9381,
+      "step": 13553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1478619601362547e-05,
+      "loss": 1.9383,
+      "step": 13554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1447133842163958e-05,
+      "loss": 1.8998,
+      "step": 13555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1415670671761255e-05,
+      "loss": 1.9693,
+      "step": 13556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1384230091639478e-05,
+      "loss": 1.9525,
+      "step": 13557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1352812103282715e-05,
+      "loss": 1.7828,
+      "step": 13558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1321416708173835e-05,
+      "loss": 2.0462,
+      "step": 13559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1290043907794987e-05,
+      "loss": 2.1108,
+      "step": 13560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1258693703626984e-05,
+      "loss": 1.9916,
+      "step": 13561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1227366097149536e-05,
+      "loss": 1.9106,
+      "step": 13562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1196061089841344e-05,
+      "loss": 2.0442,
+      "step": 13563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1164778683180286e-05,
+      "loss": 1.9335,
+      "step": 13564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1133518878642786e-05,
+      "loss": 2.0663,
+      "step": 13565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1102281677704506e-05,
+      "loss": 1.8766,
+      "step": 13566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1071067081839702e-05,
+      "loss": 1.9993,
+      "step": 13567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1039875092521975e-05,
+      "loss": 2.0486,
+      "step": 13568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1008705711223653e-05,
+      "loss": 2.0702,
+      "step": 13569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0977558939415885e-05,
+      "loss": 1.9644,
+      "step": 13570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.094643477856889e-05,
+      "loss": 2.0883,
+      "step": 13571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0915333230151933e-05,
+      "loss": 2.0249,
+      "step": 13572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0884254295632954e-05,
+      "loss": 1.9119,
+      "step": 13573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0853197976478942e-05,
+      "loss": 1.9572,
+      "step": 13574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.082216427415584e-05,
+      "loss": 2.009,
+      "step": 13575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.079115319012864e-05,
+      "loss": 1.9348,
+      "step": 13576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0760164725861008e-05,
+      "loss": 1.9594,
+      "step": 13577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.072919888281566e-05,
+      "loss": 2.0361,
+      "step": 13578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0698255662454258e-05,
+      "loss": 2.0817,
+      "step": 13579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.066733506623747e-05,
+      "loss": 1.9649,
+      "step": 13580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0636437095624794e-05,
+      "loss": 1.9832,
+      "step": 13581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0605561752074677e-05,
+      "loss": 2.1226,
+      "step": 13582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0574709037044392e-05,
+      "loss": 2.0118,
+      "step": 13583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0543878951990448e-05,
+      "loss": 2.0498,
+      "step": 13584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.051307149836801e-05,
+      "loss": 1.9732,
+      "step": 13585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.04822866776313e-05,
+      "loss": 2.0496,
+      "step": 13586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0451524491233276e-05,
+      "loss": 2.1133,
+      "step": 13587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0420784940626157e-05,
+      "loss": 2.0071,
+      "step": 13588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.03900680272609e-05,
+      "loss": 1.966,
+      "step": 13589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0359373752587286e-05,
+      "loss": 1.9871,
+      "step": 13590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0328702118054376e-05,
+      "loss": 1.9016,
+      "step": 13591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.029805312510974e-05,
+      "loss": 1.9049,
+      "step": 13592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.026742677520016e-05,
+      "loss": 1.8892,
+      "step": 13593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0236823069771204e-05,
+      "loss": 2.0736,
+      "step": 13594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0206242010267606e-05,
+      "loss": 1.9644,
+      "step": 13595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.017568359813271e-05,
+      "loss": 2.0919,
+      "step": 13596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.014514783480903e-05,
+      "loss": 2.0725,
+      "step": 13597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0114634721737745e-05,
+      "loss": 2.1016,
+      "step": 13598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0084144260359425e-05,
+      "loss": 1.9838,
+      "step": 13599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.005367645211309e-05,
+      "loss": 1.9706,
+      "step": 13600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.002323129843697e-05,
+      "loss": 1.9051,
+      "step": 13601
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9992808800768036e-05,
+      "loss": 1.9834,
+      "step": 13602
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9962408960542466e-05,
+      "loss": 2.1408,
+      "step": 13603
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9932031779195115e-05,
+      "loss": 1.965,
+      "step": 13604
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.990167725815989e-05,
+      "loss": 2.0555,
+      "step": 13605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.987134539886948e-05,
+      "loss": 2.0619,
+      "step": 13606
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.984103620275579e-05,
+      "loss": 2.0796,
+      "step": 13607
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9810749671249352e-05,
+      "loss": 2.0207,
+      "step": 13608
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9780485805779848e-05,
+      "loss": 2.0114,
+      "step": 13609
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.975024460777569e-05,
+      "loss": 2.0358,
+      "step": 13610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9720026078664467e-05,
+      "loss": 1.9589,
+      "step": 13611
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9689830219872475e-05,
+      "loss": 1.9584,
+      "step": 13612
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9659657032825074e-05,
+      "loss": 2.0256,
+      "step": 13613
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.962950651894646e-05,
+      "loss": 2.0566,
+      "step": 13614
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.959937867965983e-05,
+      "loss": 1.9092,
+      "step": 13615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.956927351638732e-05,
+      "loss": 2.0003,
+      "step": 13616
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9539191030549964e-05,
+      "loss": 2.0124,
+      "step": 13617
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9509131223567622e-05,
+      "loss": 2.029,
+      "step": 13618
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9479094096859273e-05,
+      "loss": 2.0666,
+      "step": 13619
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9449079651842782e-05,
+      "loss": 2.0535,
+      "step": 13620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9419087889934793e-05,
+      "loss": 2.0316,
+      "step": 13621
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.938911881255101e-05,
+      "loss": 2.0296,
+      "step": 13622
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9359172421106075e-05,
+      "loss": 1.9386,
+      "step": 13623
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9329248717013582e-05,
+      "loss": 1.9766,
+      "step": 13624
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.92993477016859e-05,
+      "loss": 2.0382,
+      "step": 13625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.92694693765344e-05,
+      "loss": 2.0577,
+      "step": 13626
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9239613742969563e-05,
+      "loss": 1.9825,
+      "step": 13627
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.920978080240049e-05,
+      "loss": 1.9503,
+      "step": 13628
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9179970556235438e-05,
+      "loss": 1.9726,
+      "step": 13629
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9150183005881504e-05,
+      "loss": 1.9835,
+      "step": 13630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9120418152744733e-05,
+      "loss": 2.0443,
+      "step": 13631
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9090675998230112e-05,
+      "loss": 1.9631,
+      "step": 13632
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9060956543741514e-05,
+      "loss": 2.0084,
+      "step": 13633
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.903125979068171e-05,
+      "loss": 2.0971,
+      "step": 13634
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9001585740452577e-05,
+      "loss": 2.03,
+      "step": 13635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8971934394454716e-05,
+      "loss": 2.0703,
+      "step": 13636
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8942305754087785e-05,
+      "loss": 2.0586,
+      "step": 13637
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8912699820750223e-05,
+      "loss": 1.9844,
+      "step": 13638
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8883116595839634e-05,
+      "loss": 2.0577,
+      "step": 13639
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.885355608075234e-05,
+      "loss": 2.0564,
+      "step": 13640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.882401827688368e-05,
+      "loss": 2.1123,
+      "step": 13641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8794503185627866e-05,
+      "loss": 2.0171,
+      "step": 13642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.876501080837817e-05,
+      "loss": 2.1451,
+      "step": 13643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8735541146526647e-05,
+      "loss": 2.0917,
+      "step": 13644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8706094201464297e-05,
+      "loss": 2.0083,
+      "step": 13645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.867666997458112e-05,
+      "loss": 1.9545,
+      "step": 13646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8647268467266053e-05,
+      "loss": 1.9419,
+      "step": 13647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8617889680906828e-05,
+      "loss": 2.1261,
+      "step": 13648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.858853361689028e-05,
+      "loss": 2.1887,
+      "step": 13649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8559200276601963e-05,
+      "loss": 2.0269,
+      "step": 13650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8529889661426602e-05,
+      "loss": 2.1156,
+      "step": 13651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.850060177274765e-05,
+      "loss": 1.9325,
+      "step": 13652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.847133661194761e-05,
+      "loss": 1.9687,
+      "step": 13653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8442094180407763e-05,
+      "loss": 2.073,
+      "step": 13654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8412874479508567e-05,
+      "loss": 1.8958,
+      "step": 13655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8383677510629193e-05,
+      "loss": 1.9267,
+      "step": 13656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8354503275147814e-05,
+      "loss": 1.9277,
+      "step": 13657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.832535177444139e-05,
+      "loss": 2.1104,
+      "step": 13658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8296223009886147e-05,
+      "loss": 1.9391,
+      "step": 13659
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.826711698285699e-05,
+      "loss": 2.0382,
+      "step": 13660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8238033694727706e-05,
+      "loss": 1.9614,
+      "step": 13661
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.820897314687103e-05,
+      "loss": 2.0553,
+      "step": 13662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.817993534065887e-05,
+      "loss": 2.1017,
+      "step": 13663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.815092027746179e-05,
+      "loss": 2.1072,
+      "step": 13664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8121927958649366e-05,
+      "loss": 2.1062,
+      "step": 13665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.809295838559005e-05,
+      "loss": 2.0409,
+      "step": 13666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8064011559651374e-05,
+      "loss": 1.9683,
+      "step": 13667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8035087482199676e-05,
+      "loss": 1.8777,
+      "step": 13668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.80061861546002e-05,
+      "loss": 2.0834,
+      "step": 13669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7977307578217084e-05,
+      "loss": 2.0948,
+      "step": 13670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.794845175441362e-05,
+      "loss": 1.9875,
+      "step": 13671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.791961868455183e-05,
+      "loss": 2.0789,
+      "step": 13672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7890808369992628e-05,
+      "loss": 2.2087,
+      "step": 13673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7862020812095868e-05,
+      "loss": 2.1178,
+      "step": 13674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.783325601222058e-05,
+      "loss": 1.8756,
+      "step": 13675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7804513971724445e-05,
+      "loss": 1.9863,
+      "step": 13676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7775794691964163e-05,
+      "loss": 2.021,
+      "step": 13677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7747098174295206e-05,
+      "loss": 2.0665,
+      "step": 13678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.771842442007232e-05,
+      "loss": 1.9052,
+      "step": 13679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7689773430648925e-05,
+      "loss": 2.035,
+      "step": 13680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7661145207377327e-05,
+      "loss": 2.0813,
+      "step": 13681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.763253975160889e-05,
+      "loss": 1.948,
+      "step": 13682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7603957064693922e-05,
+      "loss": 1.9868,
+      "step": 13683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7575397147981507e-05,
+      "loss": 1.9092,
+      "step": 13684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7546860002819742e-05,
+      "loss": 1.9204,
+      "step": 13685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7518345630555654e-05,
+      "loss": 1.7514,
+      "step": 13686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.748985403253528e-05,
+      "loss": 1.987,
+      "step": 13687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7461385210103376e-05,
+      "loss": 2.0552,
+      "step": 13688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.74329391646037e-05,
+      "loss": 2.0145,
+      "step": 13689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7404515897379126e-05,
+      "loss": 1.9592,
+      "step": 13690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7376115409771188e-05,
+      "loss": 1.9407,
+      "step": 13691
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7347737703120482e-05,
+      "loss": 1.8621,
+      "step": 13692
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.731938277876649e-05,
+      "loss": 2.019,
+      "step": 13693
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7291050638047647e-05,
+      "loss": 1.9923,
+      "step": 13694
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.726274128230132e-05,
+      "loss": 2.1098,
+      "step": 13695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7234454712863725e-05,
+      "loss": 2.0514,
+      "step": 13696
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.720619093107001e-05,
+      "loss": 1.9855,
+      "step": 13697
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.717794993825439e-05,
+      "loss": 1.8188,
+      "step": 13698
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.714973173574985e-05,
+      "loss": 1.9301,
+      "step": 13699
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7121536324888442e-05,
+      "loss": 2.0673,
+      "step": 13700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7093363707000874e-05,
+      "loss": 2.1021,
+      "step": 13701
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7065213883417142e-05,
+      "loss": 2.0579,
+      "step": 13702
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.70370868554659e-05,
+      "loss": 2.0892,
+      "step": 13703
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.700898262447481e-05,
+      "loss": 2.0056,
+      "step": 13704
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6980901191770426e-05,
+      "loss": 2.0243,
+      "step": 13705
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6952842558678293e-05,
+      "loss": 2.0556,
+      "step": 13706
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.692480672652291e-05,
+      "loss": 2.1065,
+      "step": 13707
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.68967936966275e-05,
+      "loss": 1.988,
+      "step": 13708
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6868803470314385e-05,
+      "loss": 2.003,
+      "step": 13709
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.684083604890485e-05,
+      "loss": 1.9377,
+      "step": 13710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6812891433719e-05,
+      "loss": 2.0724,
+      "step": 13711
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6784969626075785e-05,
+      "loss": 1.9462,
+      "step": 13712
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6757070627293204e-05,
+      "loss": 1.9297,
+      "step": 13713
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6729194438688256e-05,
+      "loss": 1.9107,
+      "step": 13714
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.670134106157667e-05,
+      "loss": 1.7611,
+      "step": 13715
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6673510497273226e-05,
+      "loss": 2.1961,
+      "step": 13716
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.664570274709154e-05,
+      "loss": 1.9817,
+      "step": 13717
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6617917812344284e-05,
+      "loss": 2.0387,
+      "step": 13718
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.659015569434291e-05,
+      "loss": 1.9874,
+      "step": 13719
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6562416394397873e-05,
+      "loss": 1.9046,
+      "step": 13720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6534699913818507e-05,
+      "loss": 2.1129,
+      "step": 13721
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.650700625391316e-05,
+      "loss": 1.9616,
+      "step": 13722
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6479335415989006e-05,
+      "loss": 2.0389,
+      "step": 13723
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.645168740135211e-05,
+      "loss": 2.0615,
+      "step": 13724
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.64240622113076e-05,
+      "loss": 2.089,
+      "step": 13725
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6396459847159375e-05,
+      "loss": 1.9979,
+      "step": 13726
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6368880310210442e-05,
+      "loss": 1.8612,
+      "step": 13727
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6341323601762548e-05,
+      "loss": 1.9964,
+      "step": 13728
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6313789723116367e-05,
+      "loss": 1.8238,
+      "step": 13729
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6286278675571696e-05,
+      "loss": 1.9173,
+      "step": 13730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6258790460427053e-05,
+      "loss": 1.951,
+      "step": 13731
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6231325078979954e-05,
+      "loss": 2.1299,
+      "step": 13732
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6203882532526747e-05,
+      "loss": 2.0797,
+      "step": 13733
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6176462822362902e-05,
+      "loss": 2.107,
+      "step": 13734
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.614906594978266e-05,
+      "loss": 2.0469,
+      "step": 13735
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6121691916079208e-05,
+      "loss": 2.028,
+      "step": 13736
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.609434072254462e-05,
+      "loss": 2.0397,
+      "step": 13737
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.606701237046998e-05,
+      "loss": 1.9798,
+      "step": 13738
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.603970686114531e-05,
+      "loss": 2.042,
+      "step": 13739
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6012424195859356e-05,
+      "loss": 2.0283,
+      "step": 13740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5985164375899973e-05,
+      "loss": 2.0402,
+      "step": 13741
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5957927402553972e-05,
+      "loss": 2.0875,
+      "step": 13742
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5930713277106932e-05,
+      "loss": 1.913,
+      "step": 13743
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.590352200084344e-05,
+      "loss": 2.1563,
+      "step": 13744
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5876353575046908e-05,
+      "loss": 2.0924,
+      "step": 13745
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.584920800099987e-05,
+      "loss": 1.9945,
+      "step": 13746
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5822085279983577e-05,
+      "loss": 1.9457,
+      "step": 13747
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5794985413278338e-05,
+      "loss": 2.0345,
+      "step": 13748
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5767908402163244e-05,
+      "loss": 2.1099,
+      "step": 13749
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5740854247916492e-05,
+      "loss": 1.9664,
+      "step": 13750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571382295181506e-05,
+      "loss": 2.1143,
+      "step": 13751
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5686814515134873e-05,
+      "loss": 1.8961,
+      "step": 13752
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5659828939150744e-05,
+      "loss": 1.875,
+      "step": 13753
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5632866225136598e-05,
+      "loss": 1.9678,
+      "step": 13754
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5605926374365032e-05,
+      "loss": 2.0675,
+      "step": 13755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.557900938810769e-05,
+      "loss": 2.0016,
+      "step": 13756
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.555211526763506e-05,
+      "loss": 1.8801,
+      "step": 13757
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5525244014216732e-05,
+      "loss": 1.9234,
+      "step": 13758
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5498395629120976e-05,
+      "loss": 2.125,
+      "step": 13759
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5471570113615162e-05,
+      "loss": 1.7657,
+      "step": 13760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5444767468965447e-05,
+      "loss": 2.007,
+      "step": 13761
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5417987696437098e-05,
+      "loss": 1.9884,
+      "step": 13762
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5391230797294097e-05,
+      "loss": 2.1118,
+      "step": 13763
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.536449677279944e-05,
+      "loss": 1.9852,
+      "step": 13764
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5337785624215004e-05,
+      "loss": 2.1576,
+      "step": 13765
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5311097352801672e-05,
+      "loss": 1.8655,
+      "step": 13766
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5284431959819213e-05,
+      "loss": 2.0238,
+      "step": 13767
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.525778944652617e-05,
+      "loss": 1.9781,
+      "step": 13768
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5231169814180267e-05,
+      "loss": 2.0422,
+      "step": 13769
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5204573064037942e-05,
+      "loss": 2.0504,
+      "step": 13770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5177999197354685e-05,
+      "loss": 1.8966,
+      "step": 13771
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5151448215384722e-05,
+      "loss": 1.8468,
+      "step": 13772
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5124920119381436e-05,
+      "loss": 2.1221,
+      "step": 13773
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.509841491059699e-05,
+      "loss": 1.9935,
+      "step": 13774
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5071932590282444e-05,
+      "loss": 2.057,
+      "step": 13775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.504547315968785e-05,
+      "loss": 2.0145,
+      "step": 13776
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.501903662006221e-05,
+      "loss": 1.974,
+      "step": 13777
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4992622972653303e-05,
+      "loss": 2.054,
+      "step": 13778
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4966232218707908e-05,
+      "loss": 2.0111,
+      "step": 13779
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4939864359471866e-05,
+      "loss": 1.9666,
+      "step": 13780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4913519396189567e-05,
+      "loss": 2.0816,
+      "step": 13781
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4887197330104796e-05,
+      "loss": 1.9765,
+      "step": 13782
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4860898162459947e-05,
+      "loss": 1.893,
+      "step": 13783
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4834621894496302e-05,
+      "loss": 1.9063,
+      "step": 13784
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4808368527454207e-05,
+      "loss": 1.9932,
+      "step": 13785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4782138062572947e-05,
+      "loss": 2.0294,
+      "step": 13786
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4755930501090642e-05,
+      "loss": 1.9549,
+      "step": 13787
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4729745844244302e-05,
+      "loss": 1.8841,
+      "step": 13788
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4703584093269829e-05,
+      "loss": 2.0803,
+      "step": 13789
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4677445249402288e-05,
+      "loss": 2.0,
+      "step": 13790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4651329313875417e-05,
+      "loss": 2.0788,
+      "step": 13791
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.462523628792195e-05,
+      "loss": 1.9719,
+      "step": 13792
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4599166172773458e-05,
+      "loss": 2.0185,
+      "step": 13793
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.457311896966068e-05,
+      "loss": 2.0489,
+      "step": 13794
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4547094679812967e-05,
+      "loss": 1.8092,
+      "step": 13795
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4521093304458776e-05,
+      "loss": 1.9557,
+      "step": 13796
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4495114844825408e-05,
+      "loss": 1.9452,
+      "step": 13797
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4469159302139156e-05,
+      "loss": 2.1124,
+      "step": 13798
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4443226677625154e-05,
+      "loss": 2.0502,
+      "step": 13799
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4417316972507421e-05,
+      "loss": 2.1163,
+      "step": 13800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4391430188008981e-05,
+      "loss": 1.9635,
+      "step": 13801
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4365566325351854e-05,
+      "loss": 1.8774,
+      "step": 13802
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4339725385756786e-05,
+      "loss": 1.9505,
+      "step": 13803
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4313907370443579e-05,
+      "loss": 2.0589,
+      "step": 13804
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4288112280630761e-05,
+      "loss": 2.0458,
+      "step": 13805
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4262340117536077e-05,
+      "loss": 1.9469,
+      "step": 13806
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4236590882376e-05,
+      "loss": 1.9617,
+      "step": 13807
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.421086457636589e-05,
+      "loss": 1.9872,
+      "step": 13808
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.418516120072011e-05,
+      "loss": 1.9945,
+      "step": 13809
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4159480756651966e-05,
+      "loss": 2.1815,
+      "step": 13810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.41338232453736e-05,
+      "loss": 2.0186,
+      "step": 13811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4108188668096155e-05,
+      "loss": 2.0236,
+      "step": 13812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4082577026029497e-05,
+      "loss": 1.8955,
+      "step": 13813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4056988320382713e-05,
+      "loss": 2.0781,
+      "step": 13814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4031422552363615e-05,
+      "loss": 1.9719,
+      "step": 13815
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.400587972317896e-05,
+      "loss": 1.9664,
+      "step": 13816
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.398035983403434e-05,
+      "loss": 2.0367,
+      "step": 13817
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3954862886134457e-05,
+      "loss": 2.0126,
+      "step": 13818
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3929388880682791e-05,
+      "loss": 1.9904,
+      "step": 13819
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3903937818881774e-05,
+      "loss": 1.9191,
+      "step": 13820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3878509701932718e-05,
+      "loss": 1.8847,
+      "step": 13821
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3853104531035943e-05,
+      "loss": 2.0209,
+      "step": 13822
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.38277223073906e-05,
+      "loss": 2.0103,
+      "step": 13823
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802363032194842e-05,
+      "loss": 2.0226,
+      "step": 13824
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3777026706645601e-05,
+      "loss": 2.1196,
+      "step": 13825
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.375171333193892e-05,
+      "loss": 1.8467,
+      "step": 13826
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.372642290926951e-05,
+      "loss": 2.0036,
+      "step": 13827
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3701155439831249e-05,
+      "loss": 1.9503,
+      "step": 13828
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3675910924816736e-05,
+      "loss": 1.8895,
+      "step": 13829
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.365068936541769e-05,
+      "loss": 2.0153,
+      "step": 13830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3625490762824489e-05,
+      "loss": 1.9732,
+      "step": 13831
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.360031511822668e-05,
+      "loss": 2.059,
+      "step": 13832
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3575162432812482e-05,
+      "loss": 1.999,
+      "step": 13833
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3550032707769277e-05,
+      "loss": 2.0362,
+      "step": 13834
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.352492594428323e-05,
+      "loss": 2.0015,
+      "step": 13835
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3499842143539386e-05,
+      "loss": 1.9966,
+      "step": 13836
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.347478130672175e-05,
+      "loss": 1.9607,
+      "step": 13837
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3449743435013318e-05,
+      "loss": 1.9519,
+      "step": 13838
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3424728529595919e-05,
+      "loss": 2.1661,
+      "step": 13839
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.339973659165028e-05,
+      "loss": 1.8997,
+      "step": 13840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3374767622356066e-05,
+      "loss": 2.0044,
+      "step": 13841
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3349821622891944e-05,
+      "loss": 2.0306,
+      "step": 13842
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3324898594435364e-05,
+      "loss": 2.0028,
+      "step": 13843
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3299998538162772e-05,
+      "loss": 2.0812,
+      "step": 13844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.327512145524945e-05,
+      "loss": 2.0527,
+      "step": 13845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3250267346869738e-05,
+      "loss": 2.0871,
+      "step": 13846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3225436214196806e-05,
+      "loss": 2.0935,
+      "step": 13847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3200628058402609e-05,
+      "loss": 2.022,
+      "step": 13848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3175842880658317e-05,
+      "loss": 1.8967,
+      "step": 13849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.315108068213383e-05,
+      "loss": 2.0305,
+      "step": 13850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3126341463997881e-05,
+      "loss": 1.9422,
+      "step": 13851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.31016252274182e-05,
+      "loss": 1.9108,
+      "step": 13852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3076931973561579e-05,
+      "loss": 2.014,
+      "step": 13853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3052261703593527e-05,
+      "loss": 2.0984,
+      "step": 13854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3027614418678557e-05,
+      "loss": 1.9202,
+      "step": 13855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3002990119980018e-05,
+      "loss": 1.965,
+      "step": 13856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2978388808660313e-05,
+      "loss": 1.9628,
+      "step": 13857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2953810485880679e-05,
+      "loss": 1.894,
+      "step": 13858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.292925515280119e-05,
+      "loss": 2.0158,
+      "step": 13859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2904722810580971e-05,
+      "loss": 2.1769,
+      "step": 13860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2880213460377987e-05,
+      "loss": 1.978,
+      "step": 13861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2855727103349146e-05,
+      "loss": 2.0161,
+      "step": 13862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.28312637406503e-05,
+      "loss": 1.9964,
+      "step": 13863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2806823373436027e-05,
+      "loss": 1.99,
+      "step": 13864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.278240600286018e-05,
+      "loss": 1.9394,
+      "step": 13865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2758011630075172e-05,
+      "loss": 2.0644,
+      "step": 13866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2733640256232582e-05,
+      "loss": 2.0331,
+      "step": 13867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2709291882482598e-05,
+      "loss": 1.9864,
+      "step": 13868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2684966509974694e-05,
+      "loss": 1.9184,
+      "step": 13869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2660664139857114e-05,
+      "loss": 1.9749,
+      "step": 13870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2636384773276833e-05,
+      "loss": 2.0245,
+      "step": 13871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.261212841137993e-05,
+      "loss": 1.9624,
+      "step": 13872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2587895055311494e-05,
+      "loss": 2.0791,
+      "step": 13873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2563684706215272e-05,
+      "loss": 1.94,
+      "step": 13874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2539497365234021e-05,
+      "loss": 1.957,
+      "step": 13875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2515333033509546e-05,
+      "loss": 2.1816,
+      "step": 13876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2491191712182382e-05,
+      "loss": 2.0138,
+      "step": 13877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2467073402392115e-05,
+      "loss": 1.9139,
+      "step": 13878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2442978105277114e-05,
+      "loss": 2.0324,
+      "step": 13879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2418905821974746e-05,
+      "loss": 1.9214,
+      "step": 13880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2394856553621326e-05,
+      "loss": 1.8874,
+      "step": 13881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2370830301352055e-05,
+      "loss": 2.0433,
+      "step": 13882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.234682706630097e-05,
+      "loss": 1.987,
+      "step": 13883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2322846849601e-05,
+      "loss": 2.0121,
+      "step": 13884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2298889652384237e-05,
+      "loss": 1.9373,
+      "step": 13885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2274955475781446e-05,
+      "loss": 2.1931,
+      "step": 13886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2251044320922332e-05,
+      "loss": 1.9346,
+      "step": 13887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.222715618893555e-05,
+      "loss": 1.9102,
+      "step": 13888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2203291080948753e-05,
+      "loss": 1.9396,
+      "step": 13889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2179448998088428e-05,
+      "loss": 1.8462,
+      "step": 13890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2155629941479896e-05,
+      "loss": 1.922,
+      "step": 13891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.213183391224748e-05,
+      "loss": 2.0335,
+      "step": 13892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2108060911514507e-05,
+      "loss": 1.9763,
+      "step": 13893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.208431094040302e-05,
+      "loss": 1.9564,
+      "step": 13894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2060584000034124e-05,
+      "loss": 1.8071,
+      "step": 13895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2036880091527758e-05,
+      "loss": 2.15,
+      "step": 13896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2013199216002801e-05,
+      "loss": 1.9666,
+      "step": 13897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1989541374577029e-05,
+      "loss": 2.0052,
+      "step": 13898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1965906568367214e-05,
+      "loss": 2.211,
+      "step": 13899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1942294798488851e-05,
+      "loss": 1.9923,
+      "step": 13900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1918706066056606e-05,
+      "loss": 2.0645,
+      "step": 13901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1895140372183866e-05,
+      "loss": 1.9188,
+      "step": 13902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1871597717983017e-05,
+      "loss": 1.9032,
+      "step": 13903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1848078104565175e-05,
+      "loss": 1.9991,
+      "step": 13904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1824581533040724e-05,
+      "loss": 2.1355,
+      "step": 13905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1801108004518668e-05,
+      "loss": 1.913,
+      "step": 13906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.177765752010701e-05,
+      "loss": 2.0957,
+      "step": 13907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1754230080912587e-05,
+      "loss": 1.8161,
+      "step": 13908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1730825688041402e-05,
+      "loss": 2.0876,
+      "step": 13909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1707444342598072e-05,
+      "loss": 1.9149,
+      "step": 13910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1684086045686327e-05,
+      "loss": 1.9598,
+      "step": 13911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1660750798408614e-05,
+      "loss": 2.0452,
+      "step": 13912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1637438601866501e-05,
+      "loss": 1.9891,
+      "step": 13913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1614149457160384e-05,
+      "loss": 1.8744,
+      "step": 13914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1590883365389549e-05,
+      "loss": 2.0897,
+      "step": 13915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1567640327652173e-05,
+      "loss": 2.0099,
+      "step": 13916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1544420345045436e-05,
+      "loss": 1.9908,
+      "step": 13917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1521223418665294e-05,
+      "loss": 1.9741,
+      "step": 13918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1498049549606814e-05,
+      "loss": 1.9247,
+      "step": 13919
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1474898738963736e-05,
+      "loss": 1.9929,
+      "step": 13920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1451770987828902e-05,
+      "loss": 1.9761,
+      "step": 13921
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1428666297293999e-05,
+      "loss": 2.0503,
+      "step": 13922
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1405584668449597e-05,
+      "loss": 1.9201,
+      "step": 13923
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.138252610238516e-05,
+      "loss": 2.0236,
+      "step": 13924
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1359490600189204e-05,
+      "loss": 1.9337,
+      "step": 13925
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1336478162948972e-05,
+      "loss": 1.9763,
+      "step": 13926
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1313488791750759e-05,
+      "loss": 2.0251,
+      "step": 13927
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.12905224876797e-05,
+      "loss": 2.029,
+      "step": 13928
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1267579251819815e-05,
+      "loss": 2.0885,
+      "step": 13929
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1244659085254183e-05,
+      "loss": 2.1354,
+      "step": 13930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.122176198906455e-05,
+      "loss": 1.979,
+      "step": 13931
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1198887964331772e-05,
+      "loss": 1.8519,
+      "step": 13932
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1176037012135598e-05,
+      "loss": 1.9681,
+      "step": 13933
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1153209133554664e-05,
+      "loss": 1.941,
+      "step": 13934
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1130404329666389e-05,
+      "loss": 2.0259,
+      "step": 13935
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1107622601547241e-05,
+      "loss": 1.8635,
+      "step": 13936
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1084863950272638e-05,
+      "loss": 2.0049,
+      "step": 13937
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1062128376916836e-05,
+      "loss": 1.9907,
+      "step": 13938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1039415882552861e-05,
+      "loss": 1.8562,
+      "step": 13939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1016726468253024e-05,
+      "loss": 2.104,
+      "step": 13940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0994060135088136e-05,
+      "loss": 2.1163,
+      "step": 13941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0971416884128227e-05,
+      "loss": 1.9861,
+      "step": 13942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0948796716441944e-05,
+      "loss": 1.9041,
+      "step": 13943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0926199633097156e-05,
+      "loss": 1.9758,
+      "step": 13944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0903625635160509e-05,
+      "loss": 2.1193,
+      "step": 13945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0881074723697481e-05,
+      "loss": 1.9987,
+      "step": 13946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0858546899772448e-05,
+      "loss": 2.0317,
+      "step": 13947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0836042164448945e-05,
+      "loss": 2.0326,
+      "step": 13948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0813560518789179e-05,
+      "loss": 1.9396,
+      "step": 13949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.07911019638543e-05,
+      "loss": 2.153,
+      "step": 13950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0768666500704405e-05,
+      "loss": 2.0756,
+      "step": 13951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0746254130398537e-05,
+      "loss": 2.0542,
+      "step": 13952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0723864853994625e-05,
+      "loss": 1.877,
+      "step": 13953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0701498672549492e-05,
+      "loss": 2.098,
+      "step": 13954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0679155587118739e-05,
+      "loss": 1.9229,
+      "step": 13955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0656835598757186e-05,
+      "loss": 2.2023,
+      "step": 13956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0634538708518327e-05,
+      "loss": 2.1462,
+      "step": 13957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.061226491745465e-05,
+      "loss": 1.9365,
+      "step": 13958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0590014226617428e-05,
+      "loss": 2.0455,
+      "step": 13959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0567786637057097e-05,
+      "loss": 1.937,
+      "step": 13960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0545582149822709e-05,
+      "loss": 2.028,
+      "step": 13961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0523400765962477e-05,
+      "loss": 2.0267,
+      "step": 13962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0501242486523287e-05,
+      "loss": 1.9443,
+      "step": 13963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0479107312551195e-05,
+      "loss": 2.0404,
+      "step": 13964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0456995245091028e-05,
+      "loss": 1.9704,
+      "step": 13965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0434906285186397e-05,
+      "loss": 1.8377,
+      "step": 13966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0412840433880022e-05,
+      "loss": 2.0029,
+      "step": 13967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0390797692213516e-05,
+      "loss": 1.9873,
+      "step": 13968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0368778061227268e-05,
+      "loss": 1.9035,
+      "step": 13969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0346781541960726e-05,
+      "loss": 1.9889,
+      "step": 13970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0324808135452057e-05,
+      "loss": 2.1355,
+      "step": 13971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.03028578427386e-05,
+      "loss": 2.0143,
+      "step": 13972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0280930664856413e-05,
+      "loss": 1.9494,
+      "step": 13973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0259026602840449e-05,
+      "loss": 2.0523,
+      "step": 13974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0237145657724655e-05,
+      "loss": 1.9528,
+      "step": 13975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.021528783054193e-05,
+      "loss": 2.0512,
+      "step": 13976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0193453122324004e-05,
+      "loss": 2.0882,
+      "step": 13977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.017164153410144e-05,
+      "loss": 1.9771,
+      "step": 13978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0149853066903803e-05,
+      "loss": 2.0849,
+      "step": 13979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0128087721759605e-05,
+      "loss": 1.9885,
+      "step": 13980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0106345499696245e-05,
+      "loss": 2.0312,
+      "step": 13981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.008462640173996e-05,
+      "loss": 2.024,
+      "step": 13982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0062930428915928e-05,
+      "loss": 2.0936,
+      "step": 13983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0041257582248331e-05,
+      "loss": 1.9986,
+      "step": 13984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0019607862760071e-05,
+      "loss": 2.0974,
+      "step": 13985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.997981271473166e-06,
+      "loss": 1.9364,
+      "step": 13986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.9763778094083e-06,
+      "loss": 1.9462,
+      "step": 13987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.954797477585375e-06,
+      "loss": 1.9263,
+      "step": 13988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.933240277022914e-06,
+      "loss": 1.8689,
+      "step": 13989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.911706208738547e-06,
+      "loss": 2.0705,
+      "step": 13990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.890195273748626e-06,
+      "loss": 2.0731,
+      "step": 13991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.868707473068616e-06,
+      "loss": 2.0434,
+      "step": 13992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.847242807712764e-06,
+      "loss": 2.0464,
+      "step": 13993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.825801278694258e-06,
+      "loss": 1.9893,
+      "step": 13994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.804382887025121e-06,
+      "loss": 1.9425,
+      "step": 13995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.782987633716433e-06,
+      "loss": 2.1707,
+      "step": 13996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.761615519778056e-06,
+      "loss": 1.8937,
+      "step": 13997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.740266546218846e-06,
+      "loss": 1.8876,
+      "step": 13998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.718940714046442e-06,
+      "loss": 2.0731,
+      "step": 13999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.69763802426754e-06,
+      "loss": 2.0332,
+      "step": 14000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.67635847788767e-06,
+      "loss": 1.9703,
+      "step": 14001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.655102075911303e-06,
+      "loss": 2.0212,
+      "step": 14002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.63386881934164e-06,
+      "loss": 1.9106,
+      "step": 14003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.612658709181154e-06,
+      "loss": 1.8459,
+      "step": 14004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.59147174643088e-06,
+      "loss": 2.1045,
+      "step": 14005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.570307932090904e-06,
+      "loss": 2.0471,
+      "step": 14006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.549167267160209e-06,
+      "loss": 1.9907,
+      "step": 14007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.528049752636714e-06,
+      "loss": 1.9464,
+      "step": 14008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.506955389517236e-06,
+      "loss": 2.0067,
+      "step": 14009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.485884178797366e-06,
+      "loss": 2.0664,
+      "step": 14010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.464836121471809e-06,
+      "loss": 2.1209,
+      "step": 14011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.443811218534049e-06,
+      "loss": 1.9523,
+      "step": 14012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.422809470976568e-06,
+      "loss": 2.0297,
+      "step": 14013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.40183087979063e-06,
+      "loss": 2.0962,
+      "step": 14014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.380875445966442e-06,
+      "loss": 2.1223,
+      "step": 14015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.359943170493213e-06,
+      "loss": 1.9124,
+      "step": 14016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.339034054358986e-06,
+      "loss": 2.0214,
+      "step": 14017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.31814809855075e-06,
+      "loss": 1.9107,
+      "step": 14018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.297285304054271e-06,
+      "loss": 2.1402,
+      "step": 14019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.276445671854428e-06,
+      "loss": 1.9689,
+      "step": 14020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.255629202934823e-06,
+      "loss": 2.1141,
+      "step": 14021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.234835898278116e-06,
+      "loss": 1.9126,
+      "step": 14022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.214065758865686e-06,
+      "loss": 1.9091,
+      "step": 14023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.193318785678029e-06,
+      "loss": 2.0,
+      "step": 14024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.172594979694415e-06,
+      "loss": 2.2203,
+      "step": 14025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.151894341893063e-06,
+      "loss": 2.0346,
+      "step": 14026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.13121687325108e-06,
+      "loss": 1.9671,
+      "step": 14027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.110562574744518e-06,
+      "loss": 1.9431,
+      "step": 14028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08993144734832e-06,
+      "loss": 2.0088,
+      "step": 14029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.069323492036207e-06,
+      "loss": 1.9982,
+      "step": 14030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.048738709781068e-06,
+      "loss": 2.0432,
+      "step": 14031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.028177101554458e-06,
+      "loss": 1.9761,
+      "step": 14032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.00763866832699e-06,
+      "loss": 2.086,
+      "step": 14033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.987123411068054e-06,
+      "loss": 1.9365,
+      "step": 14034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.9666313307461e-06,
+      "loss": 2.003,
+      "step": 14035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.946162428328353e-06,
+      "loss": 1.8152,
+      "step": 14036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.92571670478104e-06,
+      "loss": 1.9913,
+      "step": 14037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.905294161069112e-06,
+      "loss": 1.9482,
+      "step": 14038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.88489479815674e-06,
+      "loss": 2.0274,
+      "step": 14039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.864518617006767e-06,
+      "loss": 1.8784,
+      "step": 14040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.844165618580924e-06,
+      "loss": 2.0113,
+      "step": 14041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.823835803839885e-06,
+      "loss": 2.0059,
+      "step": 14042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.803529173743441e-06,
+      "loss": 1.8656,
+      "step": 14043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.78324572924999e-06,
+      "loss": 1.9467,
+      "step": 14044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.762985471316987e-06,
+      "loss": 2.1611,
+      "step": 14045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.742748400900668e-06,
+      "loss": 1.9838,
+      "step": 14046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.722534518956437e-06,
+      "loss": 1.9403,
+      "step": 14047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.702343826438363e-06,
+      "loss": 2.0923,
+      "step": 14048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.682176324299461e-06,
+      "loss": 2.0941,
+      "step": 14049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.662032013491694e-06,
+      "loss": 2.0833,
+      "step": 14050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.641910894965966e-06,
+      "loss": 1.9483,
+      "step": 14051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.62181296967196e-06,
+      "loss": 1.9363,
+      "step": 14052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.60173823855842e-06,
+      "loss": 2.0874,
+      "step": 14053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.581686702572867e-06,
+      "loss": 1.8477,
+      "step": 14054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.561658362661817e-06,
+      "loss": 2.0625,
+      "step": 14055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.541653219770684e-06,
+      "loss": 2.1195,
+      "step": 14056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.521671274843656e-06,
+      "loss": 1.975,
+      "step": 14057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.50171252882398e-06,
+      "loss": 1.9478,
+      "step": 14058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.481776982653788e-06,
+      "loss": 2.0814,
+      "step": 14059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.461864637274053e-06,
+      "loss": 1.9399,
+      "step": 14060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.441975493624687e-06,
+      "loss": 1.9515,
+      "step": 14061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.42210955264444e-06,
+      "loss": 2.0242,
+      "step": 14062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.402266815271176e-06,
+      "loss": 2.0917,
+      "step": 14063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.382447282441418e-06,
+      "loss": 2.0722,
+      "step": 14064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.3626509550907e-06,
+      "loss": 1.9328,
+      "step": 14065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.342877834153384e-06,
+      "loss": 1.9676,
+      "step": 14066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.323127920562944e-06,
+      "loss": 1.9898,
+      "step": 14067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.303401215251582e-06,
+      "loss": 2.0844,
+      "step": 14068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.28369771915044e-06,
+      "loss": 2.019,
+      "step": 14069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.264017433189497e-06,
+      "loss": 1.9827,
+      "step": 14070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.244360358297787e-06,
+      "loss": 2.1199,
+      "step": 14071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.224726495403124e-06,
+      "loss": 1.8689,
+      "step": 14072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.205115845432377e-06,
+      "loss": 2.0827,
+      "step": 14073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.185528409311028e-06,
+      "loss": 2.0182,
+      "step": 14074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.165964187963781e-06,
+      "loss": 1.9625,
+      "step": 14075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.146423182314067e-06,
+      "loss": 1.8065,
+      "step": 14076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.126905393284311e-06,
+      "loss": 1.9657,
+      "step": 14077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.10741082179578e-06,
+      "loss": 1.8654,
+      "step": 14078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.087939468768624e-06,
+      "loss": 1.8544,
+      "step": 14079
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.068491335121941e-06,
+      "loss": 2.0117,
+      "step": 14080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.049066421773831e-06,
+      "loss": 1.9908,
+      "step": 14081
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.029664729641006e-06,
+      "loss": 2.0684,
+      "step": 14082
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.010286259639454e-06,
+      "loss": 2.0181,
+      "step": 14083
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.990931012683778e-06,
+      "loss": 1.9879,
+      "step": 14084
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.971598989687634e-06,
+      "loss": 2.086,
+      "step": 14085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.95229019156346e-06,
+      "loss": 1.9395,
+      "step": 14086
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.933004619222805e-06,
+      "loss": 2.056,
+      "step": 14087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.913742273575885e-06,
+      "loss": 1.79,
+      "step": 14088
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.894503155532029e-06,
+      "loss": 1.9639,
+      "step": 14089
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.875287265999232e-06,
+      "loss": 1.9244,
+      "step": 14090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.856094605884601e-06,
+      "loss": 1.9158,
+      "step": 14091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.836925176094134e-06,
+      "loss": 2.1458,
+      "step": 14092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.817778977532608e-06,
+      "loss": 1.9041,
+      "step": 14093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.798656011103689e-06,
+      "loss": 2.0871,
+      "step": 14094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.77955627771021e-06,
+      "loss": 2.0385,
+      "step": 14095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.760479778253559e-06,
+      "loss": 1.914,
+      "step": 14096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.741426513634297e-06,
+      "loss": 2.0556,
+      "step": 14097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.722396484751704e-06,
+      "loss": 2.0287,
+      "step": 14098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.70338969250406e-06,
+      "loss": 2.0764,
+      "step": 14099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.684406137788591e-06,
+      "loss": 1.9088,
+      "step": 14100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.665445821501305e-06,
+      "loss": 2.1012,
+      "step": 14101
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.64650874453715e-06,
+      "loss": 1.779,
+      "step": 14102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.627594907790081e-06,
+      "loss": 1.9995,
+      "step": 14103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.60870431215277e-06,
+      "loss": 2.0587,
+      "step": 14104
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.5898369585170045e-06,
+      "loss": 1.9846,
+      "step": 14105
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.5709928477732945e-06,
+      "loss": 1.9382,
+      "step": 14106
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.552171980811151e-06,
+      "loss": 1.8778,
+      "step": 14107
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.533374358518974e-06,
+      "loss": 2.1323,
+      "step": 14108
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.514599981784054e-06,
+      "loss": 1.9949,
+      "step": 14109
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.495848851492515e-06,
+      "loss": 1.9628,
+      "step": 14110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.4771209685295385e-06,
+      "loss": 1.9646,
+      "step": 14111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.458416333779083e-06,
+      "loss": 1.9854,
+      "step": 14112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.439734948124055e-06,
+      "loss": 2.082,
+      "step": 14113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.4210768124463035e-06,
+      "loss": 1.9999,
+      "step": 14114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.402441927626458e-06,
+      "loss": 1.9374,
+      "step": 14115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.383830294544203e-06,
+      "loss": 1.8971,
+      "step": 14116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.365241914078003e-06,
+      "loss": 2.0719,
+      "step": 14117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.346676787105211e-06,
+      "loss": 1.9737,
+      "step": 14118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.3281349145022935e-06,
+      "loss": 1.9712,
+      "step": 14119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.309616297144384e-06,
+      "loss": 1.9952,
+      "step": 14120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.291120935905615e-06,
+      "loss": 1.9195,
+      "step": 14121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.272648831658957e-06,
+      "loss": 2.1459,
+      "step": 14122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.254199985276433e-06,
+      "loss": 2.1349,
+      "step": 14123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.235774397628847e-06,
+      "loss": 1.9984,
+      "step": 14124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.217372069585837e-06,
+      "loss": 1.997,
+      "step": 14125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.198993002016152e-06,
+      "loss": 2.0029,
+      "step": 14126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.180637195787265e-06,
+      "loss": 2.01,
+      "step": 14127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.16230465176565e-06,
+      "loss": 2.0011,
+      "step": 14128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.143995370816558e-06,
+      "loss": 2.0657,
+      "step": 14129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.125709353804355e-06,
+      "loss": 1.9274,
+      "step": 14130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.107446601592127e-06,
+      "loss": 2.0521,
+      "step": 14131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.0892071150419064e-06,
+      "loss": 2.0744,
+      "step": 14132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.070990895014562e-06,
+      "loss": 1.9513,
+      "step": 14133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.052797942370126e-06,
+      "loss": 2.0002,
+      "step": 14134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.03462825796719e-06,
+      "loss": 2.04,
+      "step": 14135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.016481842663458e-06,
+      "loss": 2.1311,
+      "step": 14136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.998358697315466e-06,
+      "loss": 2.0529,
+      "step": 14137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.980258822778751e-06,
+      "loss": 2.0671,
+      "step": 14138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.9621822199075755e-06,
+      "loss": 2.0956,
+      "step": 14139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.9441288895552015e-06,
+      "loss": 1.9499,
+      "step": 14140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.926098832573835e-06,
+      "loss": 2.0209,
+      "step": 14141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.908092049814518e-06,
+      "loss": 2.0139,
+      "step": 14142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.890108542127182e-06,
+      "loss": 1.9776,
+      "step": 14143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.872148310360704e-06,
+      "loss": 1.8182,
+      "step": 14144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.85421135536285e-06,
+      "loss": 1.9976,
+      "step": 14145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.83629767798033e-06,
+      "loss": 2.0225,
+      "step": 14146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.818407279058691e-06,
+      "loss": 2.0854,
+      "step": 14147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.80054015944237e-06,
+      "loss": 2.0247,
+      "step": 14148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.782696319974746e-06,
+      "loss": 1.8723,
+      "step": 14149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.764875761498091e-06,
+      "loss": 2.1066,
+      "step": 14150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.747078484853564e-06,
+      "loss": 1.9771,
+      "step": 14151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.729304490881272e-06,
+      "loss": 2.0451,
+      "step": 14152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.7115537804201545e-06,
+      "loss": 2.0246,
+      "step": 14153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.693826354308097e-06,
+      "loss": 1.9721,
+      "step": 14154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.67612221338193e-06,
+      "loss": 2.0429,
+      "step": 14155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.658441358477263e-06,
+      "loss": 1.9324,
+      "step": 14156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.640783790428595e-06,
+      "loss": 2.0713,
+      "step": 14157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.623149510069593e-06,
+      "loss": 2.1159,
+      "step": 14158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.605538518232534e-06,
+      "loss": 2.0776,
+      "step": 14159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.5879508157487e-06,
+      "loss": 1.9427,
+      "step": 14160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.570386403448203e-06,
+      "loss": 1.9779,
+      "step": 14161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.552845282160269e-06,
+      "loss": 2.0476,
+      "step": 14162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.535327452712792e-06,
+      "loss": 1.9259,
+      "step": 14163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.517832915932664e-06,
+      "loss": 2.0756,
+      "step": 14164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.500361672645616e-06,
+      "loss": 1.8094,
+      "step": 14165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.482913723676431e-06,
+      "loss": 2.0322,
+      "step": 14166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.465489069848674e-06,
+      "loss": 2.0659,
+      "step": 14167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.448087711984796e-06,
+      "loss": 1.9798,
+      "step": 14168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.430709650906142e-06,
+      "loss": 2.0269,
+      "step": 14169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.413354887433054e-06,
+      "loss": 1.9746,
+      "step": 14170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.396023422384711e-06,
+      "loss": 2.0244,
+      "step": 14171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.3787152565792374e-06,
+      "loss": 2.0349,
+      "step": 14172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.361430390833478e-06,
+      "loss": 1.9225,
+      "step": 14173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.3441688259634475e-06,
+      "loss": 2.1319,
+      "step": 14174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.326930562783939e-06,
+      "loss": 1.8876,
+      "step": 14175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.309715602108579e-06,
+      "loss": 1.9764,
+      "step": 14176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.292523944749884e-06,
+      "loss": 2.1406,
+      "step": 14177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.275355591519538e-06,
+      "loss": 2.0426,
+      "step": 14178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.258210543227783e-06,
+      "loss": 1.9651,
+      "step": 14179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.241088800683914e-06,
+      "loss": 1.9341,
+      "step": 14180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.223990364696119e-06,
+      "loss": 2.031,
+      "step": 14181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.206915236071531e-06,
+      "loss": 2.0862,
+      "step": 14182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.189863415616115e-06,
+      "loss": 2.0137,
+      "step": 14183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.172834904134727e-06,
+      "loss": 2.0034,
+      "step": 14184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.15582970243117e-06,
+      "loss": 2.0762,
+      "step": 14185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.138847811308135e-06,
+      "loss": 1.9519,
+      "step": 14186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.121889231567201e-06,
+      "loss": 2.0742,
+      "step": 14187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.104953964008897e-06,
+      "loss": 1.9904,
+      "step": 14188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.088042009432471e-06,
+      "loss": 2.0758,
+      "step": 14189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.071153368636339e-06,
+      "loss": 1.966,
+      "step": 14190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.054288042417699e-06,
+      "loss": 2.1693,
+      "step": 14191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.037446031572525e-06,
+      "loss": 2.0799,
+      "step": 14192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.020627336895845e-06,
+      "loss": 2.1438,
+      "step": 14193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.003831959181582e-06,
+      "loss": 1.9933,
+      "step": 14194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.987059899222435e-06,
+      "loss": 1.947,
+      "step": 14195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.97031115781016e-06,
+      "loss": 2.0402,
+      "step": 14196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9535857357352895e-06,
+      "loss": 2.128,
+      "step": 14197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9368836337873045e-06,
+      "loss": 2.0114,
+      "step": 14198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.920204852754629e-06,
+      "loss": 1.9743,
+      "step": 14199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.903549393424523e-06,
+      "loss": 1.9235,
+      "step": 14200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.88691725658308e-06,
+      "loss": 1.9782,
+      "step": 14201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.870308443015504e-06,
+      "loss": 2.0326,
+      "step": 14202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8537229535056695e-06,
+      "loss": 2.1113,
+      "step": 14203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.837160788836504e-06,
+      "loss": 2.0549,
+      "step": 14204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.820621949789717e-06,
+      "loss": 2.0174,
+      "step": 14205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.804106437146073e-06,
+      "loss": 1.9773,
+      "step": 14206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.787614251685114e-06,
+      "loss": 2.0357,
+      "step": 14207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.771145394185273e-06,
+      "loss": 2.005,
+      "step": 14208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.754699865423874e-06,
+      "loss": 1.9091,
+      "step": 14209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.738277666177294e-06,
+      "loss": 1.9884,
+      "step": 14210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.721878797220636e-06,
+      "loss": 1.9321,
+      "step": 14211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.705503259327949e-06,
+      "loss": 2.0041,
+      "step": 14212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.689151053272279e-06,
+      "loss": 2.0308,
+      "step": 14213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.672822179825454e-06,
+      "loss": 1.9937,
+      "step": 14214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.656516639758136e-06,
+      "loss": 1.9779,
+      "step": 14215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.640234433840097e-06,
+      "loss": 1.9344,
+      "step": 14216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.623975562839834e-06,
+      "loss": 2.0066,
+      "step": 14217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.607740027524899e-06,
+      "loss": 1.9382,
+      "step": 14218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5915278286615136e-06,
+      "loss": 1.9814,
+      "step": 14219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.575338967015009e-06,
+      "loss": 1.9733,
+      "step": 14220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.559173443349496e-06,
+      "loss": 1.9712,
+      "step": 14221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5430312584280886e-06,
+      "loss": 1.9406,
+      "step": 14222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.526912413012675e-06,
+      "loss": 2.1463,
+      "step": 14223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.510816907864092e-06,
+      "loss": 1.9529,
+      "step": 14224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.494744743742175e-06,
+      "loss": 1.9425,
+      "step": 14225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.478695921405485e-06,
+      "loss": 1.8306,
+      "step": 14226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.462670441611583e-06,
+      "loss": 2.176,
+      "step": 14227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.446668305116864e-06,
+      "loss": 1.9047,
+      "step": 14228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.430689512676779e-06,
+      "loss": 1.9569,
+      "step": 14229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.414734065045446e-06,
+      "loss": 2.1317,
+      "step": 14230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.398801962976041e-06,
+      "loss": 2.0074,
+      "step": 14231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3828932072206845e-06,
+      "loss": 2.1539,
+      "step": 14232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.367007798530111e-06,
+      "loss": 2.0714,
+      "step": 14233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3511457376543835e-06,
+      "loss": 1.9168,
+      "step": 14234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3353070253420174e-06,
+      "loss": 1.9805,
+      "step": 14235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.319491662340803e-06,
+      "loss": 2.0661,
+      "step": 14236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.303699649397087e-06,
+      "loss": 2.1862,
+      "step": 14237
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.287930987256495e-06,
+      "loss": 1.9151,
+      "step": 14238
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.272185676663155e-06,
+      "loss": 2.128,
+      "step": 14239
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.256463718360416e-06,
+      "loss": 2.1384,
+      "step": 14240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.240765113090295e-06,
+      "loss": 2.0595,
+      "step": 14241
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.225089861593868e-06,
+      "loss": 2.0741,
+      "step": 14242
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.209437964611042e-06,
+      "loss": 2.0839,
+      "step": 14243
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1938094228805575e-06,
+      "loss": 1.9375,
+      "step": 14244
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.178204237140161e-06,
+      "loss": 2.0544,
+      "step": 14245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.162622408126483e-06,
+      "loss": 1.9922,
+      "step": 14246
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.147063936574991e-06,
+      "loss": 2.0464,
+      "step": 14247
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1315288232201e-06,
+      "loss": 2.0088,
+      "step": 14248
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.116017068794998e-06,
+      "loss": 1.9578,
+      "step": 14249
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.100528674032046e-06,
+      "loss": 1.9741,
+      "step": 14250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.085063639662213e-06,
+      "loss": 1.9589,
+      "step": 14251
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.069621966415527e-06,
+      "loss": 1.8342,
+      "step": 14252
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.054203655020851e-06,
+      "loss": 2.004,
+      "step": 14253
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.038808706205988e-06,
+      "loss": 2.1181,
+      "step": 14254
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.023437120697638e-06,
+      "loss": 1.9786,
+      "step": 14255
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.008088899221275e-06,
+      "loss": 1.8509,
+      "step": 14256
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.99276404250143e-06,
+      "loss": 2.004,
+      "step": 14257
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.977462551261524e-06,
+      "loss": 2.1436,
+      "step": 14258
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.96218442622376e-06,
+      "loss": 1.9399,
+      "step": 14259
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.946929668109335e-06,
+      "loss": 1.944,
+      "step": 14260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.931698277638286e-06,
+      "loss": 2.082,
+      "step": 14261
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.916490255529538e-06,
+      "loss": 1.9071,
+      "step": 14262
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.901305602501016e-06,
+      "loss": 1.9265,
+      "step": 14263
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.886144319269425e-06,
+      "loss": 1.8769,
+      "step": 14264
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.871006406550416e-06,
+      "loss": 2.0914,
+      "step": 14265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.855891865058526e-06,
+      "loss": 1.913,
+      "step": 14266
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.840800695507241e-06,
+      "loss": 2.0847,
+      "step": 14267
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.825732898608825e-06,
+      "loss": 2.061,
+      "step": 14268
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.810688475074598e-06,
+      "loss": 2.0576,
+      "step": 14269
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7956674256146024e-06,
+      "loss": 1.8716,
+      "step": 14270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.780669750937939e-06,
+      "loss": 1.9477,
+      "step": 14271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7656954517525385e-06,
+      "loss": 2.037,
+      "step": 14272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.750744528765116e-06,
+      "loss": 1.9507,
+      "step": 14273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.735816982681551e-06,
+      "loss": 1.8945,
+      "step": 14274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.72091281420628e-06,
+      "loss": 2.0647,
+      "step": 14275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.706032024042961e-06,
+      "loss": 2.1082,
+      "step": 14276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.691174612893923e-06,
+      "loss": 1.9504,
+      "step": 14277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.676340581460492e-06,
+      "loss": 1.9557,
+      "step": 14278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.661529930442887e-06,
+      "loss": 2.0195,
+      "step": 14279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.646742660540215e-06,
+      "loss": 2.0432,
+      "step": 14280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.631978772450363e-06,
+      "loss": 2.1273,
+      "step": 14281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.617238266870327e-06,
+      "loss": 2.0091,
+      "step": 14282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.602521144495886e-06,
+      "loss": 1.926,
+      "step": 14283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.587827406021705e-06,
+      "loss": 1.9741,
+      "step": 14284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.573157052141286e-06,
+      "loss": 1.9694,
+      "step": 14285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.558510083547241e-06,
+      "loss": 1.9822,
+      "step": 14286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.54388650093085e-06,
+      "loss": 2.0346,
+      "step": 14287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.529286304982394e-06,
+      "loss": 1.9468,
+      "step": 14288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.514709496391045e-06,
+      "loss": 2.076,
+      "step": 14289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.500156075844919e-06,
+      "loss": 1.9652,
+      "step": 14290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.485626044030855e-06,
+      "loss": 2.0165,
+      "step": 14291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.471119401634749e-06,
+      "loss": 1.8969,
+      "step": 14292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.456636149341387e-06,
+      "loss": 1.8932,
+      "step": 14293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.442176287834443e-06,
+      "loss": 1.8829,
+      "step": 14294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.427739817796317e-06,
+      "loss": 2.0057,
+      "step": 14295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4133267399085185e-06,
+      "loss": 2.1376,
+      "step": 14296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.398937054851449e-06,
+      "loss": 2.004,
+      "step": 14297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.38457076330423e-06,
+      "loss": 1.9824,
+      "step": 14298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.370227865945042e-06,
+      "loss": 1.8352,
+      "step": 14299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3559083634509e-06,
+      "loss": 1.9493,
+      "step": 14300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.341612256497651e-06,
+      "loss": 2.0412,
+      "step": 14301
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.327339545760201e-06,
+      "loss": 2.0153,
+      "step": 14302
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.313090231912231e-06,
+      "loss": 1.9352,
+      "step": 14303
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2988643156262605e-06,
+      "loss": 2.1149,
+      "step": 14304
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.284661797573863e-06,
+      "loss": 1.9727,
+      "step": 14305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.270482678425447e-06,
+      "loss": 1.9343,
+      "step": 14306
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.256326958850198e-06,
+      "loss": 1.9676,
+      "step": 14307
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.242194639516417e-06,
+      "loss": 2.0168,
+      "step": 14308
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.228085721091124e-06,
+      "loss": 1.9536,
+      "step": 14309
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.214000204240287e-06,
+      "loss": 1.8961,
+      "step": 14310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.199938089628763e-06,
+      "loss": 1.9176,
+      "step": 14311
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1858993779203544e-06,
+      "loss": 2.0165,
+      "step": 14312
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.171884069777699e-06,
+      "loss": 1.9701,
+      "step": 14313
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.157892165862431e-06,
+      "loss": 1.9723,
+      "step": 14314
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.143923666834859e-06,
+      "loss": 1.9878,
+      "step": 14315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.129978573354398e-06,
+      "loss": 2.0712,
+      "step": 14316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1160568860793e-06,
+      "loss": 1.9138,
+      "step": 14317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1021586056666505e-06,
+      "loss": 2.0585,
+      "step": 14318
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.088283732772591e-06,
+      "loss": 1.8952,
+      "step": 14319
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.074432268051875e-06,
+      "loss": 2.0034,
+      "step": 14320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.060604212158536e-06,
+      "loss": 2.0428,
+      "step": 14321
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.046799565745108e-06,
+      "loss": 1.9906,
+      "step": 14322
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.033018329463289e-06,
+      "loss": 1.9487,
+      "step": 14323
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.019260503963562e-06,
+      "loss": 2.22,
+      "step": 14324
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.005526089895295e-06,
+      "loss": 1.9583,
+      "step": 14325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.991815087906914e-06,
+      "loss": 1.8422,
+      "step": 14326
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9781274986454566e-06,
+      "loss": 1.9876,
+      "step": 14327
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.964463322757017e-06,
+      "loss": 2.0773,
+      "step": 14328
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.950822560886691e-06,
+      "loss": 1.9831,
+      "step": 14329
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.937205213678297e-06,
+      "loss": 1.9314,
+      "step": 14330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.923611281774597e-06,
+      "loss": 1.9919,
+      "step": 14331
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.910040765817246e-06,
+      "loss": 2.0302,
+      "step": 14332
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.896493666446843e-06,
+      "loss": 1.9915,
+      "step": 14333
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.882969984302765e-06,
+      "loss": 2.0255,
+      "step": 14334
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.869469720023444e-06,
+      "loss": 1.9863,
+      "step": 14335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.855992874246095e-06,
+      "loss": 2.1137,
+      "step": 14336
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.842539447606818e-06,
+      "loss": 2.1017,
+      "step": 14337
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.829109440740719e-06,
+      "loss": 2.0259,
+      "step": 14338
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8157028542816776e-06,
+      "loss": 1.8574,
+      "step": 14339
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8023196888624657e-06,
+      "loss": 1.9562,
+      "step": 14340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7889599451149114e-06,
+      "loss": 1.9436,
+      "step": 14341
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.775623623669566e-06,
+      "loss": 1.9235,
+      "step": 14342
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7623107251559818e-06,
+      "loss": 2.1511,
+      "step": 14343
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7490212502024335e-06,
+      "loss": 1.8945,
+      "step": 14344
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.735755199436308e-06,
+      "loss": 1.9398,
+      "step": 14345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.722512573483827e-06,
+      "loss": 2.0479,
+      "step": 14346
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7092933729699907e-06,
+      "loss": 1.9871,
+      "step": 14347
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.696097598518855e-06,
+      "loss": 1.8628,
+      "step": 14348
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6829252507531997e-06,
+      "loss": 2.0228,
+      "step": 14349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6697763302948605e-06,
+      "loss": 2.1335,
+      "step": 14350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.656650837764508e-06,
+      "loss": 2.1063,
+      "step": 14351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.643548773781591e-06,
+      "loss": 1.9098,
+      "step": 14352
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6304701389647253e-06,
+      "loss": 1.9455,
+      "step": 14353
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6174149339310846e-06,
+      "loss": 1.9937,
+      "step": 14354
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.604383159297009e-06,
+      "loss": 2.0078,
+      "step": 14355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.591374815677562e-06,
+      "loss": 1.9218,
+      "step": 14356
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.578389903686807e-06,
+      "loss": 2.2066,
+      "step": 14357
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5654284239377e-06,
+      "loss": 2.0657,
+      "step": 14358
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5524903770419726e-06,
+      "loss": 2.0492,
+      "step": 14359
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5395757636103586e-06,
+      "loss": 1.9764,
+      "step": 14360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.526684584252482e-06,
+      "loss": 2.0978,
+      "step": 14361
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5138168395768556e-06,
+      "loss": 1.9403,
+      "step": 14362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5009725301907714e-06,
+      "loss": 2.0352,
+      "step": 14363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.488151656700578e-06,
+      "loss": 2.0924,
+      "step": 14364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4753542197114575e-06,
+      "loss": 1.9947,
+      "step": 14365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4625802198274826e-06,
+      "loss": 1.9891,
+      "step": 14366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4498296576515597e-06,
+      "loss": 1.9966,
+      "step": 14367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.437102533785541e-06,
+      "loss": 2.2528,
+      "step": 14368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.424398848830279e-06,
+      "loss": 1.8502,
+      "step": 14369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4117186033853497e-06,
+      "loss": 2.0107,
+      "step": 14370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3990617980492744e-06,
+      "loss": 1.9623,
+      "step": 14371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3864284334195193e-06,
+      "loss": 1.8304,
+      "step": 14372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3738185100923857e-06,
+      "loss": 1.8796,
+      "step": 14373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3612320286630636e-06,
+      "loss": 1.9683,
+      "step": 14374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.348668989725745e-06,
+      "loss": 1.9645,
+      "step": 14375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.336129393873344e-06,
+      "loss": 2.0259,
+      "step": 14376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3236132416978314e-06,
+      "loss": 2.0464,
+      "step": 14377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3111205337899575e-06,
+      "loss": 2.1903,
+      "step": 14378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.298651270739417e-06,
+      "loss": 2.1886,
+      "step": 14379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2862054531347385e-06,
+      "loss": 2.1843,
+      "step": 14380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2737830815635084e-06,
+      "loss": 1.8918,
+      "step": 14381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.261384156612035e-06,
+      "loss": 2.1241,
+      "step": 14382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.249008678865517e-06,
+      "loss": 1.9191,
+      "step": 14383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2366566489081537e-06,
+      "loss": 2.0502,
+      "step": 14384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2243280673230347e-06,
+      "loss": 1.9801,
+      "step": 14385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.212022934692027e-06,
+      "loss": 2.0648,
+      "step": 14386
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1997412515959447e-06,
+      "loss": 1.9688,
+      "step": 14387
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1874830186146007e-06,
+      "loss": 1.9491,
+      "step": 14388
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.175248236326589e-06,
+      "loss": 1.955,
+      "step": 14389
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1630369053093356e-06,
+      "loss": 1.8672,
+      "step": 14390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.150849026139324e-06,
+      "loss": 2.1786,
+      "step": 14391
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1386845993918165e-06,
+      "loss": 2.0061,
+      "step": 14392
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1265436256410763e-06,
+      "loss": 1.9622,
+      "step": 14393
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1144261054600887e-06,
+      "loss": 2.0593,
+      "step": 14394
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.102332039420841e-06,
+      "loss": 2.0472,
+      "step": 14395
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0902614280942097e-06,
+      "loss": 2.0932,
+      "step": 14396
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0782142720500174e-06,
+      "loss": 2.0917,
+      "step": 14397
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0661905718568638e-06,
+      "loss": 2.0696,
+      "step": 14398
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.05419032808224e-06,
+      "loss": 2.0229,
+      "step": 14399
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0422135412926932e-06,
+      "loss": 2.0307,
+      "step": 14400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0302602120534928e-06,
+      "loss": 1.9778,
+      "step": 14401
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0183303409289096e-06,
+      "loss": 1.9173,
+      "step": 14402
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.006423928481994e-06,
+      "loss": 1.951,
+      "step": 14403
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9945409752747952e-06,
+      "loss": 1.91,
+      "step": 14404
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9826814818682544e-06,
+      "loss": 1.9132,
+      "step": 14405
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9708454488220905e-06,
+      "loss": 2.0991,
+      "step": 14406
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9590328766950225e-06,
+      "loss": 1.9659,
+      "step": 14407
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9472437660446603e-06,
+      "loss": 2.0145,
+      "step": 14408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9354781174274482e-06,
+      "loss": 2.1347,
+      "step": 14409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.923735931398719e-06,
+      "loss": 2.0057,
+      "step": 14410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.912017208512752e-06,
+      "loss": 2.1371,
+      "step": 14411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9003219493227707e-06,
+      "loss": 1.9419,
+      "step": 14412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.888650154380723e-06,
+      "loss": 2.0989,
+      "step": 14413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8770018242376126e-06,
+      "loss": 1.9703,
+      "step": 14414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8653769594432224e-06,
+      "loss": 2.031,
+      "step": 14415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.853775560546279e-06,
+      "loss": 2.0563,
+      "step": 14416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8421976280944005e-06,
+      "loss": 1.9428,
+      "step": 14417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8306431626340945e-06,
+      "loss": 1.943,
+      "step": 14418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8191121647107575e-06,
+      "loss": 2.0813,
+      "step": 14419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.807604634868677e-06,
+      "loss": 1.9877,
+      "step": 14420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7961205736510844e-06,
+      "loss": 1.8594,
+      "step": 14421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.784659981599935e-06,
+      "loss": 1.9396,
+      "step": 14422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.773222859256297e-06,
+      "loss": 2.0172,
+      "step": 14423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7618092071600155e-06,
+      "loss": 1.922,
+      "step": 14424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7504190258498263e-06,
+      "loss": 2.1166,
+      "step": 14425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.739052315863355e-06,
+      "loss": 1.8751,
+      "step": 14426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7277090777371172e-06,
+      "loss": 1.9231,
+      "step": 14427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7163893120066286e-06,
+      "loss": 2.0836,
+      "step": 14428
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7050930192061284e-06,
+      "loss": 1.863,
+      "step": 14429
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6938201998688572e-06,
+      "loss": 1.982,
+      "step": 14430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.682570854526889e-06,
+      "loss": 1.9135,
+      "step": 14431
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6713449837112993e-06,
+      "loss": 1.8848,
+      "step": 14432
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.660142587951886e-06,
+      "loss": 1.984,
+      "step": 14433
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6489636677775042e-06,
+      "loss": 1.8682,
+      "step": 14434
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.637808223715732e-06,
+      "loss": 2.131,
+      "step": 14435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6266762562932035e-06,
+      "loss": 2.0055,
+      "step": 14436
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6155677660353316e-06,
+      "loss": 1.9722,
+      "step": 14437
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.604482753466475e-06,
+      "loss": 1.9695,
+      "step": 14438
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.593421219109882e-06,
+      "loss": 2.0581,
+      "step": 14439
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.582383163487745e-06,
+      "loss": 2.1571,
+      "step": 14440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5713685871209814e-06,
+      "loss": 1.9584,
+      "step": 14441
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5603774905295086e-06,
+      "loss": 2.0857,
+      "step": 14442
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5494098742321892e-06,
+      "loss": 1.9372,
+      "step": 14443
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5384657387467205e-06,
+      "loss": 1.9807,
+      "step": 14444
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.527545084589633e-06,
+      "loss": 1.9875,
+      "step": 14445
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5166479122764596e-06,
+      "loss": 2.0414,
+      "step": 14446
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.505774222321511e-06,
+      "loss": 2.0665,
+      "step": 14447
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4949240152381535e-06,
+      "loss": 2.1519,
+      "step": 14448
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4840972915384784e-06,
+      "loss": 2.161,
+      "step": 14449
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.473294051733521e-06,
+      "loss": 1.9673,
+      "step": 14450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.462514296333207e-06,
+      "loss": 2.0683,
+      "step": 14451
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4517580258464066e-06,
+      "loss": 2.0734,
+      "step": 14452
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.441025240780825e-06,
+      "loss": 1.9505,
+      "step": 14453
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.430315941643113e-06,
+      "loss": 1.9933,
+      "step": 14454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.419630128938699e-06,
+      "loss": 2.0163,
+      "step": 14455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.408967803172013e-06,
+      "loss": 2.0803,
+      "step": 14456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.398328964846375e-06,
+      "loss": 1.9325,
+      "step": 14457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3877136144638823e-06,
+      "loss": 1.9444,
+      "step": 14458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3771217525256905e-06,
+      "loss": 1.7878,
+      "step": 14459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3665533795316774e-06,
+      "loss": 1.9613,
+      "step": 14460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.356008495980777e-06,
+      "loss": 2.016,
+      "step": 14461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3454871023707025e-06,
+      "loss": 2.0335,
+      "step": 14462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3349891991980566e-06,
+      "loss": 2.1675,
+      "step": 14463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3245147869583873e-06,
+      "loss": 2.1253,
+      "step": 14464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3140638661461322e-06,
+      "loss": 2.02,
+      "step": 14465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3036364372545637e-06,
+      "loss": 1.9443,
+      "step": 14466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.293232500775899e-06,
+      "loss": 2.0361,
+      "step": 14467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.28285205720119e-06,
+      "loss": 2.074,
+      "step": 14468
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2724951070204893e-06,
+      "loss": 2.1365,
+      "step": 14469
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2621616507226272e-06,
+      "loss": 2.0375,
+      "step": 14470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2518516887953255e-06,
+      "loss": 2.0985,
+      "step": 14471
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.241565221725361e-06,
+      "loss": 1.9869,
+      "step": 14472
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.231302249998124e-06,
+      "loss": 1.8766,
+      "step": 14473
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2210627740981703e-06,
+      "loss": 1.9954,
+      "step": 14474
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2108467945087807e-06,
+      "loss": 2.1307,
+      "step": 14475
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.20065431171218e-06,
+      "loss": 2.0596,
+      "step": 14476
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1904853261894287e-06,
+      "loss": 2.0884,
+      "step": 14477
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.180339838420586e-06,
+      "loss": 1.8959,
+      "step": 14478
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.170217848884548e-06,
+      "loss": 2.1988,
+      "step": 14479
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.160119358059043e-06,
+      "loss": 2.003,
+      "step": 14480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1500443664208004e-06,
+      "loss": 2.117,
+      "step": 14481
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.139992874445329e-06,
+      "loss": 1.9836,
+      "step": 14482
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.129964882607138e-06,
+      "loss": 1.8994,
+      "step": 14483
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.119960391379516e-06,
+      "loss": 2.0624,
+      "step": 14484
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1099794012347516e-06,
+      "loss": 1.8797,
+      "step": 14485
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1000219126439124e-06,
+      "loss": 2.1638,
+      "step": 14486
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0900879260770666e-06,
+      "loss": 2.0675,
+      "step": 14487
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.080177442003117e-06,
+      "loss": 1.9146,
+      "step": 14488
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0702904608898564e-06,
+      "loss": 2.0323,
+      "step": 14489
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0604269832039113e-06,
+      "loss": 2.0604,
+      "step": 14490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0505870094109093e-06,
+      "loss": 2.1088,
+      "step": 14491
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0407705399753673e-06,
+      "loss": 1.9206,
+      "step": 14492
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.030977575360582e-06,
+      "loss": 1.9491,
+      "step": 14493
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0212081160287945e-06,
+      "loss": 2.0813,
+      "step": 14494
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.011462162441191e-06,
+      "loss": 1.9741,
+      "step": 14495
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.001739715057793e-06,
+      "loss": 2.079,
+      "step": 14496
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9920407743375113e-06,
+      "loss": 2.0588,
+      "step": 14497
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.982365340738146e-06,
+      "loss": 1.9988,
+      "step": 14498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9727134147163874e-06,
+      "loss": 1.9749,
+      "step": 14499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9630849967279262e-06,
+      "loss": 2.0508,
+      "step": 14500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.953480087227122e-06,
+      "loss": 1.9753,
+      "step": 14501
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.943898686667389e-06,
+      "loss": 2.1075,
+      "step": 14502
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9343407955010883e-06,
+      "loss": 2.0103,
+      "step": 14503
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9248064141791923e-06,
+      "loss": 2.0349,
+      "step": 14504
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.915295543151896e-06,
+      "loss": 2.0222,
+      "step": 14505
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.905808182868063e-06,
+      "loss": 2.0956,
+      "step": 14506
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.896344333775557e-06,
+      "loss": 1.9592,
+      "step": 14507
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8869039963210765e-06,
+      "loss": 1.964,
+      "step": 14508
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8774871709502095e-06,
+      "loss": 2.0973,
+      "step": 14509
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.868093858107489e-06,
+      "loss": 1.9931,
+      "step": 14510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8587240582362274e-06,
+      "loss": 1.9266,
+      "step": 14511
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.849377771778793e-06,
+      "loss": 2.0654,
+      "step": 14512
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8400549991763327e-06,
+      "loss": 2.0479,
+      "step": 14513
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.830755740868828e-06,
+      "loss": 2.044,
+      "step": 14514
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8214799972953166e-06,
+      "loss": 1.9556,
+      "step": 14515
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8122277688935595e-06,
+      "loss": 1.8811,
+      "step": 14516
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.802999056100374e-06,
+      "loss": 1.9831,
+      "step": 14517
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7937938593512448e-06,
+      "loss": 2.0783,
+      "step": 14518
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7846121790808246e-06,
+      "loss": 1.828,
+      "step": 14519
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.775454015722433e-06,
+      "loss": 1.9064,
+      "step": 14520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7663193697083357e-06,
+      "loss": 2.0568,
+      "step": 14521
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7572082414697433e-06,
+      "loss": 1.9902,
+      "step": 14522
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7481206314367558e-06,
+      "loss": 2.0125,
+      "step": 14523
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.739056540038253e-06,
+      "loss": 2.0453,
+      "step": 14524
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7300159677021142e-06,
+      "loss": 2.0192,
+      "step": 14525
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7209989148551098e-06,
+      "loss": 2.0845,
+      "step": 14526
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7120053819228433e-06,
+      "loss": 2.0416,
+      "step": 14527
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7030353693298085e-06,
+      "loss": 1.9624,
+      "step": 14528
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.694088877499389e-06,
+      "loss": 1.8736,
+      "step": 14529
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6851659068539138e-06,
+      "loss": 1.8336,
+      "step": 14530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6762664578146015e-06,
+      "loss": 1.9932,
+      "step": 14531
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6673905308015048e-06,
+      "loss": 2.0299,
+      "step": 14532
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6585381262335663e-06,
+      "loss": 1.9812,
+      "step": 14533
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6497092445286743e-06,
+      "loss": 1.9231,
+      "step": 14534
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6409038861034953e-06,
+      "loss": 2.0007,
+      "step": 14535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6321220513737523e-06,
+      "loss": 2.0115,
+      "step": 14536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.623363740753947e-06,
+      "loss": 2.0286,
+      "step": 14537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6146289546574155e-06,
+      "loss": 2.0731,
+      "step": 14538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6059176934965503e-06,
+      "loss": 1.988,
+      "step": 14539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5972299576825223e-06,
+      "loss": 2.1529,
+      "step": 14540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5885657476253922e-06,
+      "loss": 2.1082,
+      "step": 14541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5799250637341667e-06,
+      "loss": 2.0745,
+      "step": 14542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5713079064166303e-06,
+      "loss": 2.0308,
+      "step": 14543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5627142760796243e-06,
+      "loss": 1.949,
+      "step": 14544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5541441731287686e-06,
+      "loss": 1.9241,
+      "step": 14545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5455975979685177e-06,
+      "loss": 2.0661,
+      "step": 14546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.537074551002382e-06,
+      "loss": 2.0894,
+      "step": 14547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5285750326325953e-06,
+      "loss": 1.9417,
+      "step": 14548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5200990432604477e-06,
+      "loss": 1.9732,
+      "step": 14549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5116465832858973e-06,
+      "loss": 2.0758,
+      "step": 14550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5032176531080133e-06,
+      "loss": 2.0622,
+      "step": 14551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4948122531246444e-06,
+      "loss": 2.2203,
+      "step": 14552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4864303837325288e-06,
+      "loss": 1.9762,
+      "step": 14553
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4780720453272944e-06,
+      "loss": 1.9999,
+      "step": 14554
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4697372383035146e-06,
+      "loss": 2.0837,
+      "step": 14555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4614259630545967e-06,
+      "loss": 2.1173,
+      "step": 14556
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4531382199728937e-06,
+      "loss": 2.0078,
+      "step": 14557
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4448740094494816e-06,
+      "loss": 1.9194,
+      "step": 14558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.436633331874604e-06,
+      "loss": 2.0396,
+      "step": 14559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.428416187637116e-06,
+      "loss": 1.958,
+      "step": 14560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.42022257712493e-06,
+      "loss": 2.2018,
+      "step": 14561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4120525007248474e-06,
+      "loss": 1.9341,
+      "step": 14562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.403905958822449e-06,
+      "loss": 2.0393,
+      "step": 14563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3957829518023157e-06,
+      "loss": 1.9704,
+      "step": 14564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3876834800478077e-06,
+      "loss": 2.0215,
+      "step": 14565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3796075439413414e-06,
+      "loss": 2.0131,
+      "step": 14566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3715551438640561e-06,
+      "loss": 1.8644,
+      "step": 14567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3635262801960369e-06,
+      "loss": 2.0121,
+      "step": 14568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3555209533162582e-06,
+      "loss": 1.9761,
+      "step": 14569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3475391636025846e-06,
+      "loss": 2.0428,
+      "step": 14570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3395809114318813e-06,
+      "loss": 2.0467,
+      "step": 14571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3316461971796812e-06,
+      "loss": 1.8815,
+      "step": 14572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3237350212205178e-06,
+      "loss": 1.9174,
+      "step": 14573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3158473839278706e-06,
+      "loss": 1.9336,
+      "step": 14574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.307983285674108e-06,
+      "loss": 2.1539,
+      "step": 14575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3001427268302668e-06,
+      "loss": 1.9477,
+      "step": 14576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2923257077666062e-06,
+      "loss": 1.9371,
+      "step": 14577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2845322288519978e-06,
+      "loss": 2.0107,
+      "step": 14578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2767622904543696e-06,
+      "loss": 1.9536,
+      "step": 14579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2690158929404284e-06,
+      "loss": 2.0637,
+      "step": 14580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2612930366758813e-06,
+      "loss": 1.9614,
+      "step": 14581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2535937220252147e-06,
+      "loss": 2.087,
+      "step": 14582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2459179493519157e-06,
+      "loss": 2.0894,
+      "step": 14583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.23826571901825e-06,
+      "loss": 1.9082,
+      "step": 14584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2306370313854287e-06,
+      "loss": 2.0151,
+      "step": 14585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.223031886813497e-06,
+      "loss": 1.9286,
+      "step": 14586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.215450285661557e-06,
+      "loss": 2.0386,
+      "step": 14587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2078922282873772e-06,
+      "loss": 2.1063,
+      "step": 14588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2003577150477284e-06,
+      "loss": 2.0425,
+      "step": 14589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1928467462982706e-06,
+      "loss": 1.8787,
+      "step": 14590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1853593223935533e-06,
+      "loss": 2.0005,
+      "step": 14591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.177895443687016e-06,
+      "loss": 1.9643,
+      "step": 14592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.170455110530877e-06,
+      "loss": 2.0581,
+      "step": 14593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1630383232764664e-06,
+      "loss": 1.8638,
+      "step": 14594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1556450822737819e-06,
+      "loss": 1.9549,
+      "step": 14595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1482753878718223e-06,
+      "loss": 2.0485,
+      "step": 14596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1409292404184757e-06,
+      "loss": 2.0268,
+      "step": 14597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.133606640260465e-06,
+      "loss": 2.1063,
+      "step": 14598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1263075877434582e-06,
+      "loss": 1.9271,
+      "step": 14599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.119032083211957e-06,
+      "loss": 2.0622,
+      "step": 14600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1117801270094096e-06,
+      "loss": 1.9393,
+      "step": 14601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.104551719478153e-06,
+      "loss": 1.8758,
+      "step": 14602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.097346860959303e-06,
+      "loss": 1.9537,
+      "step": 14603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.090165551793032e-06,
+      "loss": 2.0878,
+      "step": 14604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0830077923182912e-06,
+      "loss": 1.9442,
+      "step": 14605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0758735828729215e-06,
+      "loss": 1.9816,
+      "step": 14606
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0687629237936537e-06,
+      "loss": 2.0413,
+      "step": 14607
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0616758154161631e-06,
+      "loss": 2.0501,
+      "step": 14608
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0546122580749605e-06,
+      "loss": 1.9127,
+      "step": 14609
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0475722521035014e-06,
+      "loss": 1.9005,
+      "step": 14610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0405557978340197e-06,
+      "loss": 2.0001,
+      "step": 14611
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0335628955977504e-06,
+      "loss": 1.9587,
+      "step": 14612
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0265935457248188e-06,
+      "loss": 2.0465,
+      "step": 14613
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0196477485440725e-06,
+      "loss": 2.0563,
+      "step": 14614
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.012725504383527e-06,
+      "loss": 1.9773,
+      "step": 14615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.00582681356981e-06,
+      "loss": 2.114,
+      "step": 14616
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.9895167642855e-07,
+      "loss": 1.9631,
+      "step": 14617
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.92100093284376e-07,
+      "loss": 2.1435,
+      "step": 14618
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.85272064460585e-07,
+      "loss": 2.0545,
+      "step": 14619
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.784675902795304e-07,
+      "loss": 2.0289,
+      "step": 14620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.716866710623994e-07,
+      "loss": 2.0481,
+      "step": 14621
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.649293071292697e-07,
+      "loss": 1.9471,
+      "step": 14622
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.581954987991082e-07,
+      "loss": 2.0035,
+      "step": 14623
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.514852463897161e-07,
+      "loss": 2.0403,
+      "step": 14624
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.447985502178957e-07,
+      "loss": 1.9027,
+      "step": 14625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.381354105992279e-07,
+      "loss": 2.0622,
+      "step": 14626
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.314958278482388e-07,
+      "loss": 1.8583,
+      "step": 14627
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.248798022784e-07,
+      "loss": 2.0013,
+      "step": 14628
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.18287334201906e-07,
+      "loss": 2.0755,
+      "step": 14629
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.117184239299525e-07,
+      "loss": 1.884,
+      "step": 14630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.051730717726248e-07,
+      "loss": 2.026,
+      "step": 14631
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.986512780389533e-07,
+      "loss": 2.0507,
+      "step": 14632
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.921530430366364e-07,
+      "loss": 1.9457,
+      "step": 14633
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.856783670724843e-07,
+      "loss": 1.8683,
+      "step": 14634
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.792272504521415e-07,
+      "loss": 2.1326,
+      "step": 14635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.727996934800864e-07,
+      "loss": 1.8748,
+      "step": 14636
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.663956964596876e-07,
+      "loss": 1.9094,
+      "step": 14637
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.600152596933142e-07,
+      "loss": 1.9179,
+      "step": 14638
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.536583834820588e-07,
+      "loss": 1.9876,
+      "step": 14639
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.473250681259592e-07,
+      "loss": 2.0479,
+      "step": 14640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.410153139240539e-07,
+      "loss": 2.0071,
+      "step": 14641
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.347291211741049e-07,
+      "loss": 1.9005,
+      "step": 14642
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.284664901728189e-07,
+      "loss": 2.046,
+      "step": 14643
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.222274212159042e-07,
+      "loss": 1.962,
+      "step": 14644
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.160119145977918e-07,
+      "loss": 2.0916,
+      "step": 14645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.09819970611858e-07,
+      "loss": 2.023,
+      "step": 14646
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.036515895504248e-07,
+      "loss": 2.0381,
+      "step": 14647
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.975067717045925e-07,
+      "loss": 1.7695,
+      "step": 14648
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.913855173644068e-07,
+      "loss": 1.973,
+      "step": 14649
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.852878268189145e-07,
+      "loss": 2.0375,
+      "step": 14650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.792137003558297e-07,
+      "loss": 1.7863,
+      "step": 14651
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.731631382619231e-07,
+      "loss": 1.8749,
+      "step": 14652
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.67136140822744e-07,
+      "loss": 1.9988,
+      "step": 14653
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.611327083228425e-07,
+      "loss": 2.0552,
+      "step": 14654
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.551528410456032e-07,
+      "loss": 2.0396,
+      "step": 14655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.491965392731892e-07,
+      "loss": 1.9349,
+      "step": 14656
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.4326380328682e-07,
+      "loss": 1.9771,
+      "step": 14657
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.373546333664938e-07,
+      "loss": 1.9643,
+      "step": 14658
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.314690297912097e-07,
+      "loss": 1.9232,
+      "step": 14659
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.256069928387454e-07,
+      "loss": 2.0168,
+      "step": 14660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.197685227858242e-07,
+      "loss": 1.946,
+      "step": 14661
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.139536199079477e-07,
+      "loss": 2.0769,
+      "step": 14662
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.081622844796743e-07,
+      "loss": 1.8685,
+      "step": 14663
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.023945167743406e-07,
+      "loss": 1.9831,
+      "step": 14664
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.966503170642292e-07,
+      "loss": 1.8608,
+      "step": 14665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.909296856204561e-07,
+      "loss": 2.0045,
+      "step": 14666
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.852326227130834e-07,
+      "loss": 1.9984,
+      "step": 14667
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.795591286109515e-07,
+      "loss": 1.9318,
+      "step": 14668
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.739092035819016e-07,
+      "loss": 2.0439,
+      "step": 14669
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.682828478926095e-07,
+      "loss": 2.0457,
+      "step": 14670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.626800618086959e-07,
+      "loss": 2.1774,
+      "step": 14671
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.571008455945604e-07,
+      "loss": 1.9373,
+      "step": 14672
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.51545199513659e-07,
+      "loss": 2.0393,
+      "step": 14673
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.460131238281153e-07,
+      "loss": 1.9053,
+      "step": 14674
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.405046187991093e-07,
+      "loss": 1.875,
+      "step": 14675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.350196846865996e-07,
+      "loss": 1.872,
+      "step": 14676
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.295583217496015e-07,
+      "loss": 2.0392,
+      "step": 14677
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.241205302457975e-07,
+      "loss": 1.95,
+      "step": 14678
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.187063104319268e-07,
+      "loss": 2.0895,
+      "step": 14679
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.133156625635628e-07,
+      "loss": 1.8778,
+      "step": 14680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.079485868951129e-07,
+      "loss": 1.9523,
+      "step": 14681
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.026050836798747e-07,
+      "loss": 2.0514,
+      "step": 14682
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.972851531701462e-07,
+      "loss": 1.8108,
+      "step": 14683
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.9198879561706e-07,
+      "loss": 1.9223,
+      "step": 14684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.867160112705272e-07,
+      "loss": 1.9652,
+      "step": 14685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.814668003795154e-07,
+      "loss": 2.0832,
+      "step": 14686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.762411631917708e-07,
+      "loss": 1.9326,
+      "step": 14687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.710390999539849e-07,
+      "loss": 2.0311,
+      "step": 14688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.658606109116282e-07,
+      "loss": 2.0677,
+      "step": 14689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.607056963092272e-07,
+      "loss": 1.9668,
+      "step": 14690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.555743563900317e-07,
+      "loss": 2.0587,
+      "step": 14691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.504665913962925e-07,
+      "loss": 2.0039,
+      "step": 14692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.453824015690944e-07,
+      "loss": 1.9199,
+      "step": 14693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.403217871484123e-07,
+      "loss": 2.017,
+      "step": 14694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.35284748373166e-07,
+      "loss": 1.9018,
+      "step": 14695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.302712854811098e-07,
+      "loss": 1.9759,
+      "step": 14696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.252813987088323e-07,
+      "loss": 2.0535,
+      "step": 14697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.203150882918672e-07,
+      "loss": 1.8843,
+      "step": 14698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.153723544647492e-07,
+      "loss": 2.0022,
+      "step": 14699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.104531974606253e-07,
+      "loss": 2.0222,
+      "step": 14700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.055576175118092e-07,
+      "loss": 2.0105,
+      "step": 14701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.006856148493388e-07,
+      "loss": 1.9839,
+      "step": 14702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.958371897032521e-07,
+      "loss": 2.1252,
+      "step": 14703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.91012342302255e-07,
+      "loss": 2.0782,
+      "step": 14704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.862110728742208e-07,
+      "loss": 2.0902,
+      "step": 14705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.814333816457461e-07,
+      "loss": 1.7774,
+      "step": 14706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.766792688423727e-07,
+      "loss": 2.1285,
+      "step": 14707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7194873468842105e-07,
+      "loss": 1.9716,
+      "step": 14708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.672417794072681e-07,
+      "loss": 1.9403,
+      "step": 14709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6255840322106947e-07,
+      "loss": 1.9049,
+      "step": 14710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.578986063509261e-07,
+      "loss": 1.9956,
+      "step": 14711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5326238901666205e-07,
+      "loss": 1.9492,
+      "step": 14712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4864975143726893e-07,
+      "loss": 1.9853,
+      "step": 14713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4406069383040595e-07,
+      "loss": 2.2025,
+      "step": 14714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.394952164126775e-07,
+      "loss": 2.0648,
+      "step": 14715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3495331939963357e-07,
+      "loss": 2.0397,
+      "step": 14716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3043500300560255e-07,
+      "loss": 2.1474,
+      "step": 14717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2594026744385837e-07,
+      "loss": 1.892,
+      "step": 14718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2146911292662015e-07,
+      "loss": 1.933,
+      "step": 14719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.170215396649413e-07,
+      "loss": 1.9881,
+      "step": 14720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.125975478686539e-07,
+      "loss": 2.0086,
+      "step": 14721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0819713774670195e-07,
+      "loss": 1.899,
+      "step": 14722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0382030950669726e-07,
+      "loss": 2.0203,
+      "step": 14723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.994670633553632e-07,
+      "loss": 2.0349,
+      "step": 14724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9513739949803564e-07,
+      "loss": 2.0399,
+      "step": 14725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9083131813916204e-07,
+      "loss": 2.1332,
+      "step": 14726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.865488194819689e-07,
+      "loss": 1.9442,
+      "step": 14727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8228990372862757e-07,
+      "loss": 2.0015,
+      "step": 14728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.780545710801997e-07,
+      "loss": 1.9254,
+      "step": 14729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.738428217365253e-07,
+      "loss": 2.1168,
+      "step": 14730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6965465589644533e-07,
+      "loss": 1.9251,
+      "step": 14731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.65490073757635e-07,
+      "loss": 2.0148,
+      "step": 14732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.613490755167703e-07,
+      "loss": 1.9564,
+      "step": 14733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.572316613691395e-07,
+      "loss": 2.1294,
+      "step": 14734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.531378315092537e-07,
+      "loss": 2.056,
+      "step": 14735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4906758613023615e-07,
+      "loss": 1.8817,
+      "step": 14736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4502092542432197e-07,
+      "loss": 2.0003,
+      "step": 14737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4099784958241395e-07,
+      "loss": 1.883,
+      "step": 14738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3699835879447136e-07,
+      "loss": 1.8979,
+      "step": 14739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3302245324928757e-07,
+      "loss": 2.0467,
+      "step": 14740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.290701331344903e-07,
+      "loss": 1.8036,
+      "step": 14741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2514139863665246e-07,
+      "loss": 2.0654,
+      "step": 14742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.212362499412924e-07,
+      "loss": 1.9444,
+      "step": 14743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1735468723259607e-07,
+      "loss": 1.9553,
+      "step": 14744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1349671069391684e-07,
+      "loss": 1.9922,
+      "step": 14745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0966232050727573e-07,
+      "loss": 1.9098,
+      "step": 14746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.058515168537501e-07,
+      "loss": 1.9838,
+      "step": 14747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0206429991314066e-07,
+      "loss": 2.0246,
+      "step": 14748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9830066986419326e-07,
+      "loss": 1.9942,
+      "step": 14749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9456062688465456e-07,
+      "loss": 2.0468,
+      "step": 14750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.908441711509946e-07,
+      "loss": 2.1429,
+      "step": 14751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8715130283868405e-07,
+      "loss": 1.8957,
+      "step": 14752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.834820221219725e-07,
+      "loss": 2.0661,
+      "step": 14753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.798363291741102e-07,
+      "loss": 2.0121,
+      "step": 14754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.762142241671817e-07,
+      "loss": 2.0638,
+      "step": 14755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7261570727210585e-07,
+      "loss": 2.04,
+      "step": 14756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6904077865880226e-07,
+      "loss": 2.0574,
+      "step": 14757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.654894384959694e-07,
+      "loss": 2.0679,
+      "step": 14758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.619616869513064e-07,
+      "loss": 1.9135,
+      "step": 14759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.584575241912357e-07,
+      "loss": 2.0315,
+      "step": 14760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5497695038123605e-07,
+      "loss": 1.9993,
+      "step": 14761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.515199656855649e-07,
+      "loss": 2.0438,
+      "step": 14762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.480865702674251e-07,
+      "loss": 2.1957,
+      "step": 14763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4467676428885365e-07,
+      "loss": 1.9962,
+      "step": 14764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4129054791083294e-07,
+      "loss": 1.7995,
+      "step": 14765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3792792129317953e-07,
+      "loss": 1.9615,
+      "step": 14766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.345888845945998e-07,
+      "loss": 1.8811,
+      "step": 14767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3127343797268995e-07,
+      "loss": 2.0088,
+      "step": 14768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2798158158404692e-07,
+      "loss": 1.8792,
+      "step": 14769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2471331558393537e-07,
+      "loss": 2.1206,
+      "step": 14770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2146864012667635e-07,
+      "loss": 2.1659,
+      "step": 14771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1824755536542506e-07,
+      "loss": 1.9792,
+      "step": 14772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.150500614522266e-07,
+      "loss": 2.0655,
+      "step": 14773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.118761585380713e-07,
+      "loss": 1.9055,
+      "step": 14774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.087258467726172e-07,
+      "loss": 1.9363,
+      "step": 14775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0559912630468968e-07,
+      "loss": 1.9544,
+      "step": 14776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0249599728189294e-07,
+      "loss": 1.9576,
+      "step": 14777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9941645985060985e-07,
+      "loss": 1.922,
+      "step": 14778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9636051415622414e-07,
+      "loss": 1.906,
+      "step": 14779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9332816034306478e-07,
+      "loss": 1.9589,
+      "step": 14780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.90319398554184e-07,
+      "loss": 1.8879,
+      "step": 14781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8733422893163487e-07,
+      "loss": 2.1047,
+      "step": 14782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8437265161630468e-07,
+      "loss": 2.0202,
+      "step": 14783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.814346667479705e-07,
+      "loss": 2.0391,
+      "step": 14784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.785202744653547e-07,
+      "loss": 2.0164,
+      "step": 14785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7562947490606939e-07,
+      "loss": 2.0102,
+      "step": 14786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7276226820645003e-07,
+      "loss": 1.9408,
+      "step": 14787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6991865450188827e-07,
+      "loss": 2.034,
+      "step": 14788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6709863392661008e-07,
+      "loss": 2.0265,
+      "step": 14789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6430220661373119e-07,
+      "loss": 1.8628,
+      "step": 14790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6152937269525713e-07,
+      "loss": 1.9769,
+      "step": 14791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.587801323020832e-07,
+      "loss": 2.0218,
+      "step": 14792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5605448556393899e-07,
+      "loss": 1.9544,
+      "step": 14793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5335243260949927e-07,
+      "loss": 2.0707,
+      "step": 14794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.506739735662732e-07,
+      "loss": 1.9235,
+      "step": 14795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4801910856077072e-07,
+      "loss": 1.9652,
+      "step": 14796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4538783771822493e-07,
+      "loss": 2.0908,
+      "step": 14797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4278016116292537e-07,
+      "loss": 2.0457,
+      "step": 14798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4019607901788467e-07,
+      "loss": 2.0678,
+      "step": 14799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3763559140511638e-07,
+      "loss": 1.9963,
+      "step": 14800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3509869844541278e-07,
+      "loss": 2.0569,
+      "step": 14801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3258540025862241e-07,
+      "loss": 2.1597,
+      "step": 14802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3009569696331714e-07,
+      "loss": 1.8384,
+      "step": 14803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2762958867706954e-07,
+      "loss": 2.0525,
+      "step": 14804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2518707551623098e-07,
+      "loss": 1.9214,
+      "step": 14805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2276815759609816e-07,
+      "loss": 2.0283,
+      "step": 14806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.203728350308575e-07,
+      "loss": 2.0172,
+      "step": 14807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.180011079335852e-07,
+      "loss": 2.0207,
+      "step": 14808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1565297641624728e-07,
+      "loss": 1.9614,
+      "step": 14809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1332844058964398e-07,
+      "loss": 1.9526,
+      "step": 14810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1102750056346533e-07,
+      "loss": 1.9414,
+      "step": 14811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0875015644645769e-07,
+      "loss": 2.0651,
+      "step": 14812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0649640834597962e-07,
+      "loss": 1.9295,
+      "step": 14813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.04266256368446e-07,
+      "loss": 1.9806,
+      "step": 14814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0205970061916148e-07,
+      "loss": 1.9595,
+      "step": 14815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.987674120226497e-08,
+      "loss": 2.0612,
+      "step": 14816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.77173782207852e-08,
+      "loss": 2.1127,
+      "step": 14817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.55816117766961e-08,
+      "loss": 2.0501,
+      "step": 14818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.346944197069495e-08,
+      "loss": 1.9055,
+      "step": 14819
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.138086890259079e-08,
+      "loss": 1.9377,
+      "step": 14820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.93158926709714e-08,
+      "loss": 1.9593,
+      "step": 14821
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.727451337325887e-08,
+      "loss": 1.9442,
+      "step": 14822
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.525673110582055e-08,
+      "loss": 1.9901,
+      "step": 14823
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.326254596385807e-08,
+      "loss": 2.0463,
+      "step": 14824
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.129195804162937e-08,
+      "loss": 2.1877,
+      "step": 14825
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.934496743200459e-08,
+      "loss": 1.9945,
+      "step": 14826
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.742157422696571e-08,
+      "loss": 1.9587,
+      "step": 14827
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.5521778517329e-08,
+      "loss": 2.0196,
+      "step": 14828
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.364558039268943e-08,
+      "loss": 1.9106,
+      "step": 14829
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.17929799416428e-08,
+      "loss": 1.9826,
+      "step": 14830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.996397725161918e-08,
+      "loss": 1.8762,
+      "step": 14831
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.815857240899393e-08,
+      "loss": 2.1521,
+      "step": 14832
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.637676549897665e-08,
+      "loss": 2.0463,
+      "step": 14833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.461855660566674e-08,
+      "loss": 1.964,
+      "step": 14834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.288394581205337e-08,
+      "loss": 2.0346,
+      "step": 14835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.117293319995998e-08,
+      "loss": 1.9434,
+      "step": 14836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.9485518850266316e-08,
+      "loss": 1.9412,
+      "step": 14837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.782170284251986e-08,
+      "loss": 2.0418,
+      "step": 14838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.6181485255268894e-08,
+      "loss": 1.9195,
+      "step": 14839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.456486616600698e-08,
+      "loss": 2.0321,
+      "step": 14840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.297184565095092e-08,
+      "loss": 1.9904,
+      "step": 14841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.140242378537385e-08,
+      "loss": 2.1198,
+      "step": 14842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.985660064327213e-08,
+      "loss": 1.9617,
+      "step": 14843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.833437629769843e-08,
+      "loss": 2.1343,
+      "step": 14844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.683575082048419e-08,
+      "loss": 1.976,
+      "step": 14845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5360724282295094e-08,
+      "loss": 1.9433,
+      "step": 14846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3909296752853155e-08,
+      "loss": 1.887,
+      "step": 14847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.248146830060362e-08,
+      "loss": 2.0805,
+      "step": 14848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.107723899299254e-08,
+      "loss": 2.0085,
+      "step": 14849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9696608896300224e-08,
+      "loss": 2.1395,
+      "step": 14850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8339578075641256e-08,
+      "loss": 1.9538,
+      "step": 14851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.700614659507551e-08,
+      "loss": 1.9752,
+      "step": 14852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.569631451760813e-08,
+      "loss": 2.1296,
+      "step": 14853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.441008190502304e-08,
+      "loss": 1.9231,
+      "step": 14854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.314744881799392e-08,
+      "loss": 2.071,
+      "step": 14855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.190841531625077e-08,
+      "loss": 1.9392,
+      "step": 14856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0692981458135815e-08,
+      "loss": 1.9108,
+      "step": 14857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.950114730110309e-08,
+      "loss": 1.9524,
+      "step": 14858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8332912901329888e-08,
+      "loss": 1.9132,
+      "step": 14859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7188278314105308e-08,
+      "loss": 1.8902,
+      "step": 14860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.606724359327517e-08,
+      "loss": 1.9787,
+      "step": 14861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4969808791908132e-08,
+      "loss": 1.9866,
+      "step": 14862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.389597396174059e-08,
+      "loss": 1.8832,
+      "step": 14863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2845739153454227e-08,
+      "loss": 2.0009,
+      "step": 14864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1819104416620496e-08,
+      "loss": 2.1445,
+      "step": 14865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0816069799756142e-08,
+      "loss": 2.0171,
+      "step": 14866
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.983663535010116e-08,
+      "loss": 2.0611,
+      "step": 14867
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.888080111395185e-08,
+      "loss": 2.0784,
+      "step": 14868
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7948567136494286e-08,
+      "loss": 1.8338,
+      "step": 14869
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.703993346158228e-08,
+      "loss": 1.9631,
+      "step": 14870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6154900132236973e-08,
+      "loss": 1.972,
+      "step": 14871
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5293467190147237e-08,
+      "loss": 2.0184,
+      "step": 14872
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4455634676002749e-08,
+      "loss": 2.1416,
+      "step": 14873
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.364140262938296e-08,
+      "loss": 2.04,
+      "step": 14874
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2850771088646074e-08,
+      "loss": 2.112,
+      "step": 14875
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2083740091206608e-08,
+      "loss": 1.8066,
+      "step": 14876
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.134030967320232e-08,
+      "loss": 2.0966,
+      "step": 14877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.062047986977177e-08,
+      "loss": 2.0113,
+      "step": 14878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.924250714832272e-09,
+      "loss": 2.0225,
+      "step": 14879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.251622241301938e-09,
+      "loss": 1.978,
+      "step": 14880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.602594480877635e-09,
+      "loss": 2.0118,
+      "step": 14881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.977167464257029e-09,
+      "loss": 1.7975,
+      "step": 14882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.3753412209165425e-09,
+      "loss": 2.0665,
+      "step": 14883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.797115779277885e-09,
+      "loss": 1.9654,
+      "step": 14884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.242491166652542e-09,
+      "loss": 2.0178,
+      "step": 14885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.711467409241777e-09,
+      "loss": 1.926,
+      "step": 14886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.204044532025609e-09,
+      "loss": 1.8635,
+      "step": 14887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.720222558984855e-09,
+      "loss": 1.9025,
+      "step": 14888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.260001512990108e-09,
+      "loss": 1.8991,
+      "step": 14889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.823381415801741e-09,
+      "loss": 1.8994,
+      "step": 14890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.410362287903368e-09,
+      "loss": 2.0609,
+      "step": 14891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0209441488904255e-09,
+      "loss": 2.0092,
+      "step": 14892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6551270171371044e-09,
+      "loss": 1.9669,
+      "step": 14893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.312910909851862e-09,
+      "loss": 2.1277,
+      "step": 14894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.994295843299465e-09,
+      "loss": 2.1993,
+      "step": 14895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6992818324124138e-09,
+      "loss": 1.9607,
+      "step": 14896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4278688911240067e-09,
+      "loss": 2.1126,
+      "step": 14897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.18005703225732e-09,
+      "loss": 2.0115,
+      "step": 14898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.558462675807179e-10,
+      "loss": 2.0097,
+      "step": 14899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.552366075858075e-10,
+      "loss": 1.7986,
+      "step": 14900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.782280618205071e-10,
+      "loss": 1.96,
+      "step": 14901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.248206385559783e-10,
+      "loss": 2.0947,
+      "step": 14902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9501434511969295e-10,
+      "loss": 1.8099,
+      "step": 14903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8880918756236654e-10,
+      "loss": 1.9367,
+      "step": 14904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0620517093551385e-10,
+      "loss": 2.0442,
+      "step": 14905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.720229912491547e-11,
+      "loss": 2.0212,
+      "step": 14906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1800574906128957e-11,
+      "loss": 1.9711,
+      "step": 14907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 1.9742,
+      "step": 14908
+    },
+    {
+      "epoch": 1.0,
+      "step": 14908,
+      "total_flos": 2550828196347904.0,
+      "train_loss": 2.143045965703077,
+      "train_runtime": 999376.4636,
+      "train_samples_per_second": 1.909,
+      "train_steps_per_second": 0.015
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 14908,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 2000,
+  "total_flos": 2550828196347904.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}